Overclocking is Easy! Get Results! -- v1.50 -- *Updated 10-22-2007* Introduction Overclocking is easy to do, if you know and understand what you're doing. A lot of newly registered members have been skipping past the required steps to gaining proper knowledge. This is a very serious issue when you're talking about the ammount of time and money put into your PC. Learn the basic's, learn how to obtain better information, use this site to your advantage without having to ask a question that's already been answered, and get that answer easily and quickly. It requires your time, it requires your memory, it requires your common sense, if you cannot use any of these or are ignorant, then you should not be overclocking. This guide is not just intended for Overclockers, but about anyone looking into making their PC better, and even just understanding their PC better from case to components. ----------------------------------------------------------------------------------- So You Gotta Question? There have been many new users to TPU (Tech PowerUP!) that have joined for the sole purpose of asking for help with Overclocking their personal computers. But there is a right way and a wrong way of going about this, and that is what I hope to show you if you can bear with me and read through this guide. You will be more confident to ask the right question, provide the right information and know what you are doing. Also, you will find that people are more willing to help you with your OC situation, whether it's getting that extra few MHZ or completely new to Overclocking. Many times when I find someone new asking for help, well here's an example: Does this person even understand what OC/Overclock means? Probably not more than the fact of increasing speed to make it "faster," and has no clue of what other parts are affected by trying to increase the CPU speed. As always, I respond with an inquiry of their TOTAL system specs, brands, model #'s, stock or aftermarket cooling, etc. When asking a question to people on TPU or any other PC Tech/OC Forum, they will need to know what components you have, if you're overclocking, operating system, and sometimes more details depending on the question or issue. Some people cannot go any further cause' they just don't know, and if the don't know they shouldn't Overclock. Overclocking is a science, an art, and some luck. People need to understand that there's research, reading, soaking-in a lot of information and specifications in order to better understand it. If I could get new people to read this first so they can get a better idea of how to post questions and learn what they need to do to overclock, they'd be so much happier with their results that would also happen more quickly. -------------------------------------------------------------------------------------- How To Get That Answer! A better way to ask the demo question above would be (using my personal system as a demo, system as of 5/2007): Can you see the difference? I'm sure even the blind can! But it's that kind of information, that not only ensures those reading your post that you have a clue of what you're doing, but you understand it. The two biggest differences between you and the person(s) helping you is experience and knowlege, but if you can attain the proper knowlege, they can help you with their experience. If you cannot ask a question with these specifications, or close to them, then you have a lot to learn and understand, that's the sole reason behind the creation of this guide. Once you can understand and implement the above information in posts when asking questions, you should be able to Overclock without pulling teeth to relieve your pains and frustrations. Just remember the more you can provide when asking for help, the less that will be asked, which means the fewer posts to get to the solution. And we all want to get things done faster right? Who wants to wade through a thread full of 100 posts because someone couldn't offer the information in the first post or two? Go back up and recompare with the first demo question, and with what you've learned so far, you can see the difference is staggering. Now we could tell this person to attempt to up the CPU voltage, and if cooling fans aren't set for 100% speed to do so as the cooling is already being pushed, it's effective and doing a good job, but based on recommended safety net temperatures for CPU's (STAY BELOW 60C, 55C should be MAX TEMP GOAL), Since this person's at 480FSB, they're probably pushing the envelope of their current settings, so upping FSB/NB/SB voltages by one notch and attempt stability at 485, then 490FSB. If successful, try 495, if successful try 500. If unsuccessful, increase CPU voltage by one more notch, attempt again. And this is a good idea of how the cycle goes! There's a lot of trial and error for every overclock, some have more errors than others, and there are a lot of variables, but with the right knowlege, errors can be overcome. Learning what each BIOS setting affects is EXTREMELY IMPORTANT! Increasing voltage too much can cause severe heat increase and damage in a very short ammount of time. But with tried and true methods that are used at TPU, you can be successful. So you can see where this goes with more knowlege of what one is doing. People are willing to spend a ton of money on stuff they know little about, and ruin it because they don't want to read and learn for themselves. I am sure the second demo question asked is kind of pushing it from someone who knows nothing, but if you know little about what you have, it doesn't hurt to learn about it, read reviews and other people's experience with that product. The more you know, the better off you will be, and that cannot be stressed enough, because it could be the difference between a stable and faster pc, and a short-lived smoking furnace. A little research will go a long ways...there's not a single person on TPU that's a successful Overclocker that can argue that fact. I would like to thank TPU for the great opportunity of teaching me the greater fine tuned art of Overclocking many components and being an amazing online community, there isn't a better place on the web! Remember, it doesn't hurt to ask, but if you're too lazy to research, you will end up having to anyways because we will ask questions that will make you research. It's no different than selecting the type of fuel you need at a gas station, you need to know what kind of fuel your vehicle needs/requires? Check the owner's manual! Usually the Gas Fill Cap and Message near Fuel Gauge also state it, giving you multiple chances given to know what you need to before making the wrong/right decision! Enjoy the rest of the guide as it is full of information for many, whether looking to get better performance, better understanding, or just something to read, we hope you enjoy all that we have compiled for you! Before you overclock 1. Knowlege is power, and will not only be your success but your saving grace when overclocking. 2. Know what your components are (i.e. Brands, CPU, Power Supply, Motherboard, Memory brand/speed/timing (5-5-5-15), Cooling, etc.) 3. Know your system from BIOS to Operating System, learn your BIOS, read your M/B's Owner's Manual!!! It actually has useful information in it! 4. Understand that HEAT is your enemy, and when you overclock you will be beckoning it to destroy your expensive investment, COOLING is extemely important. 4. Learn your components, research, Google is VERY usefull! Learn what each of your component's jobs are, why they're there, why you have it/need it, what the affects are of modifying it. 6. The more you research, the more you can learn about Overclocking your system... 7. Understand the hazards of overclocking, what it can harm, what the wrong settings/adjustments can do, if you learn and understand how to avoid bad situations, destroyed compenents and such, even if that means not meeting your goal, you will be a successful Overclocker. Only the brave and experienced push the envelope. Note: Remember that neither myself nor anyone else including TPU can be responsible for the decisions and actions you make while overclocking and what damage may occur, which is more reason to have a better understanding of what you are doing. -------------------------------------------------------------------------------------- Important Links for Users: Need help? Read this first! TPU Guidelines Zek's PC Building Guide (Excellent Guide) Bruins004's New PC Building Guide Programs/Diagnostics/Stress Tests for Overclockers Quick Step by Step No Boot Troubleshooting Guide OCDB Project Forum - 100s of Different OC'd Rigs! OCDB Project Page Panchoman's Power Supply Guide (Very good information!) eXtreme Power Supply Calculator (Great Utility!) Fox34's Guide to Case Modding Most of these are TPU Based Threads that are full of excellent information and resources to make Overclocking more easily accomplished and understood. Thank the great folks behind all of these threads as they deserve it, they have many, many hours of trial and error to bring you the best information possible. The least that can be done is to have read and learned from their teachings and mistakes, and there are many more problem/solution threads on TPU, you just have to look for them and the search function on this forum is quite convenient, use it to your advantage. The last link, eXtreme Power Supply Calculator, is very handy, you fill out the form and it will tell you how much Wattage your PC requires (has a good estimate algorithm) to run at different loads from ~60% to 100%! Also it will tell you how much wattage the CPU TDP is on Overclock! It has many options and selections and variables, for instance at the right are my System Specs under my name, that system requires about 328 Watts at maximum load. It's a great idea for someone building a new PC and wanting to know how much power they need to provide, I personally believe in having a buffer even if purchasing a good quality PSU, it may help to get a more powerful PSU and not have to worry about it when upgrading to newer more power hungry components. If you have a Laptop Style PC, and want to tweak it, OC it, make it perform better, do yourself a favor and read the excellent information that Theonetruewill, Ketxxx and many, many users have spent hours-upon-hours compiling to give you one of the best Laptop OC Guides out there! Also, Google is going to be your best friend, I have found Google to be the best search engine (as of 6/2007) for finding information that I need whether it's specifications, methods, settings, etc... As always, TPU is great for an insane variety and a plethora of information on damn near any kind of PC system out there, but with mininal provided information by you there will be minimal help for you, and your thread may even go overlooked with ZERO responses. Also reading the sticky threads for Intel/AMD Overclocking is very, very useful as-well-as the OC Programs Downloads Thread which has many different utilities from temperature monitoring (extremely important) to overclocking in the OS to stress testing (also very important to verify stability). Reading these should be the first steps before even posting. --Have a helpful guide/suggestions/comments? Contact me!-- -------------------------------------------------------------------------------------- Researching, Why you need to do it! Probably the most important thing you can do is research what you have or what your are planning on purchasing. Sure you can ask people in these or any other forum if a component will be good, but that's only a partial search for recommendation and should not be considerend the main deciding factor of a purchase. Researching is an easy thing to do, but is time consuming, as I recommend reading at least a couple professional reviews and a handful of consumer statements/reviews. This will tell you a few things about how the component functions, performs, any issues with compatability, etc. If you learn as much as you can, the more confident you will be when the time comes to overclock that component or know it will be stable with other components being overclocked. I will personally spend at least a couple of hours researching the product I am considering purchasing, and it's closest competitors to see if the one I am choosing is the right one for me. I see many people not looking hard enough, jumping on a purchase and having issues with that said component. The truth of the matter is that it is of their own decision that they are having issues/problems for not taking the extra time to ensure that is what they want, that they can make it work with what they have and so on. Like I've said before, Google is one of the best ways to search for information on a product. Also, on occasion Epinions.com may have a few consumer reviews on the product you're researching. There's also Newegg and Tigerdirect which allow reviews of products purchased by consumers, which are also very good to read through, read all of them or as many as you can stand to. The more you can read the better (have I emphasized this enough yet?). If you take the extra time to read what more people thought and what their experiences were on the item/component you're researching, you'll have a better idea if it will be something you want or not. Some people are willing to deal with fewer features, or want to pay a higher price for better quality. But knowing what it has, how well it does, and how certain products perform during OC's can give you an average idea of what it's capable of. Keep this in mind if you want to get the most OC for your dollar. Of course check compatabilities, because even in this day and age, there are different products are incompatable with eachother. Do yourself a favor and know more about what you have or are getting, if wanting to push it's boundries, research until you know all you need to in order to get it to your goal (or close). A lot of questions can be answered before they're asked if people would research a little more, and understand no two items will OC/Perform the exact same, different products even if similar in ratings or even design can net very different results. Even so, it helps to know more, it takes time, but consider it your solution before you ask something that you can easily answer yourself. Research and learning is extrememly important if you like to have a stable, cool-running, and enjoyable PC, the more you do research and learn, the better it will be. -------------------------------------------------------------------------------------- BIOS, What's the big deal? Ahh, BIOS, this is where the Overclocking is done, sure you can do some form of OC-ing in Windows, but it's best if left to BIOS for this procedure. This is where all your important settings are for speeds, timings, voltages, features, etc. This is where you increase settings to attain your OVERCLOCK for your PC. Be patient, take small steps at first, understand what you are doing, see what it changes. Verify your OC with SuperPI or a test to see how much your OC has increased your performance. You will start to see a difference, and that's when it becomes addictive, just don't let it get out of your hands. Take small steps, stay in control, keep on top of your overclock, monitor how things are changing so you can modify if an issue arises. Bios for newcomers can be very intimidating with it's old DOS-like appearance and settings with little to no explanation of what they do or change. Bios is the brain of your motherboard, it allows you to change how your motherboard functions in many different ways and with more overclocker friendly boards this becomes even more intensive as-far-as ammount of settings. I recommend you read your M/B's owner's manual for starters, if that doesn't make things clear enough, GOOGLE or ASK what you're not sure about before changing, find out what others are using/disabling in their BIOS, find out what settings are being used for an application similar to yours. Once you start to use your BIOS, and understand how it works, what it controls, how you can modulate it to your advantage, you will be able to overlcock with more confidence. If you are unsure of a setting, then read up on it. Settings like Intel's power saving TM 1/2, C1E, SpeedStep etc., are no good for overclocking as they modify voltage levels and clock speeds which can cause instability, there are many other features that can be changed and that depends on your motherboard, it's provided Bios, etc. They are good for the regular user, which does not want to overclock or only wants a small overclock which can genrally be provided by MFG presets in Bios (depending on MFG and Motherboard/Bios). There is also issues with updating your BIOS, it is recommended this is not done unless you are having issues and updating is recommended. All BIOS releases have some sort of reference or revision number, so research it and see what others have to say about it, and how their experiences are with it. Some add features, some add stability, some degrade stability, so it helps to have that knowlege before jumping into it. Some people prefer to use an older outdated BIOS because they have found their prime settings with that version and were unsuccessful with newer versions (if your board supports flashing with older bios versions). Flashing bios is the process of updating to a newer version, can be done via operating system on a lot of mother boards now, but is not recommended because if your OS crashes or power is lost, you most likely just ended your motherboard's life, and if still covered under warranty would have to go through the RMA process (see your components MFG and the distributor you purchased the component from for more information on RMA Service). The recommended method is via 3.5" floppy disk or if supported USB Drive upon reboot. Even then you must ensure that the PC's power is not switched during the update (takes a couple of seconds average, might be short, but it's a very important two seconds). Lastly, should a problem arise from an incorrect setting or failed Overclock, you may be required to reset your CMOS. This is generally a procedure that requires a physical change on the motherboard itself, please refer to your Motherboard's Owner's Manual for more information. Genrally you can find CMOS Reset Jumper, it's default position is pins 1 and 2, in order to reset your cmos, you take the jumper off and set it on pin's 2 and 3 for at least 15 seconds (depends on MB, I usually go at least 30 seconds), then remove the jumper, install back to default position, power up your PC and your BIOS will be set back to default, you'll have to set time/date, and all of your settings will be lost. Another method is removing the CMOS battery, which is the small circular/flat shaped watch-style battery on your motherboard (see owner's manual), remove it for at least 30 seconds, then re-install it, it is recommended you use the Jumper method first though. Also, it is not recommended to be in your PC Case and touching your components without grounding yourself and removing the power source (unplug the power cable from the back of the PSU), you can ground yourself by keeping constant contact with the frame of the PC and the PSU, or go fancy and purchase a grounding cable you attach to your wrist. Either way, it pays to be safe, and static electricity WILL do severe damage to electrical components, do not risk it! ----------------------------------------------------------------------------------- Cooling As I've stated a handfull of times before, cooling is extremely important, even for a completely bone stock system. Heat is your enemy, and you must fight this enemy in order to be successful. There is a plethora of different cooling devices out there, and most everything now-a-days has a heatsink...but not always. Learning what to cool is important, and a good way to know is to understand that when you increase something's speed, even without a voltage increase, it will create more heat. Heatsinks are rated to only remove so much of this heat, this is where stock/regular heatsinks start to faulter. They were designed to just get by at stock speeds, so if you feed more heat to them, you will have serious problems. Voltage increases only make matters worse as this increases heat substantially...even a small 0.1v bump up in voltage on a component will generally cause a significant heat increase (generally measured in Celcius). So if you plan on Overclocking, plan on spending more money now to save a lot of money later. What I mean by this is the fact that good cooling can be fairly cheap to come by, which would be the Air/Heatsink method. Similar to stock, but better quality, better efficiency based on design (generally larger than OEM). There are many brands out there such as Zalman, Arctic Cooling, Scythe, Thermaltake and many others who are primarily dedicated to just providing cooling equipment/products for PC components. There are other methods of cooling that are more advanced, efficient, affective and expensive such as Peltier or Water Cooling, and generally if you have a more advanced cooling system you have an understanding of overclocking and have no need for this guide. What all do I NEED to cool? You are probably asking about now, well then keep reading. If you decide to overclock your CPU a healthy ammount, you will need to increased voltage on it, your bus (whether HTT or FSB), your memory voltage since it will generally be faster unless you have a divider to force it to run slower, your Northbridge and Southbridge voltages, etc, so all these components are not only increasing in speed, but are becoming unstable because they do not have enough voltage to do the job..this is where increased voltage comes in to play, and like I stated before, the more voltage the more heat created. It's like when driving a vehicle, the faster you want to go, the more fuel it needs and the more you need to give it via accelerator pedal, and the more airflow is needed for the coolant system to keep the engine from overheating and blowing up. For example my Intel Core2Duo e6300 is rated at ~65W TDP at stock speeds and voltage settings (1.860GHz/1.325VCore), but once I overclock it to 3.360GHz, that increases to 117W TDP, that's almost a 100% increase! If the stock cooler is only good enough to cool a little more than the 65W TDP, then it's going to have serious issues dissapating almost 120W. And that difference is using the same 1.325 vCore, if it's increased to 1.35 vCore, the TDP rating increases to almost 140W! In this TDP rating, the more Watts = the more Heat produced. Just keep this in mind, as it will give you an idea of cooling measures that need to be taken in order to not destroy your investment. So when overclocking, you will want to ensure all of these components are properly cooled and that there is proper airflow to remove hot air and replace it with cooler air (generally ambeint room temp air from intakes). Some motherboard manufacturers have good cooling already implemented, some don't. It is recommended that the Northbridge have at least a decent heatsink installed, the northbridge is kind of a CPU in it's own right and can get very hot. A larger, more efficent heat sink is definately in order for the CPU. RAM these days is generally loaded with heatsinks installed, and there really isn't a whole lot you can do but add active cooling to your memory (Take a look at Corsair's XMS Airflow on Newegg for instance, has good airflow and is very quiet and is good for cooling up to 4 sticks of memory!), but shouldn't be required unless pushing some serious boundries on frequency and voltage. Also when mounting a heatsink to the CPU or chip that needs it, you must use a thermal interface compound such as Arctic Cooling 5 or Arctic Cooling MX-1 to help transfer heat efficiently to the heatsink. Follow recommended instructions for your application. One situation largely overlooked by newcomers to Overclocking is TOTAL CASE AIRFLOW. This is dependent not only on design, but how many fans are installed, how fast they are spinning, what their CFM ratings are. Genrally you want good intake (front/side) fan(s) to suck in cooler air, and good exhaust (rear/top) fan(s) to exhaust hot air out, that is the general PC Case Layout, but cases can vary in shapes, sizes, configurations and airflow specifications. There are quite a few different sizes, but the most popular for cases/PC Towers are 80mm/92mm/120mm. The larger the fan the lower the speed the fan has to spin to create the same ammount of CFM (cubic feet per minute) of airflow, and the higher overall CFM at top rated speeds. Another posotive note is the larger you go, the quieter the fans are, so if you're aiming for a quiet system using air cooling, look for a case with good airflow qualities and spots for 120mm fans as primairy case fans. When looking at different case fans it doesn't hurt to have filters in place for intake fans to also help keep dust to a minimum. Most filters today are made for higher flow, hence will not degrade the overall CFM performance, which means they are now a better solution then they were years ago. I personally have used the Evercool Spider Fan/Filter Combos from Xoxide (I am using 80mm in intakes, I provided 120mm link), as they come with good high-flow filters that do a good job of catching dust particles and keeping airflow restriction a minimum (until the filter gets clogged with dust that is, but they are very easy to clean). New style PC Fan Filters are cheap, easy to install, usually screw into case with fan, recommended in front of fan so the fan sucks air through the filter. If you install the filter so it pushes through the fan you will lose cooling and CFM efficiency and may cause dust and clogging issues. Another issue largely overlooked with Case cooling is Airflow Ratio. This is the ratio of CFM IN to CFM OUT. To keep dust out of your case, but also have the least ammount of effective cooling is to have More CFM IN than OUT. The most effective and most dust collecting method is to have More CFM OUT than IN, this will cause air to be sucked in through your case's seams, front panel gaps, side vents to compensate for the lesser CFM input from intake fans, which in turn means cleaning out your case more often. The Preffered method is to have a 1:1 ratio, or Same CFM IN as OUT, this creates a more equal airflow and has been argued to be more effective than More CFM IN than OUT, this method will still allow the side vents to be used (if applicable), and is also efficient at exhausting the hot air and promptly replacing it with ambient (room temperature) cooler air. This of course is more of a challenge unless you replace fans with ones you know have certain CFM specs that are matched, and taking into affect that there are Video coolers that exhaust (Average is approx. 15cfm) and power supplies (generally 15-25cfm). Like I stated earlier, for a CPU, 55-60C is about as hot as you should let your CPU get, if you surpass this, then you need to turn it down a bit or improve your cooling. GPU's are a different story as they can take up to 80C+ (depending on GPU model/design), but that heat will also radiate and heat up the ambient temperature of your case, thus heating your other components more, so good cooling for your Video card should also be seriously considered. Another commonly missed area is your Hard Disk Drive, it is very important to keep your HDD cool, the cooler you can keep it, the longer it will last, to a certain extent. That rule applies to about all electronics. HDD's are often overlooked, but simple placement can keep your HDD cool enough to have no worries about. I try to keep my HDD temps in the 30s, generally 35C or less if possible, most HDD's are rated to hit a max temp of ~50C though. If you are unable to keep your HDD cool, then purchase an HDD cooler, they're simple to install and very effective, and generally cheap, like I stated though, simple placement (i.e. bottom mount location in front of lower part of intake fans) can solve this with no more money spent. One final note on cooling is monitoring those temps! There are quite a few programs out there to use for temperature readings. I personally use Everest and Speedfan, but try them all and find one that you're comfortable with. They are not completely accurate, but will give you a good idea of what kind of temperatures you are at for different components, and some require adjusting the readings +/- C/F in order to be more accurate. A simple Google search of your Motherboard and including the selected cooling program will generally tell you if you can just install and go, or if you need to make a simple adjustment. It is very important when overclocking to monitor your temperatures in order to not overheat your system and burn it up, and to ensure you are properly cooling your system when under load, verify your idle/load temperatures, and lastly peace of mind! Only a fool will not monitor the temperatures of their system while overclocking, and that is a fact. So do yourself a favor and get used to having a program open and monitoring, it will help you learn how different loads and fan speeds can change your temperature readings, and even experiment to find the best mix of cooling/quietness you can while staying within recommended boundries! ----------------------------------------------------------------------------------- Voltage and Temperatures, Cooling Part 2 I touched some on voltage and temperatures before, and will even after this section, as both are vital. You cannot have your PC without voltage, it will create heat when it has voltage. So one tip with overclocking once you reach your goal is to ensure you're running the lowest voltage possible and be completely stable. Below this section is the stability section to help you ensure you can. The problem with increasing voltage to maintain a stable overclock is the ammount of heat you're creating. And since I touched on case fan/airflow dynamics some, and basic air/fan/heatsink cooling, you should be getting a good idea of what you will need to do in order to enjoy the performance of your PC overclocked with no worries. Or that's what I'm hoping for at this point. There are programs out there that read voltages, I personally use Everest for voltage readouts, it is fairly accurate and easy to work with. But you will want to monitor voltages, especially vCore voltage. There is a common issue on Intel Spec motherboards for running their processors that is called vDroop. vDroop comes in two parts, the first the drop in voltage from set parameter (reads 1.29v when set to 1.32v for example), the second is the difference in voltage between an idling CPU and load stressed CPU (reads 1.29 idle and 1.26 load when set to 1.32 in BIOS, for example). AMD supporting boards do not have this issue as noticable as Intel. This can create instability with overclocks if that voltage is just below the point of stability under load, and there are "mods" one can do to their boards to alleviate this, but this article will not dive into that subject, just Google your motherboard with mods or vDroop mods at the end. This could cause you to increase your voltage a step or two up and while you will now be stable, you will also be running quite a bit warmer. Keep an eye on it. Same goes for anything which you increase the voltage of, the North Bridge, South Bridge, GPU, CPU, RAM, GDDR Memory, etc. Also, better quality Power Supply Units will be more efficient and stable at producing power, which can help greatly if you notice voltages changing a lot, as lower quality PSU's are the cause of many problems, crashes, failed/RMA'd equipment. I was once using a cheap $20 500W PSU, one of the cooling fans seized (it was 6 months old!), and the unit overheated and the capacitors started leaking fluid, luckily it missed my motherboard, and did not damage my video card (but was a pain to clean!) because the system shut off once the capacators gave out. It does pay in the long run to pay a little more now and save a lot later, just remember that when building a PC. Voltage increase is also an enemy to components because the more voltage you add to the circuits, the quicker they will degrade. This is a known affect of circiutry and voltage, and there's nothing to stop it from happening, but if you are careful with increasing your voltage and not just setting them all on the highest level, you will have a good PC for years and not have to worry. But every increase of voltage you do in BIOS, the lower the life expectency of circutry/components becomes. Those who upgrade every 6 months or so may have no issues running max volts, adding more via Volt Mods because it's no big deal to them if that component lasts, but if you want a good OC'd dependable PC for years to come until you're ready to upgrade, then please increase voltage carefully, if you increase a lot, please do it for OC/Stability reasons for short term until you find your lowest voltage settings for your overclock. Just do yourself a favor and have a piece of paper or notepad handy, write some of your settings down, so you can use them for reference. Besides physically writing down can help when you're not quite sure what you had set before, as nobody has a perfect memory, it does help to have a reference to go back to. And verify that the voltages you are using to maintain your Overclocked PC is within safe temperature boundries, and are as low and stable as possible. ----------------------------------------------------------------------------------- Stability, and How Exactly do I Know I'm Stable? Stability is easy to verify. First off, if your PC reboots itself or you get the dreaded BSOD (Blue Screen Of Death), your pc crashes/errors, you have an issue. This may not necessarily be hardware/component related, could be OS or Software also, but genrally it's a good mix of all of them or improper settings, overheating. Below are some steps you can take to verify temperatures, that your PC is capable of taking heavly stress loads without faulting, and give you peace of mind that you built your PC right, have proper airflow, and have a good overclock that you can run without issues. I recommend using Prime95 or Orthos for CPU Stressing, out of those two I use Orthos (can stress dual cores quite well). Just open and click start, let it run for as long as you can bear to be away from your PC. The longer the better, but if you can successfully pass 6-8 hours, you're generally considered stable. When using Orthos, set it to Small FFT's (CPU Only), and set the Priority Tab to 9. I'd say if you can pass 24 hours, you're very stable. This will give you load temperatures, so you can test your cooling. If the test fails, then you need to make more adjustments in BIOS. Generally when the CPU stress fails you are not giving your CPU enough power to get the job done. That's when you need to decide whether to downclock your CPU or increase your voltage, when the time comes to make certain decisions, be as prepared as you can be. There's also stress tests for graphics, which is very handy when overclocking your video card. Which can be very tricky or very easy depending on the card and how easy the MFG decided to make it. If you get artifacts, then increase voltage if possible, or downlcock a few MHZ until there is no longer any artifacting on the screen (glitches, yellow pixels, etc). On AGP it helps to increase the AGPv and AGP Bus by a couple of MHZ, but on PCIe there has been no result to show proof enough to increase the PCIe Bus Speed from 100, some do it for stability and that is okay, but stay below 120MHZ to avoid SATA and other issues, or at least this has been the popular rule for a while between many professionals. Many have increased the PCIe Bus to 130+ and been able to enjoy the benefits from it, but this is not well documented as of yet, and until I can provide you with more solid information, I still recommend staying below 120MHz to avoid any problems, keep checking back to see if I've updated to verify this. Increasing PCIe Voltages has also not shown much for stability since there's already extra power being fed to the card (at least mid-level to higher end cards that suck down more than 75W that the PCIe bus provides) to compensate when the GPU is under load. --Side note on PCI-e Bus Speeds: I have been able to contact a few people who've increased their PCI-e bus speeds beyond the popularly recommended max 120. Two of them were using IDE drives and were able to use full 150MHz PCI-e bus speeds with no issues, and were able to attain what they claimed as better overclocks on their video cards. Another person whom is known and respected on TPU (D44ve), was using SATA HDD(s), an 8800GTS, ASUS Striker Extreme M/B and attained better performance and had great stability using 145MHz PCI-e Bus Speed. Remember that the people I talked to had newer components, which could be the difference, especially with video adapters. Thus far it's been NVidia's newer 8XXX Series cards. Newer video cards may be able to use the increased Bus Speed more efficiently than the first wave or two of PCI-e based graphics cards. But this does shine a light on an overlooked performance increase for many building/using new or newer systems. If anyone that reads this has been fortunate in their experiences with increasing PCI-e bus speeds beyond 120MHz and had no issues, please contact me, give me all the information you can as I want the most valid information I can attain before I can safely recommend whether or not increasing beyond 120MHz is worth it. Another important stress test to perform is memory stability. I recommend Memtest, either the boot-DOS version or version for Windows OS. Allow this to run as long as possible again, the more passes the better verifacation. If you have an error, this can be more complicated, This could be tied with undervoltage, incorrect timings, malfunctioning DIMM Slot or a bad chip on the RAM Stick. If you are getting BSOD's and crashing issues that cannot be explained this is a good place to start since RAM is a strongpoint for information transfer in your system. Do not overlook this when overclocking or you will pay the price. Also, it is helpful to read up on timings, the different settings/variables and meanings of the memory timings, it get's complicated so it's recommended to learn the basic timings, as they help with stability, bandwidth (throughput), latency (response), and overall system performance. There are many different programs and methods for testing and verifying stability, the ones I listed I personally use and so do many others. Some Intel users have the benefit of using a program called Intel TAT, it will stress the CPU higher than any other program out there, there is no other way you could stress your CPU as much. Some argue that since you get your load to 100% you are, but check your temps, and remember what that 100% load is doing, your CPU can 100% load on easier tasks, TAT uses harder and more stressing tasks to perform a severe stress. Use at your own risk! ----------------------------------------------------------------------------------- Bottleneck? What is that and why should I care? This can be the cause of much unexpected anger from many individuals who believe they understand overclocking and it's performance benefits but do not reap the entire rewards for their hard work and efforts. Bottleneck is just as the word says it is, take a beer bottle (like a Budwieser bottle for instance), and at the top it's narrow, at the base it's wide. Now to implement this into your PC, you have massive ammounts of information transactions being made, some busses are wider than others, some have to share, but they all have to travel through some of the same "intersections" to get to where they're going. This is where having good quality componenets can reap serious rewards for those who pay attantion to this. When stock, sure speeds are slower, which in turns lowers scores in tests/benchmarks, and as speed's increased scores increase. Fair enough. But when you see someone with a similar system as you with something slightly different, such as a different processor (let's say a newer processor) getting higher scores even though you're overclocked close to that person, that newer processor (try P4 Netburst vs. Core2Duo) is much more efficient and making more transactions than your "older" processor. To match the performance you would have to overclock substantial ammounts that watercooling may not even be able to handle. All systems have bottlenecks, it's part of how motherboards and components are designed. Imagine hauling ass down a highway, to your job in the morning in the city, you hit a downtown intersection, you have 1000's of others trying to do the same as you, but due to an older inefficent intersection design and only few lanes to spread out on this is considered a bottleneck situation (i.e.gridlock, slow moving traffic, stuttering, slowing down). The best way to alleviate this is with newer equipment such as motherboards, cpu's, even RAM that's faster and can attain tighter timings per clock cycle and process information more effieciently by design.A lot of people are still using the AGP graphics interface, and as such the video graphics mfg's are still releasing some new stuff their way. But depending on just how "old" that system is, they may see less performance as compared to a newer AGP based system. Such as an older 2.4 Pentium 4 Socket 478 system with DDR1 2700 with AGP 4x as compared to a newer dual core Pentium 920 Socket 775 system with DDR2 5300 with AGP 8x. There will be only so much either system can do to provide as-far-as bandwidth for the video card, which will take as much as it can get. The more that can be provided, the better performance the card will display. Bottlenecking can give user's serious headaches, but a few simple tests such as 3dMark06/05/03 and even Aquamark 3 can give a good idea of bottlenecking, since generally when referring to bottlenecking people refer to Video Graphics Adapters. Run a 3dMark of your choice and go online and compare it to similar systems, you'll be able to see what others have scored with similar and slightly different and very different setups, it's a good reference point that many miss or misunderstand. And one of the easiest ways to alleviate bottlenecking is to overclock system bus speeds, memory speeds, cpu speeds, gpu speeds, etc., even on new systems this helps a lot performance wise. So do not get yourself down if you have an older system (and if it's capable of OC-ing) and your gaming isn't what you expected, you may need to overclock your system to attain better performance. Technology Explained! Overclocking.. once upon a time it was a skill reserved for the elite, now everybody wants to give it a shot with the advent of "overclocker friendly" programs bundled with most modern motherboards. Make no mistake, with these "overclocker friendly" programs or not, its just as easy to fry your hardware. Now, with that minor introduction aside, I am Ket, your god, will, and to whom you shall owe all hommage to and worship I haven’t read all of the 25 pages or so of text.. its like a novel, but from what I have scan read all the fundamentals have been covered. So instead I’m going to focus more on how to better understand your hardware, and how it interacts with the rest of your system. So without further ado lets get started. An Introduction To The AGP (Accelerated Graphics Port) Bus Originally designed for the Pentium 2 platform, the AGP bus has undergone a multitude of evolutionary steps beginning with just AGP 1x (266MB\s) and ended with 8x (2.1GB\s) using the PCI specification as an operational baseline, the AGP specification adds 20 additional signals not included in the PCI bus. The aim of the AGP bus is to provide a much smoother framerate and 3D rendered image in a much higher detail level than previously seen before on the PCI or ISA bus standards of old. The AGP bus is able to transfer high amounts of data due to it being able to transfer on both the rising and falling edges of the 66MHz bus clock frequency. AGP allows for direct data transfer between the graphics card and the CPU and \ or system memory. Although AGP is an extension of the PCI interface, it is completely (physically, logically and electrically) separate from the PCI bus. Therefore, activity of PCI peripherals won't affect the AGP card's performance. AGP Technical Sheet: Maximum transfer rates (32bit): 266MB\s (1x) 533MB\s (2x) 1066MB\s (4x) and 2100MB\s (8x) Operating Frequency: 66MHz Pipelined Requests Address \ Data de-multiplexed Single target, Single master Memory Read \ Write only, no other I/O operations High \ Low priority queues Sideband Addressing Fastwrites Explanation of DIME Direct Memory Execute (DIME) is probably the most important feature of the accelerated graphics port. AGP graphic chips have the capability to access main memory directly for the complex operation of texture mapping. AGP provides the graphics card with two methods of directly accessing texture maps in system memory: pipelining and sideband addressing. In pipelining, AGP makes multiple requests for data during a bus or memory access. Sideband addressing offers the highest level of AGP performance. In addition to allowing multiple outstanding transactions and non-coherent access to main memory, Sideband Addressing introduces a separate address \ command bus, the Sideband Address Port (SBA). Because the SBA and data buses are not multiplexed, the graphic controller can use the SBA to make data requests without interrupting the data bus. An Example Of Uninterrupted Data AGP revision history During the lifetime of the AGP bus the port underwent a multitude of evolutionary steps forward as briefly mentioned earlier, here is a rundown of the revision history. AGP 1.0 (1996) specification defined 1x and 2x speeds with the 3.3v keyed connector. AGP 2.0 followed shortly after (1998), the specification defined 1x, 2x and 4x speeds with the 3.3v, or 1.5v keyed connector or a 'Universal' connector which supported both card types. AGP Pro specification defined 1x, 2x and 4x speeds with the 3.3v, or 1.5v keyed connector or a 'Universal' connector which supported both card types. The AGP 3.0 specification followed 4 years later (2002) and was the final evolutionary step in AGPs life, the specification defined 1x, 2x, 4x and 8x speeds with the 1.5v keyed connector, or a 1.5v AGP Universal / Pro connector. Each upgrade was a supper-set of the 1x mode, (4x would also support the 1x speed etc.) The base clock rate is 66MHz, but to achieve 2x, 4x, and 8x speeds the clock is doubled each time. AGP (1x): 66MHz clock, 8 bytes/clock, Bandwidth: 266MB/s [3.3V or 1.5V signal swing] AGP 2x: 133MHz clock, 8 bytes/clock, Bandwidth: 533MB/s [3.3V or 1.5V signal swing] AGP 4x: 266MHz clock, 16 bytes/clock, Bandwidth: 1066MB/s [1.5V signal swing] AGP 8x: 533MHz clock, 32 bytes/clock, Bandwidth: 2.1GB/s [0.8V signal swing], still uses 1.5 volt mainboard power. The AGP data bus may be 8, 16, 24, 32, or 64 bits. Due to timing requirements the maximum bus length is 9". The trace impedance is specified as 65 ohms +/- 15 ohms (no termination resistor is specified). For the 8x speed the bus requires a parallel termination or 50 ohms. Some lines may require a Pull-Up Resistor to insure the lines come out of reset in the proper state. A Pull-Up Resistor Technology Explained! Cont. An Introduction To PCI-E: The AGP Bus Replacement Like Hypertransport technology, PCI-E is a 2-way serial connection that carries data in packets, similar to the way it is transferred over Ethernet connections. The PCI-E bus is an assembly of serial, point-to-point wired, individually clocked “lanes”, each consisting of two pairs of data lines carrying data upstream and downstream. Each individual “lane” is capable of 250MB\s, giving a PCI-E x16 slot a theoretical maximum of 4GB\s bandwidth. Real performance benefits come in when more than one lane is added to a given point-to-point route. Lanes can be stacked together to increase the amount of bandwidth available to specific areas of the I/O system, such as the video card slot. Using 4 routes instead of 32 for a basic connection significantly cuts the costs of producing motherboards, since each lane is exclusively used for communication between two points, there is no sharing of the available bandwidth either. Introduction To SLi (Scalable Link Interface) SLi technology allows two PCI-E x16 slots to split 16 lanes into two sets of 8 and run two identical video cards in tandem. This often leads to dramatic boosts in 3D performance. Latest chipsets such as the nVidia nForce4x16 are capable of 2x16 lanes for graphics instead of 2x8 lanes, though performance increases are minimal and arguably well within the margin of error. The current standard PCI Express implementation allows for up to 40 lanes in total. 32 of these are used for the PCI-E x16 graphics slots and eight more are distributed between any combination of PCI-E x1, x2 or x4 slots. PCI Express X1, X2, X4, X8, X16 Connections PCI Express connectors are similar in appearance and connection method to 32-bit PCI slots. PCI Express 1X slots are about the size of current modem riser slots (about 1" long), while the X16 interface (164-pins) for graphics is very similar in appearance to the standard AGP port. The flexibility to adapt to PCI Express devices of different bandwidths is built into the X4 and X8 slots. The Technology Behind The Technology The physical communication method is Low Voltage Differential Signaling (LVDS). Differential Signaling (DS) uses a wire pair to represent logic levels as opposed to the single ended (SE) methods used by legacy hardware. The reason for using DS is that of signal integrity at very high baud rates (level changes per second). LVDS Integrated circuit At a very high Baud rate, our intended signal (bottom) would be transformed into the top image: The reasons for this are mainly capacitance between tracks (which opposes a rapid change in voltage) and slew rate (essentially how quickly the driving amplifiers can change voltage levels). With DS this has much less impact. It doesn’t matter if the voltage levels are reduced in amplitude (within reason), as long as there is a difference for the difference amplifier to work with. With DS the intended signal (bottom) is converted into the DS (middle) and transmitted along the wire pair into the difference amplifier at the other end (top). The difference amplifier then recovers the original signal by comparing the two voltages. DS also offers a immunity to interference, which SE transmission does not. Suppose in transmission electromagnetic interference (EMI) caused a large voltage spike in the DS. Because the wire pairs always run in parallel very close to each other the spike will be present in both wires. Again (within reason) the signal can be recovered. This more advanced signaling is needed due to the very high transmission frequencies necessary in replacing a large parallel interface with a narrow serial interface, for example AGP to PCI-E. To reduce inter symbol interference at the high signaling rates a method called 8B/10B is used to encode the serialized data on each lane. This represents 8-bits as 10-bits to provide a net DC ‘0v’ on the communication line. As 10-bits are used to transmit 8-bits (1-byte) of data the 2.5GBaud provides a raw throughput of 250MBps per lane, however not all of this can be used to transmit useful data. Away from the physical side PCI-E is a packet based interface, much more like USB than it is to traditional “wide bus” PCI and AGP. This means that control information in the form of headers must be transmitted at the transaction layer in addition to frame, sequencing and error checking in the 2 layers preceding. These are not set values so the throughput will depend on the commands being issued or the tasks performed. True data efficiency probably lies in the region of 80% or so which would give 200MB/s throughput per lane. Beyond the 3 layers listed PCI-E behaves like traditional PCI with regards to Windows \ enumeration etc. Topology Of PCI-E PCI-E cards are connected in such a way that each card has a point-to-point link with a switch as opposed to the bus arbitration required by PCI devices. Bus arbitrations based system: Point to point link system: In this case the only advantage of PCI-E over AGP is that of data throughput (3200MBps vs. 2100MBps) as AGP is a point to point link anyway. If a graphics card is being used for less conventional uses (maybe video capture?) the equal downstream throughput should provide a large advantage over AGP. For PCI devices, especially those requiring frequent bus accesses, the gains in latency will provide a great advantage. PCI-E Clock Frequency PCI-E has a base frequency of 100MHz, this is fed into a clock multiplier which increases the frequency by an effective 25x using one (or possibly 2) PLL. For example if the base clock is increased to 120MHz the differential signaling rate is now 3GB\s. The line coding used (8B/10B) should allow for clock recovery. Due to the limiting factors such as line capacitance and slew rate, the only way to increase the maximum PCI-E frequency is to increase the voltage output of the driver IC’s, even then the de-serialiser at the other end may be/still be the limiting factor.The following diagram shows a future (or even current?) two stage clocking system: Technology Explained! Cont. Description of the Peripheral Component Interconnect “PCI” Bus The Peripheral Component Interface (PCI, sometimes referred to as Protocol Control Information) Bus was originally developed as a local bus expansion for the PC ISA, (Industry Standard Architecture) bus by Intel, and was coined the PCI Local Bus. The spec started as an add-on to the ISA form factor with the PCI requiring its own connectors. The PCI spec defines the Electrical requirements for the interface. No bus terminations are specified, the bus relies on signal reflection to achieve level threshold. The first version of the PCI bus ran at 33MHz with a 32 bit bus (133MBps), the current version runs at 66MHz with a 64 bit bus, although the 66MHz PCI bus is not to be found on standard motherboards, its commonly found on workstations. The PCI bus operates either synchronously or asynchronously with the motherboard bus rate. While operating asynchronously the bus will operate at any frequency from 66MHz down to, and including, 0MHz. Flow control is added to allow the bus to operate with slower devices on the bus, allowing the bus to operate at their speed. PCI is an un-terminated bus, the signals rely on signal reflections to attain their final value. The PCI specification has also been ported to a number of other form factors, including: PCI: The original specification 'Peripheral Component Interface', @ Rev 2.2 PCI-X: The latest version 64 bits at 133MHz cPCI, Compact PCI: PCI in a VME form factor, 3U/6U using 2mm connectors PC104-Plus: PCI add-on to the PC104 spec, ISA in a square form factor PISA: PCI add-on with PCAT to the ISA AT form factor P2CI: PCI on the VME64 P2 connector PMC: PCI on a Mezzanine Card, 'PMC' PXI: cPCI for Instrumentation IPCI: Industrial PCI (Another version of cPCI} Serial PCI PCI on a serial link Card Bus: 32 bit PCI on the PC Card (PCMCIA) Format Each of these additional specifications rely on the PCI spec., normally only the mechanical form factor definition changes. Unlike earlier PC buses, the PCI bus is processor independent. The 64 bit PCI bus is made up of the following primary signals: Address/Data Bus: 64bit Address; 64bit Data, Time Multiplexed System Bus: 2bits; Clock/Reset Interface Control Bus: 7bits; Ready, Acknowledge, Stop. Parity Bus: 2 bits, 1 for the 32 LSBs and 1 for the 32 MSB bits Errors Bus: 2 bits, 1 for Parity and 1 for System Command/Byte Enable: 8 bits (0-3 @ 32bit, and 4-7@ 64bit Bus) 64MHz Control: 6 bits; (2) Enable/Running, (2) Present, (2) Ack/Req Cache: 2 Bits Interrupt bus: 4 bits JTAG Bus: 5 bits Power: +5, +3.3, +12, -12v, GND The Time Multiplexed Address and Data bus may exist as either 0 to 31 bits (32-bit) or 0 to 63 bits (64-bit) using the 64-bit expansion bus. Both the Address and Data line use the same bus, 32-bit PCI may also use 64-bit addressing by using two address cycles; termed Dual Address Cycles (DAC), the low order address is sent first. Additional control bits are utilized once the bus is increased to 64-bits. The specification defines both a Reset line and a Clock line. The Clock may be either 33MHz or 66MHz. A number of “Handshake” lines exist to allow communication, i.e. Ready, and Acknowledge Two Parity lines are made available, one for the 32-bit bus width (bits 0 to 31) and an additional one for the 64-bit expansion (bits 32 to 63). and two error bits; I assume, 1 for the LSB 32-bits and one for the upper 32-bits. PCI cards for a personal computer differ from the ISA type by two important factors: - Components mounted on the reverse side of the card - Edge connector is more dense, shorter and the keys reside in different locations Memory Command mode differences between PCI 2.0 & 2.1 AD1 AD0 2.0 Burst Order 2.1 Burst Order 0 0 Linear Incrementing Linear Incrementing 0 1 Cacheline Toggle Mode Reserved 1 0 Reserved Cacheline Wrap Mode 1 1 Reserved Reserved Toggle addressing used in V2.0 is similar to Intel 486 and Pentium class CPU toggle mode addressing. Rev 2.1 employs cacheline wrap, the data accesses proceed by incrementing linearly to the end of the chaceline, followed by it wrapping around and continuing to transfer data from the beginning until it fills the entire cacheline. Transaction ordering was introduced in PCI rev 2.1 to accomplish the following goals; 1. To satisfy write results ordering requirements of producer-consumer model. 2. To allow some transactions to be posted for improved performance. 3. To prevent bus deadlock when the posting buffers must be flushed to meet write results requirements. Target Initiated Termination This section of the PCI specification was expanded to clarify cases of target initiated termination, target initiated termination without data that was defined as disconnect-c in rev 2.0 is now superseded by two uniquely identifiable cases: disconnect-1 and disconnect-2. Delayed Transactions Some PCI devices may not be able to meet latency requirements, so the PCI rev 2.1 specification introduces the concept of delayed transactions by those targets that are unable to complete the initial data phase within the required clock cycles. Through the use of delayed transactions the PCI bus is not held in a wait state while a master waits for a slow device. Implementing a delayed transaction To implement a delayed transaction the target must terminate the transaction with a retry signal once the request information has been latched. The master is unaware of the reason that the target terminated the transaction and is required to initiate the transaction again, in the meantime another master may be allowed to use the bus for an independent transaction. Arbitration Both rev 2.0 and 2.1 of the PCI specification allows any bus arbitration algorithm to be implemented by the central PCI bus arbiter, as long as only one GNT# line is asserted on any clock. Rev 2.1 now requires this algorithm to be fair, a fairness algorithm allows low-priority masters access to the bus while higher priority masters are continually requesting it. Latency PCI Rev 2.0 recommends that targets be able to complete the initial data phase within 16 clock cycles of the initiation of the transaction. It also recommends that the master be able to assert IRDY# within 8 clock cycles of all data phases, both of these recommendations are requirements of PCI compliant devices. Three possibilities have been identified on pre PCI rev 2.1 in respect to meeting new latency rules; 1. The possibility that a device can meet the initial latency requirement of 16 clocks, most I\O controllers developed before rev 2.1 will be able to meet this requirement. 2. The possibility that a device can usually meet the initial latency value but sometimes can’t, these devices should use the retry target initiated termination so the request can be repeated. 3. The third possibility is that the device usually can’t meet the initial latency requirement, these devices are required to implement delayed transaction.All PCI devices have two exceptions to the latency requirements, both occur during initialisation and don’t have upper bound or initial latency. - POST code accessing the device configuration registers - POST code copying the initial expansion ROM image to memory Final exception to the initial latency rule applies to host bridges, these bridges are allowed an additional 16 clocks (32 in total) to complete the initial data phase if the access hits a modified cacheline. (above) PCI power, clock & reset timing. Address \ Data Stepping Stepping a signal allows devices with a weak output buffer to assert a set of signals over several clock cycles to reduce buffer generated ground current. Similarly a device with strong output buffers can assert a subset of these over several cycles to reduce the number of signals that must be switched simultaneously. Vital Product Data (VPD) VPD is information that identifies aspects of a systems hardware, microcode and software elements, it also provides the system with a limited method in which to monitor aspects of a device for performance and failure rate measurements, this information is particularly useful with plug and play devices. Technology Explained! Cont. PCI v2.3 The latest PCI revision remains largely unchanged from rev 2.2, the only real noteworthy differences are rev 2.3 removes support for 5v PCI devices and adds a two wire management interface for the SMbus. Mainboard Chipset Explanation For this explanation I will use the nVidia nForce4 SLi chipset AMD Edition. Like the nForce3, the nForce4 is a single sillicon chip design, as opposed to the typical North and South bridge chipset design. One reason this has been possible to achieve is because of AMDs innovative CPU design, placing the memory controller in the CPU instead of relying on a separate chip placed on the mainboard, this design holds numerous advantages such as faster data communication and fetching between the CPU and RAM (Random Access Memory). Diagram with Memory Controller on mainboard. A better explanation would be to consider a car journey, with Point A being the CPU, Point B the separate chip on the mainboard and Point D being the RAM. More conventional designs would require the necessary data to be passed along all “Points” in order for the requested data to reach its source, but by placing the memory controller directly in the CPU eliminates Point B, instead having the data pass directly from the CPU to RAM and back again over the CPUs FSB (Front Side Bus) significantly reducing request time, although this also has its disadvantages, as traces need to remain quite short and close to eachother. Diagram with Memory Controller integrated into the CPU Block Diagram of NF4 Chipset The HyperTransport Link HyperTransport interconnect technology provides up to a maximum bandwidth of 22.4GB\s and typically will run at 400MHz for I\O (Input\Output) connections, or 1GHz (1000MHz) for CPU-to-CPU links, and is double pumped, meaning data is sent at the falling and rising edges of the clock giving a maximum speed of 2Ghz (2000MHz). A 8-bit I\O link provides 800MB\s bandwidth each way and 1.6GB\s aggregate bi-directional (both ways) bandwidth. 16-bit CPU-to-CPU links provide 3.2GB\s each way, for a maximum bi-directional bandwidth of 6.4GB\s. The HyperTransport bus has been designed as a high-performance, high-speed, high-bandwidth, point-to-point link to provide the lowest possible latency, increase the communication speed between integrated circuits in computers, servers, embedded systems, supercomputers, and networking and telecommunications equipment up to 48 times faster than some existing technologies. HyperTransport technology also helps reduce the number of buses in a system, which can reduce system bottlenecks and enable today's faster microprocessors to use system memory more efficiently in high-end systems. Hyper Transport technology is designed to: * Provide significantly more bandwidth than current technologies * Use low-latency responses and low pin counts * Maintain compatibility with legacy PC buses while being extensible to new SNA (Systems Network Architecture) buses. * Appear transparent to operating systems and offer little impact on peripheral drivers. One of HyperTransport's low latency capabilities is the low data packet overhead. Compared to other packet-based approaches, HyperTransport provides the lowest packet header overhead: 8 bytes for a read operation and 12 bytes for a write operation (for a read request, there is an 8-byte Read Request control packet followed by the data packet; for a write request, there is an 8-byte Write Request control packet followed by a 4-byte Read Response control packet, followed by the data packet). This compares very favourably to PCI Express as shown below. HyperTransport Priority Request Interleaving Another aspect of Hyper Transport's low latency is a native mechanism, Priority Request Interleaving (PRI), this enables a high PRI command (8-bytes long) to be inserted within a potentially long, lower priority data transfer. While transfer 1 is underway between peripheral B and the host, the need arises for peripheral A to start a data transfer from the host. Transfer 2 without PRI would have to wait until transfer 1 completes and, should transfer 1 be the answer to a cache miss, for example, latency for transfer 2 would become prohibitive. A control packet is promptly inserted with PRI within transfer 1's data stream, instructing the link to initiate data transfer 2 on the other link channel concurrently with the completion of data transfer 1. This mechanism greatly reduces latency and is unique to HT technology. HyperTransport Multi-Processor Interconnect Below is a diagram where AMD depict a multiprocessor architecture for their 64-bit codename Hammer CPU chip that use a HyperTransport-based, high-speed interconnects between multiple processors (2 to 8 processors in glueless SMP (Symmetric Multi-Processing) arrangement. In the four-CPU diagram below, a crossbar switch exists to provide interconnection between the memory controller, HyperTransport interface(s), and a System Request Queue (SRQ) close to each CPU core. The SRQ appears to buffer data and I/O requests to \ from the core and the external world, including system memory and other processors (for cache coherency probes and data transactions). Is there any benefit to running the HyperTransport Bus over its specified maximum? Testing conducted shows there are marginal performance gains in running the HTT Bus over its specified maximum of 1GHz (2Ghz, if you consider the rising and falling implementation on the clock), for example, when conducting HD Tach performance tests, on my system, with a 1GHz HTT Bus, returned a constant result of 138.8MB\s burst result, where raising the HTT Bus to 1.25GHz (again, 2.5GHz considering the rising and falling implementation on the clock) returned a constant HD Tach result of 140.2MB\s. However when running tests which were more CPU dependent such as 3DMark2001SE, running the default 1GHz HTT Bus, returned a more or less (taking into consideration a margin of error ratio) consistent score of 28,248, but when increasing the HTT Bus to 1.25GHz, (again taking into consideration a ratio for error) returned a score of 28,567. So although there are undoubtedly performance improvements, they are marginal and the decision to raise the HTT Bus beyond spec ultimately remains at the users discretion. Introduction to DDR SDRAM (Double Data Rate Synchronous Dynamic Random Access Memory) DDR memory is commonplace in any of todays PCs or laptops, the soul purpose of DDR was to replace the aging DIMM and SIMM (Dual \ Single Inline Memory Module) with something that was capable of operating faster and providing more bandwidth, the outcome was DDR. DDR works on a fairly common principle in todays computers, allowing data transmission on the falling and rising edge of the bus, in effect almost doubling the available bandwidth, without having to worry about various problems such as timing skew that additional data lines would introduce, some examples of system busses transmitting data on the falling and rising edge of the clock signal include the Front Side Bus, Accelerated Graphics Port, and the HyperTransport link on AMDs Athlon 64. Timing skew is a problem that can occur in varying system busses, when signals are transmitted across parallel paths, they will not arrive at the exact same time due to variations in wire transmission properties and transistor sizing, which is unavoidable. If the skew is large enough, its possible that the clock signal may arrive while the data signal is still in transition between previous and current values, if this happens it will result in a functional error, making it impossible to determine what value was transmitted from the detected value. The raw bandwidth of DDR is impressive, although because of latencies and bus protocol overhead, maximum bandwidth efficiency can never be reached, the protocol overhead and latencies can roughly account for typically 12-20%, depending on the DDR frequency and programmable latencies. Technology Explained! Cont. Current JEDEC (Joint Electron Device Engineering Council) standards are as follows: PC1600, aka DDR200, providing up to 1.6GB\s Bandwidth PC2100, aka DDR266, providing up to 2.1GB\s Bandwidth PC2700, aka DDR333, providing up to 2.7GB\s Bandwidth PC3200, aka DDR400, providing up to 3.2GB\s Bandwidth Any frequencies not listed are not JEDEC standard, but moreover manufacturers catering for the enthusiast using extreme high grade performance ICs (Integrated Circuits) such as Samsung TCCD and non-standardised PCB designs, such as the Brainpower PCB. In more recent times DDR2 has become commercially available, its primary advantage is providing much higher clock frequencies than DDR can achieve, however this is done at a significant trade-off with latencies, in essence still rendering DDR more efficient and effective, as DDR can maintain consistent comparable (or better) performance than DDR2, weather it be comparing the two at “stock” (default system clocks) or running the modules head-to-head at their maximum clockable frequency on a 1:1 (CPU:RAM in sync) divider. A good example would be comparing DDR600 (PC4800) with latencies of 2.5-3-3-8, vs DDR2 800 (PC2 6400) with latencies of 4-4-4-12. An alternative to DDR is self-clocking, and has been chosen for use with PCI-Express and Infiniband. Self-clocking is a signal that can be decoded without a separate clock signal or other source of synchronization. This is usually done by including embedded synchronization information within the signal, and adding constraints on the coding of the data payload so that false synchronization can easily be detected, an example of self-clocking would be 8B\10B encoding. A Typical Enthusiast Level Mainboard BIOS (Basic Input\Output System) For this explanation I will use my personal DFI Lanparty SLI-D, explaining the more advanced configuration settings. Genie BIOS LDT \ FSB Ratio - Controls the HyperTransport Bus, default is 1000MHz, or 200x5. LDT (Local Descriptor Table) Transfer Width - Controls the size of the upload and download transfer pipe available. The LDT Bus is a memory table used in the x86 architecture, containing memory segment descriptors, size, linear memory, access privileges etc. The LDT Bus works by assigning separate address spaces for multiple user processes, normally one LDT will be assigned to each user process, when the user opens a new process the OS intelligently switches the current LDT. CPU Special Control - Defines above standard CPU voltages, when used together with CPU control just about any voltage variance is possible up to 2.1v. DRAM (Dynamic Random Access Memory) Configuration Tref - The cycle interval between the CPU and DRAM, 15.6us is the most relaxed, but also offers maximum stability with no noticeable performance loss, then for the modules that are better than others, the performance option of 3.9us, this is fairly tight and often causes problems, few modules can run this tight at high frequencies. Read \ Write Bypass - This option specifies how long an action can remain in a queue before being veto'd. For example, if set to 16x the action can remain in the queue for 16 accesses before being veto'd. Bypass Max - This option specifies how long a request can remain ignored for in the queue, for example if this option is set to 7x it can be ignored for 7 accesses before being forced to be read or veto'd. Async Latency - Refers to DRAM response times, lower equals better but at high frequencies this naturally has to be adjusted to allow the system to remain stable, for example, with a bus clock of 200MHz x5 = 1000MHz HTT this option should be able to be set at 7ns (Nano Seconds) or 8ns and remain stable, at higher frequencies such as 250 x4 = 1000MHz HTT this option should be set to 8ns, and when running DDR600 speeds or higher, setting this option to 9ns is almost mandatory. Idle Cycle Limit - This option specifies the idle clocks between accesses, by default this option should always be set to 256clks, but high performance enthusiast modules can sometimes remain stable with a lower Idle Cycle Limit, such as 128. Dynamic Idle Cycle Counter - This option primarily is for TCCD based modules, if enabled this option forces the DRAM to dynamically adjust the Idle Cycle Limit based on page conflict \ page miss (PC/PM) traffic. DRAM Drive Strength - Options are normal and weak, normal should be used for such ICs as BH-5 and UTT whereas a weak drive strength should be used for TCCD ICs, weak drive strengths are uneven values while normal drive strengths are even values. The MD Drive Strength determines the signal strength of the memory data line. The higher the value, the stronger the signal. It is mainly used to boost the DRAM driving capability with heavier DRAM loads (multiple and \ or double-sided DIMMs). If you are using a heavy DRAM load, you should set this function to Hi or High. Due to the nature of this BIOS option, it's possible to use it as an aid in overclocking the memory bus. Your SDRAM DIMM may not overclock as well as you wanted it to, but by raising the signal strength of the memory data line, it is possible to improve its stability at overclocked speeds. However, this is not a surefire way of overclocking the memory bus. In addition, increasing the memory bus signal strength will not improve the performance of the SDRAM DIMMs. Hence, it's advisable to leave the MD Driving Strength at Lo \ Low unless you have a high DRAM load or if you are trying to stabilise an overclocked DIMM. DQS Skew Value - Key in values are 0 - 255, adjusting this tweaks the sine wave, allowing modules at extremely high frequencies to be run with much greater stability. Due to this value able to be adjusted by an experienced user, its fairly safe to assume the sine wave is digital (signal varies according to a series of discrete values) and not analogue (the signal varies continuously according to the information.) CPC (Command Per Clock) - Controls if the modules will operate at 1T or 2T, 1T is common for memory up to 1GB (2x512MB) in size, where 2T is common for memory of 2GB (2x1GB) capacity or more. The reason 2T command timings are required is because the CPU controller can have issues with reading the memory ranks correctly. CAS Latency (TCL) - CAS, or column Address Strobe, refers to the column of physical memory in an array of capacitors ( a grid comprising of rows and columns) used in DRAM modules. The latency refers to the active amount of clock cycles that must be expended to take a request. Combined, TCL sends data from the memory controller, has it read to the memory location, and output to the modules output pins. RAS to CAS Delay (TRCD) - The memory controller selects a bank, followed by a row location (using the Row Address Strobe, or RAS) and a column location using CAS. It represents the time in cycles for issuing a command and active read \ write commands. Min RAS Active Time (TRAS) - Represents the amount of time taken between a row being accessed and deactivated. A TRAS row must be allowed to complete before being deactivated, setting this option too low can result in data corruption as the row is closed down too soon. Row Precharge Time (TRP) - Row Precharge Time represents the minimum allowable time taken between any active command and the read \ writes of the following bank on the memory module. Row Cycle Time (TRC) - TRC is the minimum time taken for a row to complete a full cycle, setting this too low will result in data corruption, and setting this too high will result in a loss of available bandwidth but increase stability. Row Refresh Cycle Time (TRFC) - This parameter determines the amount of cycles it takes to refresh a row in a memory bank, again if this is set too low it will cause corruption and set too high will result in a loss in available bandwidth, but increase stability. Row to Row Delay (TRRD) - Represents the time in cycles it takes to activate the next bank in the memory, this is the opposite of TRAS, the lower the timing the better, but it can cause stability issues. Write Recovery Time (TWR) - Sets the time taken after a write operation and precharge before another can start. Write to Read Delay (TWTR) - Sets the time taken before the next command can be executed. Read to Write Delay (TRTW) - Sets the time taken between a valid write command and the next read command, lowering this timing increases performance but can cause stability issues. Write CAS Latency (TWCL) - Writes to any available bank, operates at 1T, other parameters can be set, some may cause the system to become unbootable, requiring a CMOS (Complementary Metal-Oxide-Semiconductor) reset. Technology Explained Section provided by the one, the only, Ketxxx! Thanks for your support! Time to Overclock! In this section we will cover a couple of overclocking methods. This guide is more intended for proper methods and preperation than actually overclocking. And in all actuality, the method of Overclocking is simple. It's the preperation and stability methods that can add compication. The best thing you can do to be prepared is know your components, understand what you need to adjust in Bios to attain your overclock and know what too look for when maintaining stability. So for a simple explanation of overclocking your modern system (AMD/Intel), you increase the CPU/System Bus speed. This will overclock your Motherboard Chipset (North Bridge, South Bridge), your CPU and your Memory. One thing you want to do before performing this simple task is ensure your voltages are manually set. If they are not, your Motherboard will increase voltages according to it's algorithm and this can create excess heat and uneccasry extra voltage to be stable. Same goes for the PCI/AGP/PCI-e Bus lanes, you will want to manually set them in order to prevent an increase that could cause instability. I hope this section is found helpful for those looking to begin their overclocking procedures. I plan on covering the basics, which for many will suffice enough to overclock, I still recommend looking further into your specific application for overclocking methods as different boards and manufacturers will have varying bioses and settings. Basic Overclocking Tips The most recommended method of overclocking is via the Bios. When planning on Overclocking, make sure you set your Bios voltages and bus speeds manually. Examples are CPUv, DDRv, NB, SB, FSBv for voltages, and certain BUS speeds such as PCI should be set at 33.3MHz and PCI-e should be set at 100MHz. If you have AGP, set it at 66.6MHz. Generally I recommend starting with stock CPUv as my first setting as many CPU's now can take a decent OC off of stock voltage settings. On the motherboard side, I will increase most if not all by one setting to ensure stability. I have found some boards to be more willing to Overclock on little to no voltage increases also, which is great for keeping things a little cooler. Features such as Vanderpool (Unless using a Dual Boot OS setup like Vista and XP), C1E, EIST, TM, and Spread Spectrum should be disabled, as power saving/modifying settings can cause instability at higher speeds. Once you do attain your favored Overclock, you may want to try the power saving features to see if your PC stays stable, that way at Idle your PC will not only run cooler, but consume less power. Also make sure you adjust your Memory Voltage and Timings accordingly. You should have researched what your memory is capable of, and learned about your Memory Dividiers/Multiplier Features in your Bios. Generally with a newer Core2/DDR2 setup, and depending on CPU Bus Speed and DDR 2 Rated Speed, I run a 1:1 ratio, or generally the lowest ratio. If you set your memory to run at 533MHz (266x2), it will be running in sync with a 1066MHz FSB Rated CPU (266x4, yep, Intel still uses Quad Pumped FSB). That way, say you increase your FSB to 400, and you have DDR2 800 in your system, it will be running at DDR2 800 speeds (400x2), but your CPU will be Overclocked. Some CPU's can overclock beyond this point, some can't, verifying with Orthos helps a lot. Now comes the next easy part to overclocking, and this is the kicker: INCREASE YOUR FSB SPEED! There lies the main component to Overclocking your system. Sure if you unlock your CPU Multipliers (some chips are more limited than others), you may be able to go up, and overclock that way (now-a-day's it's usually the more expensive, higher end chips that do this). And even at that, your overclock will not be as substantial or noticed as raising the Front Side Bus speed. Take my P5B Deluxe/C2D 6300, at stock my MB is running a 266FSB setting in Bios. Well multiply that X4 and you get 1066FSB. When you install a CPU into your compatable motherboard, it will generally have to recognize the CPU and auto-set itself accordingly. One of these settings is the primary FSB setting. So if I were to install an e4400, my FSB would change to 200 (800FSB), and so on. So why all this prepearation and build up for something that's so simple to do? Well because it has a huge impact on your system, from performance to heat, etc. You will want to start increasing your FSB easy at first. Most modern systems are capable of taking a 50-100+MHz increase in FSB without really having to change anything from manual set stock voltage settings. I still recommend you don't leave any important voltage on AUTO when initially overclocking, some such as ICHv on the P965 series (reference, P5B Deluxe) have little to no effect on Overclocking, but every other voltage setting should be manually and properly adjusted. On a lot of boards you either have to disable Auto Overclocking (or whatever name they gave it) or unlock a special menu. Check your motherboard manual and check online for specifics if you haven't already. Then you simply push the ENTER button and type in a value. I decided to do a healthy almost 100MHz increase and went to 350FSB. I then pressed F10 (or go to last menu, hilight and press ENTER on SAVE AND EXIT), and let it restart, which it did flawlessly all the way into windows and was successful through a 30-Minute Orthos Stress. There are two keys in my eyes for intial success in overclocking, one is POST. POST is what your motherboard/Bios has to do upon power up, in fact POST= POWER ON SELF TEST. If you pushed too far, or missed a setting that is causing an issue, you may not POST. This may require resetting your Bios or using it's Failsafe to assist you. The second key is completely booting into Windows (whatever version), no errors, no random restarts, no BSOD's, etc. If you pass these two tests (the higher you go, the harder it will become, due to the fact it takes more tuning and increasing of voltages to get there), then you are ready to start stressing your CPU. I recommend running Orthos for at lest 20-30 Minutes, set at Priority 9 using Small FFT's to stress the CPU. If you error, then I would start by increasing the CPUv by one setting. If you pass without errors or issues, then it's time to get back into your Bios. Increase your speed some more, this time for example I increased by only 50, to 400MHz FSB. Now my memory which is DDR2 800, but strapped at a 1:1 ratio with the CPU at stock speeds (DDR533 = 1:1 strap on my specific board when a 1066FSB CPU is installed), was reading DDR533 at stock. Now it's reading DDR800 speeds which is what it's rated at, so I can use it's recommended timings and voltages without fear. Plus my System BUS is overclocked as-well-as my CPU. This is what I consider a good PRIME Overclock, this will be the easiest level of overclock to attain for many with newer systems. And, if you can't tell it's very easy to do. But while doing all of this and stressing your CPU, you want to monitor temperatures, notice any errors/falures/BSOD's, and realize that odds are any issues are the effect of your Overclock. A Few Tips To Remember: Know about how your BIOS works before changing settings. It's best initially to disable ALL Power Saving Enhancements for best stability, but do try them later on if you do leave your PC idle. Lock PCI Bus to 33.3MHz, AGP Bus to 66.6MHz, and PCI-e Bus to 100MHz. You may need to UNLOCK the Menu for these settings, or Disable Auto Overclocking to allow the Menu to be shown. Be prepared for this BEFORE attempting to overclock. Don't go too far initially, a 50-100MHz increase on most newer applications (especially Core 2 Systems) is allowable. You may want to start smaller with 10-20MHz increases. Don't forget to test after each increase. If you've successfully made it into Windows, you have passed 2 tests already! Don't forget to monitor temperatures. Keep a sharp eye on them! Understand your Memory Strap feature and adjust it according to how you intend to Overclock. If you plan on overclocking higher, a lower strap may be required as to not push your Memory too far. This causes many NO POST situations, many have overlooked this and been bitten by it. Read and understand specific Overclocking Methods that pertain to your Motherboard/CPU, there are different methods and settings, following the wrong one will lead to issues. Just remember, the act of Overclocking isn't hard to do, just make sure you are successful at it by doing it correctly and being prepared. Don't immediately increase your FSB to 500 because you think you CAN hit it! Don't try high overclocks with a lack of necessary cooling, that is unless you like to spend more money. For now this guide is more a guide of preperation for Overclocking, there are too many different methods in different BIOSes for me to list here. And there are so many guides that are dedicated to that part, and instead of copying them, I will refer to them. If my instructions and explanations above did not help you, or you are still unsure as to how to Overclock your application, take a look at the guides in the following Links section, and if all else fails, use GOOGLE! -------------------------------------------------------------------------------------- Overclocking in Windows? Yes, you can! But I personally don't recommend it. I have used many programs, from those provided from the Motherboard Manufacturer themselves to TPU's own Systool. They work well, but with some settings a restart is required and you can't access all the tweaks you need in order to be successful. If you want to use one of the utilities, you will find them in Programs/Diagnostics/Stress Tests for Overclockers. If planning on using any program to overclock your system in Windows, you should ensure you have certain BIOS settings in place such as BUS Speeds and Voltages that if even accessable by Windows will probably require a restart which is where you'd be if entering BIOS to Overclock in the first place. Some programs, mostly those from the Motherboard MFG can save settings to BIOS such as FSB, PCI-e, AGP, PCI, Fan Speeds, etc. Other utilities like Systool and Clockgen may not support that, hence require a profile load at start-up. If you decide to want to take this route, do so carefully as it's recommended for more experienced users that understand that they need to adjust in Bios what the program doesn't see or let you see. If you want more instruction how to use Clockgen or Systool, check the Laptop Overclocking section by Theonetruewill. -------------------------------------------------------------------------------------- Important Links for Overclockers: Overclocking - The Basics Theonetruewill's Laptop Overclocking Adventures (The Only Link you NEED for OC-ing Laptops!) Beginner's Guide to Intel Overclocking Beginner's Guide to AMD Overclocking Graysky's Core2 Quad Overclocking Guide Techpower UP! Overclocking and Cooling Section Hard Forum's Overclocking and Cooling Section MrSeanKon's OCBible and Guidemania! (Must Read For ANY Overclocker!) These links will go further into detail than I have on this guide, and if those guides don't work, check the Overclocking and Cooling sections at both forums! There are 100's if not 1000's of different threads to read. You can search to find a thread that may have some help and suggestions for you to attempt. The knowlege in these forums is priceless, as-well-as the history of Overclocking they both contain. Odds are you will find a thread pertaining to your system and a similar issue or setting you need to attain. If you do fail to find your answer, then register and post for help. As long as you learned in the first page how to better ask a question, you will get the help you need! ----------------------------------------------------------------------------------- Laptop Overclocking is Also Easy! Get More Results! So, you have a laptop (or PC with very limited bios/OC capablity in look for a way to Overclock), and want to overclock? Welcome to Laptop Overclocking is Also Easy!, provided as a culmination of primarily TPU's own Theonetruewill. We are proud to add this section to assist those who are unable to really use the rest of the guide (aside from learning, researching, understanding, monitoring temps, etc), it is full of excellent information. We hope that you will enjoy and find your question's answered by this part of the guide! ----------------------------------------------------------------------------------- Getting Started First things first, you need to know several things about your copmputer. To find out a lot of that information you can view about your Hardware in the Devices manager or similar time consuming methods or you can simply download CPU-Z. I also need to stress the dangers of overclocking through the following methods. Unlike within the BIOS you are not able to set the PCI frequency at a static number thus it will increase. This increases the chances of data corruption if something goes wrong, so please be careful and ensure you're taking the right steps and prevention methods. Although I have never encountered this. Please overclock AT YOUR OWN RISK. I cannot be held responsible. I also recommend a cooling pad if you are overclocking laptops, as the increased heat output may overheat your processor unless it has good standard cooling, and since there isn't much to be had for aftermarket heatsinks for laptops, you must be extremely careful, always monitor your temps. If you have a Laptop you will have to disable Cool n' Quiet (AMD), or Speedstep (Intel). You can do this with Notebook Hardware Controls (NHC) and setting MAX Performance in the CPU performance tab, OR by using Rightmark Clock Utility. You may also be able to disable it in the BIOS - probaby under 'Power Management' or similar. This must be done to allow your processor to overclock successfully. Here is a list of other recommended programs that will be useful when undergoing the methods described below. ----------------------------------------------------------------------------------- Overclocking Tools: CPUFSB Basic FSB changer. Use this to overclock your Laptop. SoftFSB 1.7 -A substandard FSB changer - I'm not fond of it. Use this to overclock your Laptop. SpeedSwitchXP -A tool to disable Speedstep on Intel processors. Clockgen -The most user-friendly overclocking tool. *Recommended* CPUMSR -A utility that can alter a lot of processor options. !Use with caution! CPU-Z -Displays useful Computer Identification information. *Recommended* Notebook Hardware Controls (NHC) -Best multipurpose program for Laptops. *Recommended* Systool (Alpha release) -An early alpha release prone to crashing. ATiTool 0.24 -The preferred tool for overclocking Laptop graphics cards. *Recommended* ATiTool 0.26 -A newer but more troublesome version of ATiTool. Rightmark Clock Utility -Useful Processor throttling and Speedstep controller ----------------------------------------------------------------------------------- METHOD 1 - Clockgen - The Simple Life The First method we will try is to overclock your CPU (processor) by increasing the FSB with a tool called clockgen. It works by finding your pll clock generator. It then manipulates this to increase FSB frequency, which in turn increases your CPU speed. For the following method I will be using a 1.86GHz Pentium M 750 as the example processor. Where speeds are mentioned with *'s around them this denotes that you this speed is from the example and that you should apply the speed of your chip there. You can find this by ujsing CPU-Z. It is on the front page. As a rough guide to which speed is relevant these are some approximate speed values for a variety of different chips at default speeds. The speed changes are varied depending on your particular model in the series. AMD Sempron - 1.6-2.2GHz~ AMD Athlon64 - 1.8-2.4GHz~ AMD Athlon64 X2 - 1.8-3.0GHz~ Intel Pentium M - 1.6-2.3GHz~ Intel Celeron M - 1.2-2.0GHz~ Intel Celeron D - 2.2-3.4GHz~ Intel Pentium 4 - 1.6-3.4Ghz~ Intel Core Duo - 1.6-2.3Ghz~ Intel Core 2 Duo - 1.6-2.9GHz~ First things first - you need to find YOUR pll generator among clockgen's list in order to actually manipulate it. There are only two ways to find out your pll generator, one of them can void your warranty (Laptops and retail machines only), and the other is time consuming. The latter is probably more preferable. 1. Open up your laptop, find it on the motherboard and then look at what it says it is. I don't advise this. 2. Instead you can just try with every pll generator in clockgen, until you get the correct one. The correct one will show your CPU frequency as identical to the one in CPU-Z. Find it via trial and error In order to accomplish this accurately you'll need to open Clockgen and use these EXACT instructions. a) Go to pll setup b) Then tick the box named "ignore GSB/PCI" c) Now select the first pll generator and then click read clocks. d) Open pll control and see if it reads ~*1860* (1.86GHz = 1860MHz) ,or very close, and I mean very close. (Remember to replace *XXXX* with your particular clock speed) e) If it does great continue to step 3. f) If it doesn't, change the pll generator and click read clocks again. Check the Pll control cpu speed. and continue if necessary down the list. g) If none of them seem to work, do them all again (ps some crash your computer, don't worry just reboot. This may require actually pulling the plug which adds some risk) h) If none of them work again try METHOD 2, but this will be more challenging. 3. As always, you will want to ensure your settings are stable, in order to ensure this it is recommended you have Notebook Hardware Control installed and set. With this you need to set your CPU to Max performance (You should have done this already - if you didn't your speed will have been bouncing around!). This removes the Intel speedstep function that slows down your cpu to save battery life. But its not necessary when you're connected to the power plug. Next, go to the Hard drive performance and set noise level up to the highest it can go(performance and number-wise) as this gives an added performance boost to your machine. Don not worry if this option is disabled it is not necessary. Then download Orthos Prime, this will test the stability. Now you are ready to increase your cpu speed by about 10Mhz using the "pll control" slider in clockgen. Then do a stability test in orthos for about 30mins. The go to another 100Mhz and another 30mins, etc. after you reach your limit increase by smaller amounts. Then finally test for about 4 hours.As soon you get an error in orthos or your computer crashes (BSOD - Blue Screen Of Death) it's too high; so lower it to the previous good one and test orthos on that speed for about 2 hours. If no errors, then great! Do not skimp on this verifactaion of stability, or you may pay dearly for it. Stability and proper temperature under load are key to maintaining your OC'd Laptop for a long time. This may be time consuming, but when you are done and considered stable, you'll be happier for a lot longer. Notes: NEVER set the clocks to appear at startup, just do it before each time you want a performance boost (i.e. gaming) Also Notebook Hardware controls (NHC) has a great feature to show the temperatures of your cpu and HDD. Don't let the CPU ever go above 80 (Intel Laptop figure) in my opinion (AMD Laptop chips never over 70). It shouldn't really be above 60 celcius most of the time. (My Pentium M 750 runs at around firstname.lastname@example.orgGHz). Overclocking using clockgen increases the RAM bandwidth also, thus further speeding up your PC! Systool crashes A LOT! because its an early release, so if your computer crashes mulitple times just restart. Clockgen will not fuction properly if CPU-Z is open at the same time. ----------------------------------------------------------------------------------- METHOD 2 - Systool - Crash Install Systool then run it, and go to; a) Hardware monitoring > CPU overclocking. b) Now you need to do the same thing as you did with clockgen. Go thorugh all of the pll's until you get the right one. There are two ways you can know if you have found it. Firstly systool will say " clock generator detected." However DON'T ASSUME THIS IS THE TRUTH! You will get a lot of false positives, so you also have to look at what clock readings it will give you. If you have set the Notebook Hardware controls to Max Performance (I should not have to keep reminding you!) for CPU speed. This is *1860MHz* approximately with the example cpu. c) If it does say the correct speed, great. You can then try and move the fsb slider. Now only do this by small bits at a time, as otherwise you are going to crash your pc instantly and maybe even monumentally! d) If this is successful, then do the same as in METHOD 1 with clockgen when increasing speed and testing for stability. For your information I can achieve a relatively stable overclock to about 2250. But I don't like to go to the limits in a laptop, so I am at 2200 most of the time. Notes: Systool is an early alpha release and may crash/freeze regularly. This may be frustrating but it is still a very powerful program, be patient with it and understand how powerful it is. ----------------------------------------------------------------------------------- METHOD 3 - CPUFSB - The Long Way Round Install CPUFSB - this requires a restart as it installs a driver just as with ATiTool. Note this program is not free but can be used as an unregistered version. a) Run the program and prepare for a long ride. EITHER b) Select you motherboard manufacturer and model (find model in CPU-Z 'Mainboard' tab) and then click "Get PLL output." If it is selected correctly, you will be able to select # OR c) If it does say the correct speed, great. You can then try and move the fsb slider. Now only do this by small bits at a time, as otherwise you are going to crash your pc instantly and maybe even monumentally! d) If this is successful, then do the same as in METHOD 1 with clockgen when increasing speed and testing for stability. For your information I can achieve a relatively stable overclock to about 2250. But I don't like to go to the limits in a laptop, so I am at 2200 most of the time. Notes: This program can take a long time, and it still might not yield results. However if this doesn't work for you, I'm beginning to run out of suggestions. ----------------------------------------------------------------------------------- METHOD 4 - SoftFSB 1.7 - Trial and Error Run the program after placing it in a folder, it creates a lot of files. Then press the 'Y' key three times in conjunction with instructions from within the MSDOS Prompt. I believe this is all to do with how it mages to alter the FSB speed via the pll generator, but I can't work out quite how it does it yet. a) Now run the created program (It has a SoftFSB 1.7 icon). It will either detect you motherboard instantly or fail to. if the latter is the case select it yourself, from the drop-down list. you can find your Motherboard model under the 'Mainboard' tab in CPU-z. Alternatively if your model is not their, use a similar method as METHOD 1 to find your pll generator. However click 'Get FSB' to let it work out your FSB. As far as I can tell, the program crashes automatically if the incorrect pll is selected. It generally does not crash your whole computer, just the program. If the pll is correct you will be able to move and set the slider freely. b) Move the slider to the desired setting c) Again use the method outlined in METHOD 1 to test for stability if you make any changes d) If this is successful, then do the same as in METHOD 1 with Clockgen when increasing speed and testing for stability. Notes: I personally don't like this program, partly because I don't admittedly understand how it works and actually changes settings. It's not quite as simple as it looks. But many have found it useful, it is recommended for more experienced users. ----------------------------------------------------------------------------------- Fix Locked Clocks on ATi MObility Radeon cards!- Original documentation by Twisterdark, edited by Theonetruewill If ATITool displays the correct clocks but refuses to change the speeds ---> change the drivers. Force the desktop drivers even if you've got a mobility card (POWERPLAY OPTION WILL NO LONGER BE AVAILABLE) so that ATITool can unlock the clockspeeds. You will have to download the latest Catalyst drivers, and then modify them (step 2). You can download ATITool here at TPU thanks to it's creator and TPU's commander, W1zzard (also responsible for Systool, OCDB, and much more! Check out his amazing work on TPU!) Perform the following tasks (to overclock Mobility & Mac Radeon cards only); 1) Download latest Catalyst/display drivers. 2) Mod them http://www.driverheaven.net/modtool/ 3) Install from Windows control panel (Start> Control Panel> display driver, searching them MANUALLY into the .INF that you can find into the modded MODDED ATI DRIVER - select x1600 or x1650 (no mobility) 4) Restart 5) Install Catalyst Control Center Modded 6) Restart 7) Install ATI Tool and check the clock, if it's strange (like 7400 mhz clock - impossible) go to settings and check, on the x1000 overclocking section, "use driver level clocking". This will alleviate many headaches with OC-ing your ATI Mobility based Radeon. 8) Before starting to overclock, write somewhere clocks you see after the point 7 because the new "defaults" botton will go to 85.9/85.9....and this mean that if you unluckily click it a lot of articrafts will appear (because of too much downclocking), but don't be scared...it's all ok, nothing dangerous, the problem is that it's hard to find something to click on the screen . So, to be prepared, create a new profile with you real default settings and us it like defaults before starting to joke with clocks. Notes: As always with ATiTool, check for stability with the 'Artifact scan' feature. If any are present reduce the clocks. Use ATITool 0.24 for Mobility Radeons. You will encounter less problems. Note From the Creator/Editor Thank you for taking the time to read this guide, and hopefully it taught you something new, or helped in a situation or decision. I created this guide in the beginning May 2007 as a place for newer and inexperienced users to learn how to gain more of the basic information and understanding of their PC's and it's components. It started as a place to help them ask questions for help in a more proper way, and has in short time grown into a fairly large guide for many to use, from the new to the more experienced. Remeber, even the experienced users can go into an arena they're unfamiliar with, but they know how to take the right steps to be successful. I believe my goal has been accomplished, with the help of TPU, it's users, and even un-registered viewers showing support. There are many guides out there, and I hope you see my intention of creating a different style of guide, not to be better, but to be easy to read and understand. This guide is a constantly changing and growing thing, and with your help it can only get better. I thank everyone for their help, support and suggestions thus far, so please don't stop! If you know something I don't, or think I should do something, then please let me know. I would like to thank Theonetruewill for his submission of the entire Laptop OC section, I would like to thank Ketxxx for his submission of the entire Technology Explained section, they have done an excellent job and I am greatful for their submission of all that extremely useful information. Also, I would like to thank TPU for it's support, and Multi-IT for also hosting this guide as an article!