IntelBurnTest - The new stress-testing program

[Konokrad]

Howdy.

I've noticed weird behaviour in IBT(both 2.4 and 2.5)

If I run ibt under safe mode, it shows one peak performance.
If I run it under normal mode, it shows another performance, lower by almost 10 GFLOPS.

Now, there is nothing to mess with calculations(I've killed all killable and non-essential processes).

My main Specs:
Win7 x64
4 GB DDR2 RAM @ 800 Mhz
Core2Quad Q6600 @ 3.0 Ghz

No matter how I tweaked voltages, the results were the same.
I guess the case is definitely not the voltages.

Oh, and I ran IBT in Xtreme Stress Mode.

[Konokrad]

What are the differences in Linpack parameters between stress-testing with and without "Xtreme stress mode" ?
I suspect it IBT sets realtime priority to Linpack with that flag.
Are there other changes?

[cbjaust]

hi, when I run ibt on my pc (Athlon 64 X2 6000+ @ 3420MHz) the program stops after about 20sec saying everything is fine but there is no data output. see pic:

Any ideas?

[Leeghoofd]

What happens if you lower the clocks ? Have you tried the LinX variant for a test ? But honestly , I have no idea why you want to hammer such an old setup...

[charden]

On a i7 860 with HT ON I test with 8 now correct since I'm using version 2.50?

[Konokrad]

Guess so.
AgentGOD, how about adding 24 threads(or the ability to manually set any number of them) ?
Would be particularly good for SR-2 owners with two 980X .

[overdoze]

Is this any better than LinX .. I only use LinX and not try this before

[Erklat]

Afaik it does the same job.

[j00ker85]

i'm having a strange error, it finishes in 0.11sec no matter what-
tried a fresh reinstall of win7, same happens! have .net 3.5.1 installed also vc++2005 and 2008, even 2010 !
it just says it finished no matter how much ram or rounds i give it...
legacy version and linx work great and i'm as stable as it can get
----------------------------
IntelBurnTest v2.50
Created by AgentGOD
----------------------------

Processor: Intel(R) Core(TM)2 Duo CPU E8200 @ 2.66GHz
Clock Speed: 2.67 GHz
Active Physical Cores: 2
Total System Memory: 3071 MB

Stress Level: Standard (1024 MB)
Testing started on 17/09/2010 11:46:36
Time (s) Speed (GFlops) Result
Testing ended on 17/09/2010 11:46:36
Test Result: Success.
----------------------------

[smut]

Love the program AgentGOD. Been using it for awhile but since I just updated from 2.3 to 2.5, I figured I'd finally give you props! Just installed a q8400 so stress testing my overclock. I literally just installed the CPU and started my overclocking adventure at 3.3ghz. Seems to be running stable using your program so far, so gonna bump it up 100mhz each time until I reach the limit. Hoping I reach at least 3.5ghz because that is what I ran my dual core at.

[MetaBaron]

i'm having a strange error, it finishes in 0.11sec no matter what-

Same here, though it's 0,44 sec for me

It's definitely a software problem (for me anyway), I restored my os drive from a five month old os-image, and then everything worked perfectly.

The strange thing is that the program runs correctly up to 4295 mb, 4296 and up finishes in 0,44 sec no matter what, and tells me that everything is great!

I7-920 @ 3.8, 12GB RAM, W7U-64

[Neova]

Same here, though it's 0,44 sec for me

It's definitely a software problem (for me anyway), I restored my os drive from a five month old os-image, and then everything worked perfectly.

The strange thing is that the program runs correctly up to 4295 mb, 4296 and up finishes in 0,44 sec no matter what, and tells me that everything is great!

I7-920 @ 3.8, 12GB RAM, W7U-64

This happens to me also but only when I OC a step up from the last stable setting. I thought it may be unstable settings but it does not crash OS. It just says it says completed in a few seconds.

[Cornelious0_0]

My initial impression is that it would be software related as opposed to a stability issue but I'm not sure. One would think if it were to fail or generate an error that soon you would be able to recreate some form of instability within Prime95 or other such testing.

[MetaBaron]

My initial impression is that it would be software related as opposed to a stability issue but I'm not sure. One would think if it were to fail or generate an error that soon you would be able to recreate some form of instability within Prime95 or other such testing.

In my case I have confirmed that it is a software issue. 5 month old os-image: Everything works as intended, Ran for about an hour w/10300MB allocated, no problem. Ghosting back to my latest backup: Fails immediately with the same memory allocation.

I'm guessing that there's a windows update or something like that, that messes up the memory allocation of Linpack on my current setup, since every single program I have tested that utilizes Linpack fails when allocating above 4200 and some MB: OCCT, LinX, IntelBurnTest.

Prime95 and other cpu-tests works fine on both the old os-image and current setup.

It is by no means a dealbreaker - but it would indeed be nice to know what's causing this behaviour. I can always revert to the old os-image if I need to run stability tests, but that's a somewhat tedious affair, in the long run.

Maybe I'll just reinstall W7, but that's quite a downer when you have configured everything exactly the way you like it

[TiesTorN]

i'm having a strange error, it finishes in 0.11sec no matter what-

I'm having exactly the same problem finishes in 0.11 sec and says "successful"... but you haven't even try?

[KURTZ]

right now IBT is more reliable than LINX?

[zalbard]

right now IBT is more reliable than LINX?

Well, this is just a user interface for Intel's Linpack, same as LinX... So it is a matter of preference which one you like most.
I seem to have less issues with LinX, personally.

[Zaskar]

Have a question about Xtreme Stress Mode, is it something that is stable enough to run for really long stability tests, like 150-300 passes? (Heat is well under safe, so thats not a concern)

Ive noticed that when I run Xtreme Stress Mode, I always end up having my computer lock or reboot after like 100-150 passes.

I can speed it up by running IBT for a pass or 2, then stopping it, it seems to free up more system ram that way (about 200-250mb more then a fresh run on initial bootup, windows offloading some stuff to the pagefile?). If I then run Xtreme Stress Mode again with the additional 200-250mb of ram available to it, it usually will lock up/reboot after 10-15 passes.

So could this be that Xtreme Stress Mode is locking away too much ram to keep the computer stable for long tests like that, and should only be used for 30min to 1 hour runs?

My OC might not be 100% stable yet but when your dealing with an error after that many hours/runs I figured id check.

[glen]

For use with Sandy Bridge you have to download the latest linpack_Xenon64 from Intel, then rename it Linpack_64 and place it in the LinData/64 folder of Intel burn test.

This raised my gflops from 47 to 83 with an i2500. I had been thinking my new cpu was no better than my old overclocked Q9550.

Would someone put this info on the first post?

[koltbuut]

Google this.

Intel® Math Kernel Library – LINPACK Download

rename and replace the original linpack64 with the linpack_xeon64 found in the latest Windows package

[OVERK|LL]

Excellent little app.

[stasio]

Google this.

Intel® Math Kernel Library – LINPACK Download

rename and replace the original linpack64 with the linpack_xeon64 found in the latest Windows package

Linpack , 10.3.3.006.
http://registrationcenter-download.i...10.3.3.006.zip

[ralliart12]

May I know if the figure(s) in the "Results" column matter when it comes to stability? i.e. I'm able to complete 25runs of IBT using different OC settings, but sometimes I get 2.something instead of the more "common" 3.something in the results column. Does that make a difference?

Also, since I can complete IBT on a few different set of OC settings, I should use the one that takes the shortest time per run & give the highest Gflops right?

[AgentGOD]

Updated to v2.51
- Linpack binaries updated + patched for AMD CPUs
- Official support for Six-Core CPUs and Sandy Bridge Architecture
- Added new algorithm to detect calculation errors faster
- Fixed a bug that caused blank results to be valid

[AgentGOD]

Hopefully this update helps address issues and concerns raised by you guys. Haven't had time to maintain the program until now. I appreciate the support and encouragement that you guys have given me