OFFICIAL: M17x-R2 Benchmarking Thread

[ahmadmud]

i think this is an interesting article regarding the degradation of cpus :

Intel's 45nm Dual-Core E8500: The Best Just Got Better - AnandTech :: Your Source for Hardware Analysis and News

[DR650SE]

Hmm the more I read, and look at my BSOD codes, it's starting to look like as bad core. Stupid benching I think it runs fine at stock voltage up to about 80W, anything higher, and it BSOD. Thinking I may have a bad thread. That would be two XMs I have killed, wow, i'm on an expensive roll. At least the work at stock right? Though I don't think I'd feel right ebaying them

I can stll run a wPrime1024 at 100W on 4 threads so I guess as far as real world goes, the CPU is plenty stable.

Edited March 15, 2011 by DR650SE

[StamatisX]

2 XMs killed? wow that's the cost of a laptop if you think about it, very expensive hobby you have indeed

[DR650SE]

yea it is. Well not killed completely thankfully, just past their prime , Still using the 940XM since it has the default 25x, and my system boots on average 11.5sec with 78-80 processes running at boot. 100% load will BSOD if the TDP is over 80ish, but I still run it at 90TDP since the CPU will never really see a 100% load on it.

[StamatisX]

you will be surprised to see how low you have to set TDP/TDC for prime95... I was shocked

[DR650SE]

Hmm, never ran that one, may have to give it a shot, I'm thinking I might need to just get a dice pot. That could make things interesting, than the increased voltage could work in my favor....

[Brian]

I don't think the CPU is dead or anything, you might just have to settle for stock TDP with it or slightly above it for stable speeds.

[DR650SE]

Well the only thing that makes it BSOD is wPrime at 80W+ TDP. So for everyday use, I still keep it at 90W with no issues, so not dead by any means. havn't really tested out the OEM 920XM, but the 940 runs fine stock voltage, 5% OC, 90W TDP. I can run a 50% load on wPrime1024/4 threads and thats fine, so for anything but benchmarking, it should be plenty fine. I might see if it BSOD while gaming. If it does, I still have a really solid 920XM ES I wasted way too much money

[mw86]

Intel Burn Test and Prime95 smallfft plus the benchmark tool thats built in to prime95 is great if your overclocking with different turbo frequencies with more or less cores active because it stress 1 core then keeps adding cores till the test is done great for showing a cpus ranges when not fully loaded over all 4 cores.are great for checking stability they push your speeds down a lot even at high tdp. At 5% Bios I've seen much less stability when multipliers are raised too much or tdp is raised up too much I've even got bsod's i believe mean the Intel memory controller speeds are too high. The base-clock effects the whole system. I've been trying to hone in on some numbers and found mine not to far off from stams. My heat is much higher i have yet to do a repasting but the numbers are valid. I found more stability from StamatisX's example of 3%overclock. Before I couldn't push the system as hard in 5% or 4%. So for my system as long as my total overclock is at < 3.57 and at 5% or 4% it is stable with tdp's high enough to maintain those speeds. But to reach 3.57 on stock volts on all 4 cores and the lower 3% was needed for me to get stability. Adding voltage for a small range gave stability at higher clocks on all 4 cores but temps are exponentially to high under full loads dropping my speeds down way farther than on stock volt. So unless benching no need to add voltage. You could get >3.6<3.7 4 core full load with 75mv. I think CPUZ is the only thing that benefits from the 150mv only due to it not putting stress on the cpu you guys can get 4.3+ghz validations. I've got BSODs with high overclocks that completely erase stuff that was in the memory at the time of the crash, preventing windows from starting or even causeing random BSOD's after. DR650SE could there be any system errors that arent showing up but could be causing your BSOD's early. I don't think your CPU is broken. I've had mine scorching hot on the keyboard and done 98tdp prime95 benches for almost 24hours. Though these tdp and tdc are linked to the wattage your cpu is getting i have noticed you can get amazing stability at lower clocks with extremely beefed up tdp. Like keeping 26, 25 or 24 turbo multipliers and cranking the tdp up. If the speeds seem low kick up the bios overclock but if you keep the max turbo frequency low enough you can use tdp and tdc to lock it in place no matter what load you have on it like having an unyielding 3.4ghz that wont budge on prime95 or intel burn test. DR650SE could you try doing the tests you just did with the same settings that StamatisX uses?

[ahmadmud]

these are some benchmarks for my 920xm I posted in a M17X i7 processors thread in NBR, might as well post it here since this is a benchmark thread :

tests using my 920xm with TS and a no factory o'c (bus speed is default), and here are some of my benches:

(Cinebenchmark 10.5 & 11 , 32&64bit), superpi, wprime, 3dmark vantage, just so you can compare with the benches at notebookcheck:

Intel Core i7 2920XM Notebook Processor - Notebookcheck.net Tech

Cinebench R10: Rendering Single&Multiple 32Bit :

ImageShack® - Online Photo and Video Hosting

Cinebench R10: Rendering Single&Multiple 64Bit :

ImageShack® - Online Photo and Video Hosting

Cinebench R11.5: Rendering Single&Multiple 32&64Bit :

ImageShack® - Online Photo and Video Hosting

Super Pi 1,2,32M: http://img535.imageshack.us/i/superpix.png/

wprime 32&1024M: http://img576.imageshack.us/i/1024mx25and32.png/

3DMark Vantage: P CPU no PhysX 1280x1024

ImageShack® - Online Photo and Video Hosting

now although I am comparing an unlocked 920xm vs. stock 2920, please note that these scores were achieved without any o'cing to the bus speed, many users out there who are looking to get the best out of there cpus have even way better scores such as this:

ImageShack® - Online Photo and Video Hosting

[mw86]

FurMark - OpenGL benchmark and VGA Stress Test - Scores | oZone3D.Net

[StamatisX]

many users out there who are looking to get the best out of there cpus have even way better scores such as this:

ImageShack® - Online Photo and Video Hosting

Check my signature for a better score than that

[StamatisX]

FurMark - OpenGL benchmark and VGA Stress Test - Scores | oZone3D.Net

I am on Linux right now but I am tempted to reboot on Windows just to test and see mine how high it can get.

[ahmadmud]

Check my signature for a better score than that

your gpu score is really great, but his cpu score hasn't and may never be beaten ...by another 940xm/920xm

[StamatisX]

your gpu score is really great, but his cpu score hasn't and may never be beaten ...by another 940xm/920xm

This score can be beaten, especially with the use of external cooling (my 20K score is without cooling, just room temperature). You are missing the point though, if you aim for a good CPU score you will never get a very good GPU score and the total score will suck, instead if you aim for the best GPU score then you have many chances to make a very good total score since Vantage is a GPU depended benchmark.

[mw86]

Help guys I can't figure out the power supply limits! I have 900mhz core 1200mhz memory set on the HD5870 crossfire @ 1.15volt. 90+ tdp stock volts on cpu. I run furmark bench at these graphic card speeds and the first one went fine the second i did the entire system shuts off my guess i am surpassing the powersupply's capacity. Any suggestions on what should be lower so I can survive? I am willing to drop speed on graphics cards, lower the 1.15 volts to those cards or even lower my tdp. Please if anyone has suggestions on what will bring me back into safe limits it would be greatly appreciated. Maybe list what causes this failure on your setup so those here know about the limits of the powersupply. I have the 220watt one thats marketed to the extreme processors on the R2.

[mw86]

Given the opportunity I will explain my point of view on that matter.

The reason we use thermal paste and we don't apply the copper directly to the CPU die is because the surfaces are not completely flat so they don't make perfect contact, therefore we don't have the maximum heat transfer, thermal paste fills those gaps, but if we add an extra layer of copper between the CPU die and the heatsink, for the reason mentioned above we would have to apply an extra layer of thermal paste between the copper and the heatsink, the more the gaps we have to fill and the more the thermal paste we apply, the less heat we dissipate, so it wont be that efficient compared to the retention mod with only the absolutely necessary amount of thermal paste between them which is only one layer. (as a general note, excessive amount of thermal paste will increase the temperatures instead of lowering them, it is there to fill the gaps, not to replace the metal contact)

What would be more efficient to do, is to somehow channel the heat to the chassis of the laptop since it is from aluminum, or I was also thinking to apply thermal pads between the heatsink and the magnesium cover (there is a point that those two are almost touching each other).

[ATTACH=CONFIG]303[/ATTACH]

I think somebody should try this I noticed some brown burn like marks on the aluminum part of the keyboard. I feel this could be solved by your suggestion of thermal pads where the copper heat sinks come close to contact of the magnesium chassis. When I can get paste and thermal pads I will order extra pads to try this.

[StamatisX]

I don't expect to see a dramatic drop in temperatures, but even a very small one is better than nothing...

About furmark and the psu, since you have a limit of 240W it's up to you how to balance your system, more towards gpu or cpu, like if you have the cpu at 90TDP you can't have 900MHz for the gpu, if it was for gaming i would prefer higher gpu clocks and a lower TDP since the cpu is not the bottleneck and for everyday usage I would keep lower or stock GPU clocks and higher TDP for the CPU, unfortunately you can't have them both very high.

[mw86]

Thanks for that clarification StamatisX. I guess I might try OCCT power-supply test to hone in on my limits or I may try Furmark again... or Furmark with a basic Cpu stress test at the same time. As for drops in temps I agree most likely very unnoticeable difference but possibly less heat lingering all around that copper tubing.

[Brian]

Help guys I can't figure out the power supply limits! I have 900mhz core 1200mhz memory set on the HD5870 crossfire @ 1.15volt. 90+ tdp stock volts on cpu. I run furmark bench at these graphic card speeds and the first one went fine the second i did the entire system shuts off my guess i am surpassing the powersupply's capacity. Any suggestions on what should be lower so I can survive? I am willing to drop speed on graphics cards, lower the 1.15 volts to those cards or even lower my tdp. Please if anyone has suggestions on what will bring me back into safe limits it would be greatly appreciated. Maybe list what causes this failure on your setup so those here know about the limits of the powersupply. I have the 220watt one thats marketed to the extreme processors on the R2.

The PSU should not trip with those settings as long as the CPU isn't being run at 100% on all 4 cores either simultaneously. If that isn't the case, then double check your temperatures to make sure they are okay. If it continues and the temps are fine, I'd ask dell for a new PSU--the delta ones are better than flextronics. I can run mine w/the above settings all day long in furmark. I've put on a sustained load that exceeded 240W and the PSU didn't trip.

Edit: I see you ran furmark BENCH. What if you run furmark burn in? I never run the bench so I'm not sure if it utilizes the CPU or not. Also which version are you using?

[mw86]

I use 1.90 Furmark and on the 1080P Benchmark setting im not sure which features are utilized during the benchmark. Either way Furmark puts quite a load on those cards mostly the heat being the prob. So noting what you mentioned Brian I will try the Burn in instead and I think it may be TDP is too high still for 900mhz crossfire like Stam mentioned.If the temps on the GPU's r too high will they throttle like normal or does rbe bios editing disable throttling. And if they throttle or not if the temps raise real high is there a thermal trip point for the gpus that shuts the system down like the cpu will do if its temp goes well over 100celcius.

Edited March 30, 2011 by mw86

[StamatisX]

I run metro 2033 with the cards overclocked @ 950/1220

at 1080p with everything on except advanced PhysX

at 1200p with everything on except advanced PhysX

at 1080p with everything on except advanced PhysX and default CCC settings

[mw86]

Nice Stam Those are some nice frame rates, tha'ts a beast of a game. Shocking though with those 3-5frames per second minimum on some of those runs. Seems like there isn't anything out right now that requires as much. Ill have to get it I've only heard good about it.

FurMark - OpenGL benchmark and VGA Stress Test - Scores | oZone3D.Net

900mhz core 1225mhz ram ccc default settings 97tdc eist and c1 e disabled 25x140-4/3core 26x140-1/2core

Result

900 core 1250 memory 97tdp eist and c1 e disabled 25x140-4/3core 26x140-1/2core

the 1250 doesn't prove to be stable consistently so ill use the 1225 and the 900, 910 showed signs of artifacts occasionally. If I use the lowest tdp necessary to not shutdown and i should be good for something like Metro 2033.

[StamatisX]

Result

900 core 1250 memory 97tdp eist and c1 e disabled 25x140-4/3core 26x140-1/2core

the 1250 doesn't prove to be stable consistently so ill use the 1225 and the 900, 910 showed signs of artifacts occasionally. If I use the lowest tdp necessary to not shutdown and i should be good for something like Metro 2033.

Now go to CCC and turn off tessellation

[StamatisX]

Another interesting result:

winrar -> tools -> benchmark (leave it for 1min) take a screenshot