The Ultra-Stable PC clubhouse

[mastrdrver]

Add This

As promised.....one day late.

mastrdrver | i7 920 @ 3.8Ghz | 5+ Hours | HIS 4870x2 800/950| 2+ Hours

Now I'm off to see if 21x181 is any better.

 

[mastrdrver]

Update........that I forgot about.

181x21 isn't any better as I can use the same settings for 19x200. The 140mm on my NH-D14 is the only fan used. The 120mm wasn't mounted nor was any other fan.

BTW, I love the Noctua NH-D14 that I got. I tried running passive in the HAF case. I left the 140mm in the middle of the D14. Only other fans were the stock case fans. The 140mm rear case fan was still exhausting. Temps never exceeded 70C after 30 minutes of Prime95 large FFTs. Board gives ~1.12 vcore after vdroop. It is a very impressive cooler even more so after you consider you get 2 fans with the cooler. Nothing anyone else does with similar price and performance.

 

[mastrdrver]

So can we maybe update this thread for DX11 capable hardware? TessMark: OpenGL 4 GPU Tessellation Benchmark

 

[pantherx12]

So can we maybe update this thread for DX11 capable hardware? TessMark: OpenGL 4 GPU Tessellation Benchmark

*head explodes*

But it's open GL D:

 

[mastrdrver]

They have always been OpenGL though. Is there some other gpu stress tester that uses DX11 parts?

Anyway was messing around and forgot about it even though the system is right next to me.
Sadly the system was only able to increase temperatures from 22C ambient to 30C out the top of the HAF. Its a start, maybe move on to LinX and TessMark now to see what happens.

 

[Mussels]

*head explodes*

But it's open GL D:

DX11GL4 *boom*

 

[mastrdrver]

bahahahaha

We'll see.....

 

[LAN_deRf_HA]

I think this needs a revamp to be called "ultra stable". Furmark is great for heat, doesn't tell you much about stability, especially on newer cards. Same with linpack. I've had plenty of 94% memory tests pass on that and still fail the regular OCCT large data test.

My recommendation:

Core system:
Prime - 8 hours
Memtest86+ - 5 passes at least
IBT - over 90% mem usage for 20 passes
OCCT - Large data set for 3 hours

Graphics:
30 minutes of crysis gameplay, just letting it sit there idle doesn't have the same effect
OCCT GPU test - error scan checked, at least 25 minutes

Consistently I've found crysis will find an error when furmark didn't. Its remarkable how good of a stresser that game is despite lower heat and core usage.

You need true variety to claim you're actually stable, something few bother with.

 

[Bundy]

I think this needs a revamp to be called "ultra stable". Furmark is great for heat, doesn't tell you much about stability, especially on newer cards. Same with linpack. I've had plenty of 94% memory tests pass on that and still fail the regular OCCT large data test.

My recommendation:

Core system:
Prime - 8 hours
Memtest86+ - 5 passes at least
IBT - over 90% mem usage for 20 passes
OCCT - Large data set for 3 hours

Graphics:
30 minutes of crysis gameplay, just letting it sit there idle doesn't have the same effect
OCCT GPU test - error scan checked, at least 25 minutes

Consistently I've found crysis will find an error when furmark didn't. Its remarkable how good of a stresser that game is despite lower heat and core usage.

You need true variety to claim you're actually stable, something few bother with.

I agree with what you are saying, but there are so many possible errors that I guess Mussels settled on the ones most likely to fail and what covered a wider range of stresses.

On my system, heat was the limiting factor and linpack would be the one that failed, not prime. I can see however where other systems might be limited by other factors and therefore different tests being needed.

Welcome to the clubhouse! The party has kinda been over for a while but maybe some more people might rock on in.

Now the questions
What typically was the crysis error that furmark did not find?
You have three memory/NB tests there, do you use them for specific reasons?

 

[mastrdrver]

Consistently I've found crysis will find an error when furmark didn't. Its remarkable how good of a stresser that game is despite lower heat and core usage.

You need true variety to claim you're actually stable, something few bother with.

Have you ever compared 3D06 looped compared to Crysis?

Fwiw the DX11 mark thing isn't very stressful so far. I think some settings don't work too as I could change AA levels and there was no impact on fps.

 

[mastrdrver]

Need to swap in my two 5870s to see if I get the same numbers. Using the DX10 path I was able to load up both gpus and keep them around 100% the entire time. Using OGL 3.0 would only load one gpu. I used Afterburner to kick the clocks up on both gpus other wise they would just sit there at 500/507.

Kombuster: 1440x900 DX10, stress + post + 8x AA
P95 custom: FFTs 2048 to 4096 using 3072MB ram not run in place

 

[Hayder_Master]

this is my last test

 

[mstenholm]

I don't see many folders and crunchers in here. It's stable if it has run 24/7/30

 

[ERazer]

hmm gonna post my crunchers later on

 

[mstenholm]

hmm gonna post my crunchers later on

No need, you are a cruncher