• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

What are the odds of an actual program/game causing as much heat as prime95?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Brando

Member
Joined
Jan 9, 2006
Sorry if this has been done to death but here's the thing. realbench and aida64 are ok for showing extreme instability but even after using them i was still getting random instability issues despite passing multiple runs. using prime95 on my 5820k will show instability almost instantly and if i can pass prime95 for 5-10 minutes i don't seem to get any random issues whatsoever. i know people say don't use it on cpus with avx and i can see why since the heat instantly spikes 20c over realbench stress test at the same cpu speed and voltage. my question is this. why is the avx feature there if it's not supposed to be used? isn't there a chance of running something at some point that heats up just like prime95 and roasts your cpu when you thought you were safe? it seems like prime95 is just telling the ugly truth and nobody wants to know but i really hope i'm wrong because ever since i faced my side fan inwards and got a video card with an exhaust style blower i have about 10c extra temp headroom to overclock with which lets me (hopefully) finally get my full oc back and full xmp memory speed. realbench stress test just ran 15 minutes with 4 hash matches at 4375mhz 1.265v and hovered around 67-72c. before this i was getting to 79-80c max at 1.23v since my airflow wasn't as good as i thought. then i run prime95 large ffts and hit 88c in less than a minute and blue screen even though cooling is much better now. i don't know what to believe but i want to think i can leave it like this and nothing will ever be as hot or brutal as prime in actual usage. what do you think?
 
OP
Brando

Brando

Member
Joined
Jan 9, 2006
well that was a bad idea. i think playing with prime for just a few minutes actually degraded my cpu. i just said screw it and went back to my old conservative oc of 4300mhz at 1.23v and it's not stable with real bench any more. that is wild. i really hope there's nothing out there that people use for legitimate purposes for long periods of time that work the system like prime. you might come back to a molten lump of a cpu. i should have listened lmao. gonna run 1.235v and hope it's enough for my old oc. i wouldn't have thought being at 88c for a few seconds would cause problems. i had my finger on the button the whole time.
 

Dermen

Member
Joined
Dec 23, 2001
Location
Florida
I've been a little confused by it as well since I started OCing my 6700k. For me P95 Small FFT has the highest temps, but it is only 7C higher than intel burn test. P95 blend is 15-20C lower that small FFT.

I don't know if it is just AVX that is causing the heat or if other stress tests are testing AVX. I've read older posts where people said it didn't matter since nothing used AVX, but AVX is being used now so it matters. I've only read a little bit about it so I still don't know.

Another reason people stopped use P95 is with Haswell it would increase your voltage a lot if you were using adaptive/offset mode. From my tests on Skylake P95 small fft reports lower voltage than P95 blend (IBT shows same voltage as P95 blend).
 

wingman99

Member
Joined
Dec 10, 2003
I use Prime95 for stress testing. With Sky Lake using Adaptive core voltage, my voltage peaks the same sometimes gaming as it does running prime95. There was a review that I saw a while back showing what programs use AVX like prime95 close to the same temperatures. I can't find that review now.
 

mackerel

Member
Joined
Mar 7, 2008
At least in the cooler months, I do run on all my Haswell or newer systems software that is comparable to Prime95, 24/7, overclocked. The other software actually uses the same math library as P95, but there are also examples which don't. This is also in part why my overclocks don't go anywhere near what other's do. My Skylakes I leave at 4.2 GHz 1.25v set, since the increase in voltage to go further really turns up the heat for not a lot of gain.

As for temps, do note that small FFT will fit on the CPU cache, thus the compute units can work near enough flat out. I never investigated if there's a difference in heat/performance between L2 and L3 cache size work... maybe another day. If you go to blend, it will mix in large tasks. For practical purposes, these will likely be ram bandwidth limited on faster CPUs with 4 or more cores, thus the compute units can't reach their full potential. This will still be useful as it could expose weakness in the memory system.
 

Kenrou

Member
Joined
Aug 14, 2014
Mass Effect: Andromeda is been getting really close to my Prime95 temps (within ~3c), but might just be because i got a crappy GPU and the CPU is working overtime to compensate :p
 

Kenrou

Member
Joined
Aug 14, 2014
1080p (1.0 ratio) is hardly "low resolution". Demanding game, uses all 8 cores 80%+ (seen it hit a perfect 100%) and the GPU 80%+ (depending on settings).
 
OP
Brando

Brando

Member
Joined
Jan 9, 2006
are most people still using aida64/realbech stress test temps of 80c or less as an acceptable heat limit for 24/7 use? i feel like i can make it to the next mhz level just barely if i use 1.28v for 4375 or 4400mhz. i just have the lingering doubt that i'll run into some random program that will run as hot as the new prime95 and F my S up. i tried folding at home to test it out but when i set it to use the cpu at max it just stops using the cpu and uses only the video card. i don't have mass effect andromeda to test. i'm starting to think i may even be able to go for 4500mhz with 1.3v since temps are so much better. on everything except prime that is. by the way i'm NOT using adaptive voltage. this temp spike is with locked voltage so that's not the problem.

EDIT: google says h265 encoding uses avx and im compressing a bluray now. cpu is around 80-85% usage and 60-65c. i cant find anything in setting to enable avx so maybe it's automatic?
 
Last edited:

caddi daddi

Godzilla to ant hills
Joined
Jan 10, 2012
I would guess your need for a vcore increase is due to some os corruption so you might try a reinstall.
I have some enterprise software that will approach the load of p95 but, that is few and far between.
one of the CFD programs I use will bring my 9590 rig to it's knees with heat and the highest clock I can run, now that winter is over, is 4.5 even with the 4x120x60mm rad.
the cooling system takes about 2 hours to heat soak and it settles at about 60c for the most part.
 

bruceGH4

Registered
Joined
Jan 17, 2017
Location
North Vancouver
I was having the same issue with Prime95, which I now consider to be the "hard" test for stability. There is a setting in my BIOS that will detect when avx is present and it will automatically knock down my multiplier. It's like a multiplier offset. So I can run at 4300mhz and I set it to drop to 4200mhz when AVX is present (which is my best stable o.c. In Prime95). I verified that this is working by watching my speeds running different and programs.
 
OP
Brando

Brando

Member
Joined
Jan 9, 2006
I was having the same issue with Prime95, which I now consider to be the "hard" test for stability. There is a setting in my BIOS that will detect when avx is present and it will automatically knock down my multiplier. It's like a multiplier offset. So I can run at 4300mhz and I set it to drop to 4200mhz when AVX is present (which is my best stable o.c. In Prime95). I verified that this is working by watching my speeds running different and programs.

that's a great feature i never heard of. wondering if my motherboard has it.