Lenovo Y580 - maxing one CPU

[crazydawid]

Hello!
I have a Lenovo y580, GTX660m flashed bios and oc to 1290/2925 stable, temps under 80 degrees, i5-3210m, 8GB DDR3

In some games that cannot or dont want to use more then one core i have preatty bad performace, so im thinking, is it possible to set up just one, first most used core to highest turbo mode?
In throttlestop i can only set up my both cores to max 2,9T mode so its 2 x 2894mhz, is there any way to set up only one core in max 3,1 turbo mode?

Becouse games like Rome 2 are unplayable for me (i bought it 2 days ago, DAMN YOU STEAM SALE!), where minimal requirements is a Core 2 duo processor...

MSI tells me that game in big battles is loaded 50% on GPU, and its randomly jumping on my cores, 1st one being below 80% all time while fps are below 20 o.0

 

[P4-630]

I have an i7 4700HQ , I disabled the turbo and runs now at 2.4GHz and I can still run my games smoothly (such as GTA IV), are you sure that it's not your graphics card?

 

[crazydawid]

Cant be sure, problem is that a lot of games run perfect, witcher 2 maxed out 45+ fps, skyrim ultra 60 almost all time and then there is Rome 2.

 

[de.das.dude]

thermal throttling is your problem then. overclocking laptop stuff isnt a really good idea to be honest. they are already working at the limits of their thermal capabilities as is.

 

[crazydawid]

There is a lot of room for OC in a lot of laptop models.

My temps after 3 hours of Wicther 2 are never more then 80 degres, so I have huge reserve till throttling

 

[P4-630]

@up
Dont post if have small or none idea what are You talking about.
There is a lot of room for OC in a lot of laptop models.

My temps after 3 hours of Wicther 2 are never more then 80 degres, so I have huge reserve till throttling

Check your PROCHOT value in throttlestop, mine is -15 which means mine will throttle at 85 degrees, this value is different for each manufacturer.

http://www.techpowerup.com/forums/t...4700hq-with-throttlestop.207621/#post-3201277

 

[P4-630]

I'm not sure if a laptop processor can run on it's highest turbo speed on one core all the time, but set your power scheme on maximum performance.
With Intel XTU you can set the multipliers of each core.

 

[unclewebb]

The Core i5-3210M will use the 31 multiplier when a single core is active and will use the 29 multiplier when both cores are active. It does this automatically. If you are playing a game that is mostly single threaded then your CPU will automatically spend a large percentage of time using the 31 multiplier. It is not possible to use the 31 multiplier when both cores are active.

ThrottleStop does a great job of tracking the multiplier as the CPU constantly adjusts how many cores are active. Turn on the Log File option in ThrottleStop, go play a game for a while and then exit ThrottleStop and post the log file somewhere convenient so I can have a look. Use www.pastebin.com or attach a log file to your next post. Play long enough so your computer goes through an episode of lagging. A log file that covers 15 minutes or 30 minutes while gaming is a lot more useful than a log file that only covers 30 seconds. Turn on Nvidia GPU monitoring in the ThrottleStop - Options window before you start logging so that info is included in the log.

Some Lenovo laptops like my Y510P have hidden features that reduce CPU performance when gaming. A log file will help get any problems like this sorted out. Also post a screenshot of how you have ThrottleStop setup so I can have a look.

ThrottleStop 7.00
https://www.sendspace.com/file/oquhg3

 

[crazydawid]

Hi unclewebb!
Here is my log file from 40 minutes of Rome 2, campain map and campain battle:
http://pastebin.com/PVnkTDAt

And my Throttlestop 7.0 settings:
http://i.imgur.com/0167DjG.png

Hope thats all data You need ?

 

[unclewebb]

The amount of data in your log file is perfect. I will start with some background info.

Many monitoring apps do not give you an accurate look at what your CPU is doing internally. A Task Manager graph or similar that is sampling your CPU once every second can be misleading when tasks are being scheduled from core to core in your CPU hundreds of times a second.

ThrottleStop does things differently. It uses high performance monitoring timers within your CPU so it can determine exactly what your CPU is up to. It does not depend on any inaccurate Windows API functions. It gets its performance monitoring data directly from the CPU using methods recommended by Intel.

The CO% in the ThrottleStop log is a measure of what percentage of time your CPU needed to be in the C0 state processing data. When a CPU has nothing to do, it usually drops down to a low power state like C1, C1E, C3 or C6. It does this continually even while you are gaming. Each core and each thread continually bounces back and forth between these various C States.

A Core i5-3210M is a dual core CPU and also uses hyper threading so it can process 4 threads at a time. When you run a single threaded application, this will keep one thread busy in the C0 state the entire time that app is running. 1 out of 4 threads would always be busy so the C0% for this app would be 25.0%. Windows also has 500+ threads of background junk constantly running and this typically adds another percent or so to the total percent of time a CPU needs to be in the C0 state. If your computer was mostly idle and you ran a single threaded benchmark such as Super PI Mod or even a single thread of the built in TS Bench, you should expect to see a C0% of about 26.0%. If you have a lean system with very little background junk running you might see slightly less than that. C0% should not go under 25.0% since it needs at least that to run the benchmark. If it goes way higher than that then either you have a pile of junk running on your system or your anti-virus program or something else decided it was time to get to work.

When people claim that their games are single threaded, I look at the C0% data and I can see if that is accurate or not. I put your log file into Excel and I grabbed 400 lines of data while you were gaming and the average C0% was 40.14%. That means that Rome 2 is not as single threaded as you think. With your CPU, the first 25% of time that the CPU needs to be in the C0 state can be handled by the first core. As soon as you start going beyond this, then the CPU brings the second core out of a low power C State and puts it to work. The CPU does this because it makes more sense to start using an under utilized core than it is to start using hyper threading. You could use the Task Manager and you could force this game onto a single core but the end result will likely be a decrease in performance. A hyper thread is not an entire core. It is only a partial core. Forcing everything onto a single core will hurt performance. Even if it is not perfect, I tend to trust the scheduling algorithm in the CPU to maximize performance.

During the chunk of log I looked at, the average multiplier was 29.63. That confirms that part of the time the CPU was using a single core and the second core was in one of the deep low power C States like C3 or C6.

( 29 X 0.685 ) + ( 31 X 0.315 ) = 29.63

That calculation is just a rough approximation but it shows that your CPU was using the single core multiplier (31) about 31.5% of the time while playing this game. That is really good considering that Rome 2 is not a single threaded task. This also confirms that this game did not spend all of its time only using a single core. 68.5% of the time the CPU was using both cores. The maximum multiplier when that happens is 29. The random squiggly lines in Task Manager usually don't show anything like that because the Task Manager is not sampling fast enough to be accurate.

The short version of this long story. Your CPU is running perfect and it is physically not possible to get anything more out of it. The CPU and GPU temps averaged about 70°C during that part of the log and peaked at about 75°C over the entire log which is excellent compared to some never laptops that typically hit over 90°C these days. There were no signs of any CPU performance throttling issues and Intel Turbo Boost was working just as Intel intended.

The CPU is good so I would suggest that you do some logging with GPU-Z or MSI Afterburner. A 1290 MHz overclock on a GTX 660M is aggressive so you need to make sure that your GPU is not having a fit while gaming and dropping down to a lower power state. If the bios you flashed includes some extra voltage for the GPU, perhaps power consumption of the GPU is triggering the GPU clocks to drop back. I am interested to see what you can find out.

 

[crazydawid]

Hey uncle!
Thanks for Your reply.

Out of sudden I make the game runing switftly.
I dont really know how it is possible, but changing GPU OC from 1290/2925 to 1265/3000 make it performing better.
And yes, the bios is a bit overwolted, becouse without it OC stuck at 1200/2800. ( voltage is changed from 1.1 to 1.150)
Funny thing, that most games gave me more FPS when 1290/2925, but Rome 2 is diferent

Additionally I installed some windows updates, now Argus monitor is showing me that one of the cores is going to 31 multiplier more offten than before.
Also I went to BIOS and change the ''Long Duration TimeWindow'' from ''28'' to the maximum of ''56''
This maximises the stock Turbo Time Limit and gave me performance boost.

Cheers !

 

[rtwjunkie]

You definately needed the longer turbo time. Total War games have always been extremely CPU-intensive, and on a single core mostly. Lot's of CPU power more than anything else is what you need for a TW game, no matter how nice they make the graphics.

 

[unclewebb]

Your log file showed CPU power consumption peaking at about 16 watts which is well under the Core i5-3210M TDP limit of 35 watts. That means that changing the turbo time limits will not make any difference because you were already getting full turbo boost.

I have not used Argus monitor for a while. As I mentioned, most monitoring apps are way too slow to react to what your CPU is really doing internally so try to ignore the fancy colored graphs. Your CPU is constantly switching between the 29 and the 31 multiplier hundreds of times a second. Any software that reports a snapshot of that can totally miss what the CPU is really doing. Kind of like going to the drag strip and trying to take a picture of a funny car at the end of the quarter mile. Most of the pictures you take are going to be blank.

Different games use different parts of the GPU. Perhaps 1290 MHz is not fully stable so performance starts to decrease when you push the GPU that hard. Even if high GPU MHz doesn't crash a game, the GPU might be spending time internally doing error correction when you overclock too much.

Enjoy Rome 2.

Total War games have always been extremely CPU-intensive, and on a single core mostly.

That was already proven to be false. Approximately 68.5% of the time his CPU was using both cores. The single core stuff is a myth and in his situation, the longer turbo time does not do anything. It was not a limiting factor so increasing it changes nothing.

 

[de.das.dude]

There is a lot of room for OC in a lot of laptop models.

My temps after 3 hours of Wicther 2 are never more then 80 degres, so I have huge reserve till throttling

1. laptops are meant to be portable not performance machines. you think you can overclock them just because they have a good cpu / gpu chip but those just one of the few variables when overclocking. Laptops are designed to a cost and have lower tolerances in every aspect possible because they are meant to be run with a fixed set of hardware.
This means that power delivery systems are tailored specifically to be enough for delivering enough power to run stock clocks, with a bit of tolerance to account for aging.
Sometimes thermal design has a slightly higher leeway as its cheaper to use a generic solution for all platforms rather than custom solutions for each model. However you will be given the shorter stick as soon as you pick high end hardware.

2. Laptop h/w is design to work better with power fluctuations, so even if you are not giving proper power to an overclocked entity, it will automatically lower its clocks to be stable. This is an advantage that comes directly from making laptops efficient.

3. Just because one temp probe at a very specific location says that its 80C does not mean that its not more than 80C on some discrete component. This is the most basic of knowledge that one knows when overclocking.

the fact that you yourself said that its working better after you downclocked your GPU shows that i was right all along.

 

[rtwjunkie]

Looks like we need @crazyeyesreaper regarding the cpu use of TW games.

 

[crazydawid]

I think that Rome 2 is able to use more cores.
Friend of mine have i7 CPU on the same laptop model, and have way better perfromance than I do.

In few weeks i will change my i5 to i7-3610QM when I will find one a bit cheaper (used one maybe), and I will give You guys know about TW quad core performance

 

[rtwjunkie]

Back to topic, I think that Rome 2 is able to use more cores.
Friend of mine have i7 CPU on the same laptop model, and have way better perfromance than I do.

In few weeks i will change my i5 to i7-3610QM when I will find one a bit cheaper (used one maybe), and I will give You guys know about TW quad core performance

Because it is FASTER. More cores is of little help in TW games. CPU speed above all else is important.

 

[MxPhenom 216]

I think that Rome 2 is able to use more cores.
Friend of mine have i7 CPU on the same laptop model, and have way better perfromance than I do.

In few weeks i will change my i5 to i7-3610QM when I will find one a bit cheaper (used one maybe), and I will give You guys know about TW quad core performance

Total War games use the same engine, and it is not multi threaded. It is a VERY single threaded game. @crazyeyesreaper can explain that more if he wants.

Reason your friends i7 runs faster is because it has a higher clock speed.

 

[erocker]

big words from someone who doesnt know simple grammar and comes to ask for help here in the first place.

Don't report posts claiming rudeness when you post garbage like this to begin with. Keep your posts directed towards the OP and/or topic at hand.

Everyone keep on topic.

 

[unclewebb]

I think that Rome 2 is able to use more cores.

I think you are right.

Total War Rome II - Performance Testing
http://www.overclockers.ru/lab/55992_2/Total_War_Rome_II_testirovanie_proizvoditelnosti.html

That site is in Russian but they did some thorough testing of Rome II so if you are interested in the facts, head to Google translate, copy and paste that link and check it out.

They tested a wide variety of processors and varied things like hyper threading on and off, Turbo Boost on and off, overclocking, dual cores vs quad cores, etc.

Huge CPU overclocks of 50% only made a difference of a couple of frames per second in the minimum and average frame rates.

Going from a hyper threaded dual core ( 2 real cores + 2 virtual cores ) like crazydawid has to 4 cores made a much bigger difference. The minimum frame rate increased by 50%. That is the kind of difference that will make a game much smoother to play. With the 4 core processors, enabling hyper threading to go beyond that made virtually zero difference. Their conclusion is that Rome II appreciates an Intel quad core processor.

If Rome II was only a single threaded game then you would not be seeing any significant increase in performance going from a dual core to a dual core with hyper threading to a quad core. The significant increase in performance confirms that Rome II is a multi-threaded game. I have no idea if it is well programmed or not but I do know that it likes 4 real cores.

 

[rtwjunkie]

Have you played it yourself, extensively, and monitored your own core usage? For the record, The TW engine has not been extensively updated since the RTW days. Through all the games since then, included. How many of those quadcores were running faster than the dual cores? And I said it is MOSTLY single thread, not entirely.

To the OP, PM @crazyeyesreaper, ask him to explain the ins and outs of the TW engine. Few people know it better than he.

 

[crazyeyesreaper]

Fact built in Rome 2 benchmark = worthless for testing CPU performance.

Fact Intel dominates in Rome 2

Fact Clock speed matters in Rome 2 so does memory speed and cache performance.

Game is single thread heavy it spawns many threads but 1 thread always gets overloaded causes a bottleneck.

Rome 2 and Shogun 2 function similarly

Shogun 2 actually featured a CPU benchmark test for a worst case senario aka large Melee battles. The game engine used for Empire / Napoleon / Shogun 2 / Rome 2 is fundamentally the same. It was a game engine meant for musket fire and limited melee. It struggles with large numbers of soldiers in Melee due to far more complex animations that need to all be synced and are loaded onto a single core.

this image explains better than I care to Notice the graphics test showed no real change notice CPU limited aspect changed with just improved memory speeds.

You can see clock speed performance change here:

http://www.bit-tech.net/hardware/2013/06/01/intel-core-i7-4770k-cpu-review/6

This still applies to Rome 2 Total War as well.
No matter the CPU Rome 2 will hit 100% on a single thread causing a performance cap in large melee battles which is the bread and butter of the game.

 

[unclewebb]

In few weeks i will change my i5 to i7-3610QM when I will find one a bit cheaper (used one maybe), and I will give You guys know about TW quad core performance

I think you will see better performance when switching from a hyper threaded Dual Core CPU to a hyper threaded Quad Core CPU. If you want to do a fair comparison, use ThrottleStop to slow down your new Core i7 so it runs at the same speed as your old Core i5. If your game plays better on your Core i7 at the same speed as your Core i5 runs at, that would prove that the extra cores really do make a difference. When you get your new CPU send me a PM if you need some help with this test.

If anyone else has this game and a Quad Core CPU, go into the bios and turn off 2 cores and you can do some testing that way to see if extra cores makes a difference or not. You can also try using msconfig to disable half of a Quad Core CPU for testing purposes.

How many of those quadcores were running faster than the dual cores?

The first test on the Russian website I posted equalized all of the CPUs at 3000 MHz so they could do a fair, apples to apples, comparison. Their testing showed a clear improvement when going from 2 cores, to 2 cores with hyper threading to 4 cores. The Russian testing showed no performance increase when going from 4 cores to 4 cores with hyper threading and the bit-tech graph showed the same thing.

The bit-tech graph does not show any Dual Core CPUs with hyper threading. It does show larger percentage increases in FPS compared to the Russian testing when overclocking from 3.4 GHz to 5.0 GHz but this is irrelevant to crazydawid. He has a laptop and is considering getting a Quad Core CPU that in game will run at most 10% faster compared to his Core i5. Just an opinion but I think a 100% increase in the number of cores is going to make a bigger difference in FPS compared to a 10% increase in MHz.

The graphs that show the load concentrated on the first core or first thread are interesting. I am more of a programmer than a gamer so I decided to modify my i7 Turbo GT program that I wrote a few years ago. It was just a quick edit but now it is able to precisely log the load on each thread.

i7 Turbo GT - Version 1.4
https://www.sendspace.com/file/70ciqi

If anyone that plays Rome II wants to do some testing, run i7 Turbo GT and check off the Log option. Go play the game for a while and when finished, exit i7 Turbo GT and attach the log or post the data somewhere like www.pastebin.com

By monitoring each thread, it should be easier to see how bad this overloading the first thread problem really is.

Slightly off topic but here is a look at the data this program can record.

   DATE      TIME    PEAK  CMULTI STDEV  LOAD%  C0%_0  C0%_1  C0%_2  C0%_3  C0%_4  C0%_5  C0%_6  C0%_7  NOTES
12/03/2014 23:57:21 33.393 32.948 0.370   1.14   4.26   0.26   0.54   0.75   2.01   0.19   0.59   0.54 
12/03/2014 23:57:22 33.665 33.248 0.262   0.30   0.51   0.07   0.48   0.27   0.43   0.05   0.28   0.29 
12/03/2014 23:57:23 33.614 33.221 0.274   0.47   0.66   0.08   1.62   0.19   0.30   0.04   0.33   0.56 
12/03/2014 23:57:24 33.803 33.241 0.364   0.38   0.55   0.05   0.33   1.16   0.38   0.12   0.22   0.25 
12/03/2014 23:57:25 33.546 33.126 0.276   0.55   0.90   0.18   0.51   0.59   0.66   0.21   0.49   0.86 
12/03/2014 23:57:26 33.706 33.204 0.259   0.32   0.59   0.07   0.30   0.16   0.48   0.16   0.40   0.39 
12/03/2014 23:57:27 33.697 33.333 0.184   0.31   0.55   0.05   0.35   0.16   0.29   0.22   0.53   0.32 
12/03/2014 23:57:28 33.782 33.267 0.254   0.30   0.48   0.06   0.33   0.18   0.46   0.13   0.45   0.26 
12/03/2014 23:57:29 33.698 33.192 0.269   0.27   0.57   0.08   0.31   0.13   0.33   0.04   0.44   0.29 
12/03/2014 23:57:30 33.742 33.309 0.273   0.38   0.77   0.12   0.51   0.41   0.53   0.08   0.35   0.26 
12/03/2014 23:57:31 33.760 33.302 0.205   0.28   0.52   0.04   0.36   0.15   0.50   0.06   0.37   0.26 
12/03/2014 23:57:32 33.248 32.744 0.541   4.45  18.32   0.92   1.82   1.04   3.57   0.39   5.98   3.59  Move_Mouse_Test
12/03/2014 23:57:33 33.499 32.806 0.556   7.60  29.68   2.53   2.34   1.93   5.12   0.42   9.60   9.17 
12/03/2014 23:57:34 33.688 32.900 0.571   5.58  25.68   1.89   2.69   1.80   4.66   0.51   1.48   5.91 
12/03/2014 23:57:35 33.553 33.082 0.317   6.78  27.58   2.01   3.78   2.45   2.26   0.45   1.79  13.94 
12/03/2014 23:57:36 33.434 33.052 0.305   7.02  27.28   1.70   4.30   2.42   2.15   0.83   4.22  13.24 
12/03/2014 23:57:37 33.182 32.664 0.327   7.25  26.37   1.27  15.18   3.04   2.02   0.18   1.13   8.81 
12/03/2014 23:57:38 33.189 32.925 0.138   3.54  11.76   0.51   6.49   2.41   1.10   0.27   1.68   4.07 
12/03/2014 23:57:39 33.674 33.242 0.254   0.51   0.50   0.24   0.42   1.41   0.37   0.11   0.49   0.54  Test_Finished
12/03/2014 23:57:40 33.744 33.322 0.238   0.33   0.67   0.07   0.26   0.38   0.34   0.18   0.41   0.30 
12/03/2014 23:57:41 33.662 33.285 0.189   0.29   0.54   0.06   0.36   0.21   0.44   0.12   0.32   0.26 
12/03/2014 23:57:42 33.562 33.172 0.199   0.35   0.64   0.12   0.48   0.21   0.53   0.15   0.36   0.31 
12/03/2014 23:57:43 33.711 33.285 0.218   0.26   0.43   0.05   0.44   0.15   0.34   0.12   0.31   0.26 
12/03/2014 23:57:44 33.835 33.285 0.252   0.29   0.59   0.06   0.42   0.20   0.33   0.12   0.36   0.26 
12/03/2014 23:57:45 33.662 33.302 0.282   0.42   0.61   0.07   0.23   0.41   1.12   0.12   0.48   0.35 
12/03/2014 23:57:46 32.958 32.294 0.587   1.47   2.62   0.49   1.05   0.62   1.11   0.37   4.48   1.04 
12/03/2014 23:57:47 33.741 33.225 0.331   0.27   0.65   0.07   0.25   0.26   0.31   0.05   0.29   0.29 
12/03/2014 23:57:48 33.756 33.249 0.270   0.28   0.62   0.06   0.33   0.22   0.35   0.05   0.32   0.30 
12/03/2014 23:57:49 33.722 33.261 0.251   0.28   0.57   0.05   0.32   0.17   0.51   0.06   0.26   0.27 
12/03/2014 23:57:50 33.646 33.429 0.178   0.48   1.34   0.09   0.38   0.36   0.97   0.13   0.33   0.26 
12/03/2014 23:57:51 33.751 33.326 0.189   0.28   0.50   0.08   0.34   0.19   0.44   0.12   0.27   0.26 
12/03/2014 23:57:52 33.820 33.298 0.290   0.26   0.42   0.05   0.23   0.12   0.58   0.13   0.27   0.25 
12/03/2014 23:57:53 33.735 33.299 0.199   0.26   0.59   0.06   0.25   0.14   0.38   0.11   0.33   0.26

You can see how before I start moving the mouse on my laptop, the CPU is very calm across all 8 threads.

Simply using the track pad and moving the mouse around in circles creates a huge CPU load which is concentrated on the first thread and there is also a lot of load that ends up on the last thread. This mostly proves that Elan needs to get their head out of their butt and release some better drivers that do not use the CPU excessively to poll the track pad. It should not take billions of CPU cycles to determine if someone moved the mouse using the track pad.

Hopefully this tool might help someone understand Rome II better. It would be even better if the programmers of Rome II used tools like this so they don't lock everything to the same core. That is a sign of old school programming from the single core era.

 

[rtwjunkie]

Unsubbing. Unclewebb, just because you are brilliant enough to create a useful program, by no means qualifies you to give "facts" on a game engine design that apparently you know little about. This actually translates into you giving poor advice to the OP.

The fact that by your last post sound, you seem to be dismissing the overloaded single core FACT, by saying maybe we will see how bad it really is, indicates you think years of first-hand experience counts for nought. Finally, none of us has said that the game is not multi-threaded, which you have chosen to ignore, just as you have chosen to ignore many years of experience with this game engine and its mechanics.

This means you are giving bad advice to the OP, by making him think that JUST because he goes to a quad on his laptop he will see a world of difference in Rome 2. He won't. He will see some, and if he had the ability to really overclock the thing, which he won't, because he is on a laptop, he would see even more difference.

My system gets brought to a crawl with 4 armies at a time in a siege, with naval troops landing. If you'd care to run that scenario yourself you'd see that one core is completely maxed out, and the others are lazily on a Sunday stroll. You'll find the same situation at 3.5Ghz and 4.1 Ghz, although your performance will have increased a little bit at 4.1.

Really, the OP deserves better advice than yours.