Failing Prime95 torture test with the same error in the same spot (PC Crash)
 

Hello, I'm new to this forum, hope I'm in the right section, I apologize in advance if I'm not.

Going straight to the point I've had some crashes with my pc, NEVER under stress, always during normal usage/web browsing, and I've been suggested to test my ram with Prime95 Large FFTs.

With version 306b3 (downloaded some weeks ago) I keep getting the same error always on the first test within a few minutes, whatever test I run, and I never passed the first test.

I've read the stress.txt and it suggests that If your machine cannot generate correct
results, you have a hardware problem. HOWEVER, if you are failing
the torture test in the SAME SPOT with the SAME ERROR MESSAGE
every time, then ask for help at [url]httpmersenneforum.org[url] - it is
possible that a recent change to the torture test code may have
introduced a software bug.

With other versions of Prime95 running for hours I didn't get a single error (blend, large, small, smallest). I didn't get any error with Windows memtest or Memtest86 either, ran it for 4 hours session each time (limit of the free version) for 2-3 times.

Is it possible that I'm running into a Prime95 bug as the text suggests
Attacching a screenshot of two large ffts tests I ran just now.

Also I've had a system crash today after about 1 month since the last one (screen frozen, mouse and keyboard stuck while web browsing, OS crash after 10-15 seconds, had to hardware reset, temps are fine) not sure if I can discuss and try to find a solution on this forum, in this case I'll post my pc specs and be more specific about the issue.

Thanks.

[QUOTE=rapture237;580385]Hello, I'm new to this forum, hope I'm in the right section, I apologize in advance if I'm not.[/quote] This is a fine spot. Welcome to the forum.

[QUOTE]With version 306b3 (downloaded some weeks ago) I keep getting the same error always on the first test within a few minutes, whatever test I run, and I never passed the first test.[/QUOTE]

See number 10 [URL="https://www.mersenneforum.org/showpost.php?p=568252&postcount=2"]here[/URL]. There was a bug related to the torture test fixed in the version after you have.

Try downloading the latest version from here: [url]https://www.mersenneforum.org/showpost.php?p=568251&postcount=1[/url]

You should monitor your temps while running it.

[QUOTE=Uncwilly;580390]This is a fine spot. Welcome to the forum.

You should monitor your temps while running it.[/QUOTE]

Thanks! I might have accidentally posted twice (same title same text), delete the other one just in case.

Attaching a screenshot of all temps after 4 hours of Prime95 blend (no errors of course). It also includes full specs of my pc.
As you can see CPU temps don't go higher than 65°C.

Another crash today, happened while chatting on discord (previous one was while browsing web). Here's the bluescreenview dmp.

I have a question in case anyone knows: people have suggested to test using memtest86 running for at least 8 cycles but the current free version allows to run only 4 cycles (has been like this for a few years).
Does anyone know how to download an older version of memtest86 that allows the test to run indefinitely? Or do you think that running multiple sessions of 4 cycles each is reliable? Or maybe another software to use?

In your case, I'd suggest checking the GPU also! Your bluescreen sounds very much like a graphics or graphics driver problem.

[QUOTE=kruoli;580616]In your case, I'd suggest checking the GPU also! Your bluescreen sounds very much like a graphics or graphics driver problem.[/QUOTE]
Hey, thanks for the reply! How would you suggest checking the gpu? I've been playing lots of games (gpu intensive) the past 2 years and [B]never[/B] ever had crashes under stress. GPU temps under heavy stress for hours don't go higher than 65°C. And all my crashes were under normal pc usage. Tell me how to test the gpu and I will do it.

Update: another crash (3 in a row after 1 month without issues) while watching a youtube video: mouse and keyboard stuck > monitor goes black > audio keeps going > audio distortion (~15sec) > pc restarts.
Bluescreenview reported the exact same drivers above.

When having always exceptionally acceptable temperatures (e.g. < 80 °C) and the problem occured spontaneously, I would think that some component reached its end of life.

I even had this with a motherboard once. Everything else was okay after replacing it.

But let's start with one thing at a time:
For graphics, I would suggest mfaktc (this is a computation optimised CUDA program) and Firestrike. Under Windows, I use these and FurMark. [c]mfaktc[/c] has a built-in self test. If this fails, and if the occuring failures are not always the same, we have a strong point that your point of failure is your GPU.

In a lot of cases, there will be multiple programs that will not show you what is wrong with your hardware. I personally cannot help you with memtest. If you still want to test the CPU and memory subsystem, please also try y-cruncher with all tests enabled.

I have tested for about 20 minutes with furmark (all options checked for max stress) and no errors, GPU temps always below 73-72°C. (Side note: I've never seen anything make the gpu use this much power (240-250w) as furmark, and temps under heavy stress in games have never gone above 66°C even in summer).

I am using Windows 10 Pro by the way.

I'll try mfaktc tomorrow. My question is how long should I run these stress tests for? Furmark test runs for about 1 minute and stops, I ran it twice and then customized time for 10 minutes.

[QUOTE=kruoli;580624]I personally cannot help you with memtest.[/QUOTE] I have "solved" the issue with memtest86, I've found a post that says that all test cycles are the same so basically running 4 cycles and then restarting the test or doing 8 cycles should be the same, just a waste of time.

IIRC, the "big" self test of mfaktc might run an hour or a few (the command line option is -st2).

Having no abnormalities at all with Furmark might be a hint, but of course I can't say this for sure. When done with my suggested stress tests and still not having occured a specific problem, I'm interested in what others might suggest.

[QUOTE=kruoli;580629]IIRC, the "big" self test of mfaktc might run an hour or a few (the command line option is -st2).

Having no abnormalities at all with Furmark might be a hint, but of course I can't say this for sure. When done with my suggested stress tests and still not having occured a specific problem, I'm interested in what others might suggest.[/QUOTE]

I have run lots other gpu stress tests in the past and never had a crash/artifact/error, I want to make sure that is clear that all the crashes happend while watching youtube, browsing the web, using discord (normal usage) and never under stress. Also all the applications that I mentioned make use of hardware acceleration, I have tried disabling that now and see if it makes any difference.

Here's what I'm doing now: I have already run memtest86 for 8 cycles with no errors, I will run 8 or more cycles overnight.

Tomorrow I will try furmark for 1+ hour and then the other tests that you have suggested. :smile: