Prime 95 BSOD

[Blind Krill 796]

I was playing around overclocking my I7-4790K seeing what i could get out of it and ran into some weird issues with Prime 95 while stress testing it.  I had a stable 4.8 Ghz at 1.29 V, i.e. ran cinnebench and games well without crashing, but for some reason as I try to run prime 95 the system BSODs.  So after this happened i started backing down the clock 1 Ghz at a time and retesting until eventually i just hit default and put it back to stock settings.  Now at stock settings i tried Prime 95 again and instant BSOD.  

 

Any idea why my system is doing this?

 

---

I7-4790k

EVGA GTX 970

32 Gb Corsair Vengance

Corsair H100i V2

[porina]

Prime95 stresses the CPU and also the ram subsystem. Check all those clocks, voltages, timings. Heat will also affect stability, but usually this isn't "instant" crash causing.

[Sack]

I don't think Prime95 is recommended to stress test CPUs anymore, if I'm not mistaken it can literally put your CPU on fire, but you know, you do what you want.

[Simon771]

Test it with Aida64 instead. I would stay away from prime95

[Silver47]

44 minutes ago, Blind Krill 796 said:

I was playing around overclocking my I7-4790K seeing what i could get out of it and ran into some weird issues with Prime 95 while stress testing it.  I had a stable 4.8 Ghz at 1.29 V, i.e. ran cinnebench and games well without crashing, but for some reason as I try to run prime 95 the system BSODs.  So after this happened i started backing down the clock 1 Ghz at a time and retesting until eventually i just hit default and put it back to stock settings.  Now at stock settings i tried Prime 95 again and instant BSOD.  

 

Any idea why my system is doing this?

Prime95 is just absolute trash. Rather use Intel XTU, Aida64 or similar for stability testing. Somehow Prime95 really fucks around with some automatic overclocking/stability tools and mechanisms in some manufacturers BIOSes, once you overclock your chip. I found that to be especially true with Gigabyte and Asus mainboards. I had an incident in which my overclocked 4770k suddenly started pulling way too much voltage and running scorching hot, but only when I ran Prime95 for testing. All other programs I used behaved just fine.

[Blind Krill 796]

5 hours ago, porina said:

Prime95 stresses the CPU and also the ram subsystem. Check all those clocks, voltages, timings. Heat will also affect stability, but usually this isn't "instant" crash causing.

Yeah i was running the CPU stress only in it.  I know it wasn't heat since I was watching Temps the whole time (under 70c on the cores and 40 on the package).  I'm thinking its just Prime 95 as some of the other guys suggested.  Gonna check out Aida 64

[Blind Krill 796]

4 hours ago, black0utm1rage said:

Prime95 is just absolute trash. Rather use Intel XTU, Aida64 or similar for stability testing. Somehow Prime95 really fucks around with some automatic overclocking/stability tools and mechanisms in some manufacturers BIOSes, once you overclock your chip. I found that to be especially true with Gigabyte and Asus mainboards. I had an incident in which my overclocked 4770k suddenly started pulling way too much voltage and running scorching hot, but only when I ran Prime95 for testing. All other programs I used behaved just fine.

I'll give Aida 64 a try.  I have an Asus motherboard so it might be the program being screwy if you have had similar issues.  Thanks

[porina]

4 hours ago, black0utm1rage said:

Prime95 is just absolute trash. Rather use Intel XTU, Aida64 or similar for stability testing. Somehow Prime95 really fucks around with some automatic overclocking/stability tools and mechanisms in some manufacturers BIOSes, once you overclock your chip. I found that to be especially true with Gigabyte and Asus mainboards. I had an incident in which my overclocked 4770k suddenly started pulling way too much voltage and running scorching hot, but only when I ran Prime95 for testing. All other programs I used behaved just fine.

Pure rubbish. If your system is unstable running Prime95, it is unstable, full stop. You can choose to run an unstable system if you want, and many people do by making the assumption they wont ever run anything else like Prime95 in normal use. Call it "stable enough".

 

Prime95 is userland software only. It doesn't attempt to or do anything to fiddle with hardware settings. If any hardware settings change, it is due to something else. All it does is run some finely tuned instructions which do real calculations. Look up FFTs if you want to know more. It just happens to extract a high level of work out of modern Intel CPUs, and thus run them hot while doing so. Expect overclocks to be significantly lower than if you don't care about it, that is normal.

 

BTW the only mechanism I can think of that could cause higher voltage than expected is load line calibration setting in bios. It is there to prevent voltage droops under load, but set incorrectly it could over-correct. Thermal runaway is a possibility also, where when things get hot, they take more power, and get hotter, and... you can see where that goes. That's usually a result of over-ambitious settings for the cooling ability of the system.

 

As for alternative tests, I found Asus Realbench in its stress mode to work well at finding instability quickly for systems that lack or are weak in AVX2 (celerons, pentiums, non-K OC, Ryzen). I don't find Aida64 to be that fast at identifying instability unless you're doing something really stupid.

[Blind Krill 796]

7 hours ago, porina said:

Pure rubbish. If your system is unstable running Prime95, it is unstable, full stop. You can choose to run an unstable system if you want, and many people do by making the assumption they wont ever run anything else like Prime95 in normal use. Call it "stable enough".

 

Prime95 is userland software only. It doesn't attempt to or do anything to fiddle with hardware settings. If any hardware settings change, it is due to something else. All it does is run some finely tuned instructions which do real calculations. Look up FFTs if you want to know more. It just happens to extract a high level of work out of modern Intel CPUs, and thus run them hot while doing so. Expect overclocks to be significantly lower than if you don't care about it, that is normal.

 

BTW the only mechanism I can think of that could cause higher voltage than expected is load line calibration setting in bios. It is there to prevent voltage droops under load, but set incorrectly it could over-correct. Thermal runaway is a possibility also, where when things get hot, they take more power, and get hotter, and... you can see where that goes. That's usually a result of over-ambitious settings for the cooling ability of the system.

 

As for alternative tests, I found Asus Realbench in its stress mode to work well at finding instability quickly for systems that lack or are weak in AVX2 (celerons, pentiums, non-K OC, Ryzen). I don't find Aida64 to be that fast at identifying instability unless you're doing something really stupid.

 

So i know its nothing in the BIOS since i put everything to default and tried to run Prime95 and got a BSOD.  I can't say its thermal runaway either, i have a H100 i V2 on my system and it stays less than 50C running cinnibench and everything else i throw at it. I'm going to try Aida 64, Asus Realbench, and pinning the CPU rendering a CATIA model and see what happens.  All my gameplay tests and benches operate flawlessly.

[Silver47]

16 hours ago, porina said:

Pure rubbish. If your system is unstable running Prime95, it is unstable, full stop. You can choose to run an unstable system if you want, and many people do by making the assumption they wont ever run anything else like Prime95 in normal use. Call it "stable enough".

 

Prime95 is userland software only. It doesn't attempt to or do anything to fiddle with hardware settings. If any hardware settings change, it is due to something else. All it does is run some finely tuned instructions which do real calculations. Look up FFTs if you want to know more. It just happens to extract a high level of work out of modern Intel CPUs, and thus run them hot while doing so. Expect overclocks to be significantly lower than if you don't care about it, that is normal.

 

BTW the only mechanism I can think of that could cause higher voltage than expected is load line calibration setting in bios. It is there to prevent voltage droops under load, but set incorrectly it could over-correct. Thermal runaway is a possibility also, where when things get hot, they take more power, and get hotter, and... you can see where that goes. That's usually a result of over-ambitious settings for the cooling ability of the system.

 

As for alternative tests, I found Asus Realbench in its stress mode to work well at finding instability quickly for systems that lack or are weak in AVX2 (celerons, pentiums, non-K OC, Ryzen). I don't find Aida64 to be that fast at identifying instability unless you're doing something really stupid.

Ok, that part about the BIOS was just an assumption based on stuff I experienced plus what I heard from friends with similar setups. I didn't really know if it could really overwrite your voltage settings or similar, but nevertheless it is up to this date the only stress testing program that instantly caused my system to run extremely hot and eventually BSOD. Neither Aida64 nor Intel XTU stress tests caused similar problems. Ergo somethings gotta be wrong with P95. 

[porina]

5 minutes ago, black0utm1rage said:

Neither Aida64 nor Intel XTU stress tests caused similar problems. Ergo somethings gotta be wrong with P95. 

Niether Aida64 or XTU stress as much. This does not prove anything is wrong with P95. I run software similar to P95 (same math library) 24/7 on all my modern systems without problem most of the time. Any problems are usually due to cooling degradation over time or optimistic OC.