Intel Haswell Overclocking Clubhouse.

[Kursah]

I wonder if it comes down to Haswell compatible memory kits? My memory kit said it was IB compatible, but nothing about Haswell. Been a solid kit overall imho...but could it be the cause of some instability? I dunno. It runs fine at 1600 as well, but I lose ~10,000 mb/s bandwidth when doing so...not sure if that really makes a noticable difference or not...but I would assume it does and may depends on the game/application.

Well I replaced the stock 120mm exhaust fan in my case with a Corsair SP120, and that has made a nice difference. Lower temps load and idle but a couple degrees if not several and not as loud while also moving what feels like more air. The stock fan wasn't bad...but the Corsair is a solid performer! I'm impressed!

Hunter, I will disable that and give it a shot!

 

[Womper]

I wonder if it comes down to Haswell compatible memory kits? My memory kit said it was IB compatible, but nothing about Haswell. Been a solid kit overall imho...but could it be the cause of some instability? I dunno. It runs fine at 1600 as well, but I lose ~10,000 mb/s bandwidth when doing so...not sure if that really makes a noticable difference or not...but I would assume it does and may depends on the game/application.

Running at 1600 is fine for games. You're going to miss out on a couple fps in situations where you are CPU memory bandwidth limited. Like if you're at 60fps with 2133, you'll be at 57fps with 1600 (~5%). At least that's how the memory kit reviews on Ivy Bridge went.

 

[Kursah]

Well I've gotta assume it'll be similar with HW then. Back to finding stability with 4.3Ghz. 1.26v, disabled the VR setting Hunter suggested and also in that setting's description is a recommended Disable while overclocking. I'm hoping that Dave's guide covers some of these...as I hope they are kind of standard across Z87 OC MB's. Looking forward to reading it! Looking forward to maybe...just maybe finding 4.3 on air with stability!

Good news with the new exhaust fan, instead of loading into the 80's during the heat at this time of day (and only get's worse)...I'm just into the 70's! That makes me happy for being on air and dealing with cooling a Haswell.

 

[Jstn7477]

I pulled my rig back to 4.3GHz/4.3GHz (172*25) since running a Crossfire setup caused me to get occasional 124 BSODs that I didn't have at 4.4GHz/4.23GHz. Need to lower the volts a bit (1.230v for core/ring) but should be alright.

 

[Kursah]

Still can't seem to pass even a couple hours of Intel's Extreme Tuning Utility test. Well I did run the memory test and that's what failed, it seems to load the CPU similar as the CPU test and also utilize more memory for testing.

It was well over an hour in when I walked away for work. Come back to a failed restart and ETU saying it wasn't closed properly. GRRR.... I thought I was closer than ever to finally finding OC stability with this POS chip. Gonna drop memory back to 1600 again I guess. This K series chip can stay marked in the disappointment in the OC category for a while longer yet.

EDIT: Back at 4.2 for now, was able to play Wargame for a while, and did not stress test any more...leaving town for a couple days. I'll come back at it then and hope it'll lock down stable.

 

[Kursah]

Well I'm almost convinced my chip is one of the worst 1%'ers out there...I can't find true stability while overclocking for anything. Ram at 1600, 1866 or even rated 2133...cache at 35 or 39, or even auto...voltages, VRIN 1.75-1.90v, cpuv 1.2v-1.36v, memory 1.58-1.60v (bumps up by .02v on all readings), vccsa .864 (stock) to 1.15v, vcc digital and anaog auto to +.100v, RingV auto to 1.25v, memory timings anywhere from cl9 to looser than rated cl12, and everything I've discussed in previous posts in this thread. None of it seems to keep this thing stable..and when it decides it's gonna go to hell with stability games crash, tests crash and honestly I don't think I've every found stability maybe just a lucky time-frame or who knows.

I do however have my Batch for ya Dave: L307B189 It is a Malay as well. Guess it's back to stock till I find some more patience for this thing again or just decide to risk an RMA for what can't possibly be a worse chip imho. Anyone got suggestions for something I haven't tried yet?

 

[Womper]

Well I'm almost convinced my chip is one of the worst 1%'ers out there...I can't find true stability while overclocking for anything. Ram at 1600, 1866 or even rated 2133...cache at 35 or 39, or even auto...voltages, VRIN 1.75-1.90v, cpuv 1.2v-1.36v, memory 1.58-1.60v (bumps up by .02v on all readings), vccsa .864 (stock) to 1.15v, vcc digital and anaog auto to +.100v, RingV auto to 1.25v, memory timings anywhere from cl9 to looser than rated cl12, and everything I've discussed in previous posts in this thread. None of it seems to keep this thing stable..and when it decides it's gonna go to hell with stability games crash, tests crash and honestly I don't think I've every found stability maybe just a lucky time-frame or who knows.

I do however have my Batch for ya Dave: L307B189 It is a Malay as well. Guess it's back to stock till I find some more patience for this thing again or just decide to risk an RMA for what can't possibly be a worse chip imho. Anyone got suggestions for something I haven't tried yet?

Honestly I'm surprised it works at stock speeds. If it does work at stock speeds, I guess that rules out any major incompatibilities with memory or something.

I think you should either run it at stock/light overclock, or get a new chip.

 

[d1nky]

I got a friend that's about to build an intel rig.

he overclocks and benches like a bitch, its his thing.

what would the best choice be out of 3770k/4770k/3930k

btw he likes overkill and money isn't really a problem.

edit he went for 3930k and rivf

 

[SimpleTECH]

edit he went for 3930k and rivf

Good choice as the CPU will yield a higher score in most benchmarks and the platform has enough PCIe bandwidth for multiple graphics cards.

 

[EarthDog]

Good choice as the CPU will yield a higher score in most benchmarks and the platform has enough PCIe bandwidth for multiple graphics cards.

That depends on the benchmark actually... For example, 3DMark 01, 03, and 05 are single threaded to which you want the fastest out of those (the 4770K on LN2). 06/11/V will respond to multiple cores. However a 6.3GHz 4770K will smoke a 5.5GHz 3930K. He has to go cold with the 3930K to make it work better for him.

Bandwidth for GPU's isnt a huge issue either. Check out HWbot and note that most 3D records were not done on X79 platform.

 

[FireKillerGR]

3dmark05 isnt single threaded
Try the following: 1c/1t vs 1c/2t vs 4c/4t vs 4c/8t.
It gains from cores but not from hyper-threading

 

[Kursah]

Honestly I'm surprised it works at stock speeds. If it does work at stock speeds, I guess that rules out any major incompatibilities with memory or something.

I think you should either run it at stock/light overclock, or get a new chip.

This rig is solid at stock CPU, XMP memory speeds. I threw every test in the book all night last night at it stock just to make sure yet again there wasn't another issue like the memory. Every test passed. CPU runs 3.9 turbo on all cores at 1.21v (everything on Auto)...but I still would have figured being an Intel Core that that same voltage would allow me at least 4.2-4.3Ghz OC ...but then again I missed out on SB and IB OC-ing except for some minor boosting for customers that didn't want much.

I am seriously considering the Intel Performance RMA...just worried that I may get a worse chip...though not entirely sure how that could happen lol.

 

[Womper]

This rig is solid at stock CPU, XMP memory speeds. I threw every test in the book all night last night at it stock just to make sure yet again there wasn't another issue like the memory. Every test passed. CPU runs 3.9 turbo on all cores at 1.21v (everything on Auto)...but I still would have figured being an Intel Core that that same voltage would allow me at least 4.2-4.3Ghz OC ...but then again I missed out on SB and IB OC-ing except for some minor boosting for customers that didn't want much.

I am seriously considering the Intel Performance RMA...just worried that I may get a worse chip...though not entirely sure how that could happen lol.

If I do 39x cores synced, cache 39x, and leave voltage on Auto, then I get 1.195v under load. Not a huge difference, and could just be the motherboard.

But if you're telling me that a manual vcore of 1.3v doesn't even get you past 4.3GHz, then it's a waste of time to try to overclock that chip.

 

[cadaveca]

Could just be the XMP profile. I've seen lots of little ram oddities with Haswell; kits that just barely passed verification with SNB/IVB might fail with Haswell since subtimings are quite different. Kits that failed with SNB/IVB, work fine with Haswell, too.

IMC differences are grossly understated, too.

 

[Kursah]

If I do 39x cores synced, cache 39x, and leave voltage on Auto, then I get 1.195v under load. Not a huge difference, and could just be the motherboard.

But if you're telling me that a manual vcore of 1.3v doesn't even get you past 4.3GHz, then it's a waste of time to try to overclock that chip.

Ya a manual vcore of 1.3v does not get me stable at 4.3...

My voltage under load is all stock, so cores and cache are auto...and loading at 1.21v so maybe it's just a bunk chip?

I have primarily been using XMP profiles but have also tried manually overclocking as well. Though come to think of it I may have still had XMP settings loaded when I switched over to Manual settings as it wasn't a clean swap so-to speak with either F5 optimized settings or a cmos clear. Maybe I'll try that here in a bit and see what I can get...but as Womper suggested with 1.3v...it might be a waste of time trying to overclock that chip which is what I feel like after trying since late June to get this thing stable at anywhere from 4.2 to 4.4. Maybe this chip just wasn't meant to work out...wish I had a place locally I could go buy one and have exchanged it or had actually OC'd my rig before Newegg's return policy expired so I didn't have to waste my Intel Performance RMA. But I'm almost set on RMA-ing...but then part of me doesn't want to give into defeat. Though it's awfully frustrating that anything past stock is at best intermittently stable maybe a roll of the dice on an RMA will bring me better luck than the guy in the Newegg warehouse did?

 

[Womper]

Ya a manual vcore of 1.3v does not get me stable at 4.3...

My voltage under load is all stock, so cores and cache are auto...and loading at 1.21v so maybe it's just a bunk chip?

I have primarily been using XMP profiles but have also tried manually overclocking as well. Though come to think of it I may have still had XMP settings loaded when I switched over to Manual settings as it wasn't a clean swap so-to speak with either F5 optimized settings or a cmos clear. Maybe I'll try that here in a bit and see what I can get...but as Womper suggested with 1.3v...it might be a waste of time trying to overclock that chip which is what I feel like after trying since late June to get this thing stable at anywhere from 4.2 to 4.4. Maybe this chip just wasn't meant to work out...wish I had a place locally I could go buy one and have exchanged it or had actually OC'd my rig before Newegg's return policy expired so I didn't have to waste my Intel Performance RMA. But I'm almost set on RMA-ing...but then part of me doesn't want to give into defeat. Though it's awfully frustrating that anything past stock is at best intermittently stable maybe a roll of the dice on an RMA will bring me better luck than the guy in the Newegg warehouse did?

Dave knows best- the memory controller is a big player if he says so. You can officially declare defeat after trying a few memory kits. But that's not always an option. If you have access to a different set of memory, I'd give it a try. Otherwise, my feeling is that your chip is just a slow one, and I'd swap it via that Intel plan.

After several solid stress testing sessions I've firmed up my settings as far as non-AVX stuff goes. For 24/7 I'm going with 1.275v adaptive mode (that is, 1.275v Additional Turbo Mode for us Asus owners) on both the core and the cache. This gets me 4.7GHz core and 4.5GHz cache. Min cache ratio is 8x, SVID is disabled, input voltage at 1.8v, XMP mode, DDR 2133, command rate reduced to 1, and TFAW increased from 32 to 40. All other settings are left at optimized default. I'm not sure SVID, input voltage, or TFAW have any impact; my earlier complaints about stability turned out to be low core voltage and/or using 4.6GHz cache. 4.8GHz is still possible with a manual 1.34v for non-AVX stuff, but I can't get it to work with adaptive at any setting. Although 1.34v adaptive would pop up to 1.45v for AVX, which is a little too high for my liking anyways.

From what I can tell, my chip is similar to a lot of people running at 4.4-4.5GHz with 1.25v manual...the difference is that I don't care about AVX, so +headroom for me. I might be a little more fortunate that I have the margin to hit 4.8GHz at 1.34v manual. 4.9GHz is a no-go at 1.4v, and 46x cache is north of 1.325v, and not worth pursuing. Interestingly, I saw significant droop on a 1.325v cache with input voltage at 1.8. It needed 1.86v to keep the core voltage on target. Gotta watch for where that droop starts to happen.

 

[Kursah]

Dave knows best- the memory controller is a big player if he says so. You can officially declare defeat after trying a few memory kits. But that's not always an option. If you have access to a different set of memory, I'd give it a try. Otherwise, my feeling is that your chip is just a slow one, and I'd swap it via that Intel plan.

Ya I have one more older G.Skill 2X2GB 1600 CL9 kit I can slap in, but beyond that...that's all I have for memory and I'm pretty much broke at this point. I did do some research before buying the sniper memory, and it seemed that most that used it with Haswell were running just fine at the 2133 and CL11 timings...with a lot of guys buying 2 kits because they were so happy with them. I'll see if I can't find a 3rd different branded set. I can't really afford different memory, nor does this memory fail memtest or any stress test when the CPU is at stock speeds with the memory is at XMP settings.

I agree my chip is a slow one...I'm really leaning towards the swap via Intel plan.

After several solid stress testing sessions I've firmed up my settings as far as non-AVX stuff goes. For 24/7 I'm going with 1.275v adaptive mode (that is, 1.275v Additional Turbo Mode for us Asus owners) on both the core and the cache. This gets me 4.7GHz core and 4.5GHz cache. Min cache ratio is 8x, SVID is disabled, input voltage at 1.8v, XMP mode, DDR 2133, command rate reduced to 1, and TFAW increased from 32 to 40. All other settings are left at optimized default. I'm not sure SVID, input voltage, or TFAW have any impact; my earlier complaints about stability turned out to be low core voltage and/or using 4.6GHz cache. 4.8GHz is still possible with a manual 1.34v for non-AVX stuff, but I can't get it to work with adaptive at any setting. Although 1.34v adaptive would pop up to 1.45v for AVX, which is a little too high for my liking anyways.

From what I can tell, my chip is similar to a lot of people running at 4.4-4.5GHz with 1.25v manual...the difference is that I don't care about AVX, so +headroom for me. I might be a little more fortunate that I have the margin to hit 4.8GHz at 1.34v manual. 4.9GHz is a no-go at 1.4v, and 46x cache is north of 1.325v, and not worth pursuing. Interestingly, I saw significant droop on a 1.325v cache with input voltage at 1.8. It needed 1.86v to keep the core voltage on target. Gotta watch for where that droop starts to happen.

Ya you have a very nice chip compared to mine man! I have tried the TFAW and SVID settings and everything else I can research in this thread and others and what-not. I can't even boot into windows at 4.5!

I am running 1.8 for input voltage, and even tried 1.85, 1.90, 1.75 and none of them changed stability except 1.75 caused crashing sooner and the higher values causing higher temps.

Thanks for the input and constant support guys! I am heavily leaning towards RMA-ing the chip. I just wanted to see if there was just one more thing I could try yet again that just might...maybe...give me a damn break with OC-ing this chip!

It's all good though! I think you're right Womper and seeing your settings, and what Dave's recommended over the past weeks/months I've been utilizing this thread as my Haswell OC resource and report area I think I may have exhausted my limited resources.

Edit: Well Intel's Warranty page is down...guess I have some more time to mess with it.

Edit part deux: RMA initiated...here we go...

Edit part tres: RMA is now active, gonna tear down and ship the CPU out. Will report back in a couple weeks. May be hectic with school, owning a business, family and a job (was hoping to have this PC good to go before school started lol!) but dammit I will figure it out! I'm gonna cross my fingers for an improved replacement!

 

[Womper]

Edit part tres: RMA is now active, gonna tear down and ship the CPU out. Will report back in a couple weeks. May be hectic with school, owning a business, family and a job (was hoping to have this PC good to go before school started lol!) but dammit I will figure it out! I'm gonna cross my fingers for an improved replacement!

Good luck, hope to see big numbers from the replacement!

 

[HammerON]

Ya I have one more older G.Skill 2X2GB 1600 CL9 kit I can slap in, but beyond that...that's all I have for memory and I'm pretty much broke at this point. I did do some research before buying the sniper memory, and it seemed that most that used it with Haswell were running just fine at the 2133 and CL11 timings...with a lot of guys buying 2 kits because they were so happy with them. I'll see if I can't find a 3rd different branded set. I can't really afford different memory, nor does this memory fail memtest or any stress test when the CPU is at stock speeds with the memory is at XMP settings.

I agree my chip is a slow one...I'm really leaning towards the swap via Intel plan.



Ya you have a very nice chip compared to mine man! I have tried the TFAW and SVID settings and everything else I can research in this thread and others and what-not. I can't even boot into windows at 4.5!

I am running 1.8 for input voltage, and even tried 1.85, 1.90, 1.75 and none of them changed stability except 1.75 caused crashing sooner and the higher values causing higher temps.

Thanks for the input and constant support guys! I am heavily leaning towards RMA-ing the chip. I just wanted to see if there was just one more thing I could try yet again that just might...maybe...give me a damn break with OC-ing this chip!

It's all good though! I think you're right Womper and seeing your settings, and what Dave's recommended over the past weeks/months I've been utilizing this thread as my Haswell OC resource and report area I think I may have exhausted my limited resources.

Edit: Well Intel's Warranty page is down...guess I have some more time to mess with it.

Edit part deux: RMA initiated...here we go...

Edit part tres: RMA is now active, gonna tear down and ship the CPU out. Will report back in a couple weeks. May be hectic with school, owning a business, family and a job (was hoping to have this PC good to go before school started lol!) but dammit I will figure it out! I'm gonna cross my fingers for an improved replacement!

Good call. The volts you required were really high for a slight overclock
I am really interested to see how your new 4770K will overclock...

 

[Kursah]

Thanks guys. I have my fingers crossed the next one is at least average! Granted I would love a 4.6Ghz @ 1.20-1.25v that can run on air! Ha! We shall see! I hope to report back nothing but good news! I also hope to see more good news on the Haswell OC front in this thread while I'm eagerly awaiting my replacement. Thanks again for the suggestions and support folks! I greatly appreciate it!

And Dave, any word on your Haswell OC guide yet?

Also I've been spreading the word on the $25 Intel Performance Tuning Protection Plan as much as I can here locally and to many of my online buddies...it's amazing how many people had no idea about this program. Can't say I did until I read your OP in this thread.

 

[cadaveca]

And Dave, any word on your Haswell OC guide yet?

Been done for some time, just awaiting it's turn on the front page. I have three reviews and the guide waiting, working on review #4 right now; I try to keep a bit ahead.

Also I've been spreading the word on the $25 Intel Performance Tuning Protection Plan as much as I can here locally and to many of my online buddies...it's amazing how many people had no idea about this program. Can't say I did until I read your OP in this thread.

I'm probably the only reviewer that mentions it now, it seems. Too bad, too. Thanks for letting people know; it's a cheap service that to me is valuable if you have a conscience.

 

[d1nky]

Ill beat marsey to this one, plugging for the new hwbot tpu team.

some may know the last started in 2005 and died 2010. so im trying to get it back up and active again.

we would love to have many intel guys join and bench for fun. atm its mainly vishera entries.


link in my sig, thanks!

 

[TheHunter]

How far is it safe to push blck @ 100MHz, is 101.6MHz still in the safe zone?

 

[Womper]

How far is it safe to push blck @ 100MHz, is 101.6MHz still in the safe zone?

I was just about to ask the same thing, I'm running 101.2MHz right now.

Also, does anyone know about this microcode stuff for the 4770k? I'm on microcode 9, but some people on hardforum got microcode 12 in a beta BIOS and seem to think it makes a difference with overclocking.

Edit: Didn't last too long in that stress test. The bclk is increasing all of the frequency bins, and with an adaptive voltage, the non-turbo regions don't get extra voltage they might need to handle the increase.

 

[freeboy]

Im hitting a wall in the 4.6 / 4.7 range.. I keep having the system boot into windows and either A) downclock to 44( the default) or B)up clock to ram to 2133... or both note this is some oddity where bios setting are ignored or changed

For example I put in 46 multiplier, and bclk 100 , with mem at 1333///
can come out either of the above.. or blue screen

I'm hoping that this can be eliminated as Im good and stable at 4.5 and low temps and dont have the faintest idea whats happening here Ive done settings for the various power at "auto" no auto, with more or less power...
no idea why this is going on.. should I re install bios..????
fyi bios is f5 updated to F7 no changes

THEN
I realized in GB boards like mine if you leave Uncore at stock it auto boots it to 40, this was screwing with my settings ..
WOW how simple, things are better with higher uncore, still tuned down... that was a weeks worth of headbanging
Im over the wall and going
again thanks

thanks to Dave and all