Realbench LuxMark-x64.exe stopped working

[madmeatballs]

When luxmark-x64.exe stops working can it be possible that my ram isn't stable? or this has something to do with GPU overclocks?
Although, I highly doubt it is GPU instability though since I passed Realbench 8hr stress with 1.4v DRAM Voltage. It is just that I am testing my ram because I believe it is faulty I keep on getting page fault/system PTE misuse BSOD when I power up my computer from a cold boot. G.Skill asked me to test at Bios optimized settings with XMP enabled and disabled. Well in this one I have XMP enabled. I just want to know if this LuxMark-x64 crash can also be caused by unstable ram.

[MMikhail]

Chino wrote:
Running the latest NVIDIA drivers and benchmarking in RB without a problem.

Benchmarks work, but stress tests fail after 1 - 2 hours.

[Brighttail]

Main thing is to make sure you disable any ramdisk/ramcache programs when running your benchmark tests. They can interfere.

[Arne_Saknussemm]

While there may be one in a million cases where some conflict is causing crashes...by and large if RB is crashing it is because your PC is not stable...either CPU or GPU overclock or RAM settings (XMP is overclocking too) or all of them combined...since that is exactly what this test does...real-world load on all subsystems at once.

Sometimes it's OS corruption that goes unnoticed running day to day or driver corruption that can go unnoticed too...

For 99.99999% blaming the test isn't the way to think about it.

[MMikhail]

Well, it looks like it's not a GPU or iGPU problem for me, since the OpenCL(Blender) test fails according to RB output.
And this test uses only GPU rendering. Disable each video and run the test on iGPU, on GPU and both. Result is the same, fail.
So it's either CPU, RAM or MB. Going to try in single dimm mode and see how it works out.

[MMikhail]

Arne Saknussemm wrote:
While there may be one in a million cases where some conflict is causing crashes...by and large if RB is crashing it is because your PC is not stable...either CPU or GPU overclock or RAM settings (XMP is overclocking too) or all of them combined...since that is exactly what this test does...real-world load on all subsystems at once.

Sometimes it's OS corruption that goes unnoticed running day to day or driver corruption that can go unnoticed too...

If it's not a hardware issue, but a software bug, should not it popup in regular day to day workload?
eg. the system can be running 24/7 for 30 days without any noticeable problems without rebooting.
So I would assume the OS is in very good condition.

Also weirdly, if you mix RB stress test with something simple as a video player playing some simple SD video (very low overhead)) together
stability improves considerably to almost 100%. So this raises one more question, does RB have 0% cpu loads during it's test, and we are getting the problem when cpu clocks down for power saving ? How to test this case and what to tweak in UEFI ?

[Arne_Saknussemm]

Hi MMikhail

My comments were not directed at anyone in particular. As I say some software conflict may be responsible for some cases....just making the point that generally this is not the case...

Put it this way...in the RB forum there are more threads questioning RB integrity than are real instances of RB integrity failing. The vast majority are unstable OCs or unstable systems.

I have had a GPU driver running and gaming for three weeks no problem and then not been able to run RB...which has prompted me to clean install driver and RB works fine.
And yes, OS can jog along day to day fine but be in a terrible state.

Instability can indeed occur in transition to lower power states....just an example being negative offsets that work fine at load but crash at idle.

I've found cache OCs to be stable for everything and then fail some RB runs....or Firestrike runs...

[MMikhail]

Talking of software conflicts...

Currently pinpointed a 100% reproduceable situation that cause instability warning failures.
And as it is hard to believe, it is default Win10 ScreenSavers, after it turns on, we fail a few minutes later.

Any ideas ?

[Arne_Saknussemm]

Well...not apart from the obvious...dont use screensaver while benchmarking.

New RB version out by the way...

[MMikhail]

You are referring to RB 2.53b (b for Beta?)

[Arne_Saknussemm]

Yes that's the one....not sure b means beta...more likely version letter because Nodens is working on 3 versions...