Stability?

demondrops

demondrops

Limp Gawd
Joined
Jul 7, 2016
Messages
422
I tested with intelburntest, aida, intel xtu test and also occp before. but prime seem to just kill my cpu, tempratures go off the charts really. and it looked like i had to drop 200 mhz on my oc to even run it over a period of time. but i got no issues with any other tests and temps go at like 85c in those tests for prime, and i see it go to 90c and i did the mistake of leaving to go store and i saw on hwmonitor it fucking hit like 100c probs for brief moment tho.. and ambient here is like 20c or so like what the hell, this is at 1.26v.. i even changed tim and reset fan just to make sure nothing was wrong, got the noctua dh15 and 4 intake fan and the noctua it exhaust rigth into a corsair ml 140 as exhaust out of cabinet, packet is like 40, but most cores run 31c idles. with that said i have no random crashes or anything odd at all happen like ever running as i did before and that is why i have been doing exactly that. and i have gamed heavily, demanding games and also not so demanding games. i hear some ppl say go with it and other ppl are like 24/7 prime for a day? what y guys say?
 
Last edited:
The new prime uses avx which is hot and needs lot of power

Some Bios have avx offset just for this reason

Use the old version of p95, 26.6 iirc as it doesn't have avx
 
Last edited:
  • Like
Reactions: Nokia
like this
avx? im bit noob i want to learn, but that i know is it is something to do about electricity going in to cpu. but is it ok for me to think it is 24/7 stable when i have 100% stability except from prime95? because for last 3-4 months ive run what is in my signature and i never had one problem at all. i did run it longer then that, but i had some issues that were caused by win10 settings, but as soon as i fixed it i never had any bsod or any failure whatsoever. 100% stable except from prim95. im sry i had some beer today lol som im talkative and extra weird pls forgive me haha. literally 0 problems at all, and ive gamed bf1 wich is very heavy on cpu and everything, and alot of other things. and i had several days i sit all day long when im awake. and 0 problem.
 
i just ran the latest version of prime 95 btw. ok but if it is loading the cpu more then the power i would allow it normally it wouldnt be realistic?
 
rigth now i sit at 1.26 and it is game stable, aida stable and intel burn test stable.. but prim 95 autocrash unless i lower to 4.2 then it cna go to 100c as i posted earlier :S
 
using a weaker stress test to hide cpu instability is not making you system more stable. You are just ignoring the issue.

If you can live with that. Yhen thats fine, its your freaking computer


Also if peope think the only difference between 26.6 and 28.5 (or newer) is only avx clearly does not understand how codepath and code optimizng works.
26.6 is weaker because its not meant nor optimized for any newer cpu's so it keeps you cpu stalled for more time.
You can run any prime version without AVX but people tend to not really care to understand things. once they have gained an opinion one it. funny because in my mind you should learn how things works before gaining the opinion.
In short: 28.5 without avx still runs hotter/better for testing than 26.6. There is no logical reason for running outdated code for a stress test besides trying to ignoring an issue.

But again: Select how much (in)stability you want. But unstable during anything = unstable.
 
ok running 26.6 it is just like other tests np xD and it dont look like i can boil water either.
 
SvenBent said:
using a weaker stress test to hide cpu instability is not making you system more stable. You are just ignoring the issue.

If you can live with that. Yhen thats fine, its your freaking computer


Also if peope think the only difference between 26.6 and 28.5 (or newer) is only avx clearly does not understand how codepath and code optimizng works.
26.6 is weaker because its not meant nor optimized for any newer cpu's so it keeps you cpu stalled for more time.
You can run any prime version without AVX but people tend to not really care to understand things. once they have gained an opinion one it. funny because in my mind you should learn how things works before gaining the opinion.
In short: 28.5 without avx still runs hotter/better for testing than 26.6. There is no logical reason for running outdated code for a stress test besides trying to ignoring an issue.

But again: Select how much (in)stability you want. But unstable during anything = unstable.
Click to expand...
Suffice to say, you can run the latest version and disable avx...no need to downgrade.
 
demondrops said:
ok running 26.6 it is just like other tests np xD and it dont look like i can boil water either.
Click to expand...

In case you get isseus with 26.6 you con continue downgrade you test again to ignore that instability. e.g. go down to just open 5 words and make that a pass.
You didn't fix you issue. you just decided to look away from it.

Again its your computer and your choice to do so. Just don't think you fixed the issue.
 
Speedeu4ia said:
The new prime uses avx which is hot and needs lot of power

Some Bios have avx offset just for this reason
Click to expand...

Historically, I've been a stickler for pure stability testing but to me this AVX thing is a different animal. The fact that mobo manufacturers offer an offset now tells me that AVX and the way its stressed are broken for any poor enthusiast who happens to have a motherboard without the offset. That's I've why ignored AVX and will until I have a mobo with the required offset and can overclock without worrying about it.

Even if AVX matters to your workflow is there a 1:1 correlation to what these stress tests do with AVX heat generation that matches real-world scenarios? E.G. if you almost melt your chip with AVX stress test and it fails, can you really not do video editing under a normal workload? Something tells me you can.

So yeah if you can run AVX stress tests crash free, I say congrats, but I also say without an offset, you are probably gimping your system for every other use case.
 
..
So if i don't put racing suspensions in my Volvo, it's not 'stable', right? Damn, nobody told me that man..
I gotta feel each and every bump in the road crushing my back because hey, it IS unstable, i mean who knows, next turn might take me straight into the Rockies! Would it survive them? It wouldn't, now would it?
See? Unstable! Useless those cars.
..

You are your CPU's user; no other. Adjust it to your needs and your needs alone. That's the whole point with PCs.
Do you have any program that utilises advanced vector extensions? Do you think it's possible you might? Ever?
If the answer is no, then stop hammering your CPU to death; literally. Just because the cool kids in the block do it, doesn't mean you should too. Most especially considering the process involved; do a thorough search over at overclock.net, see how many motherboard reps report degraded CPUs within mere hours of Prime testing :)
I never understood why we (humans) were made so; one little monkey does something 'x' way, the next one doesn't even think. It just goes and does exactly what it saw the first one doing.

Don't be that person.
Your only reason to use Prime is if you're lacking both in knowledge and inclination to ammend that flaw by informing yourself. Even then, you need be aware of just how stressing it is. Needlessly so.

As always, just my opinion.
 
I'm personally fond of x264 as a gauge. Sure, my system may fail a test on Prime95 that it passes on x264, but I don't buy that it needs to pass Prime95. In my limited experience, if x264 doesn't crash, nothing else I run on my computer will - and that, to me, is "stable."
 
I'm in the group "if it can't run everything and anything stable, it's not stable". I have to point out that even these unrealistic synthetic test aren't the end all be all stability test. I had rigs that run Prime95, OCCT, RealBench and IBT for days without problems, but still failed in real life workloads.
 
Meeho said:
I'm in the group "if it can't run everything and anything stable, it's not stable". I have to point out that even these unrealistic synthetic test aren't the end all be all stability test. I had rigs that run Prime95, OCCT, RealBench and IBT for days without problems, but still failed in real life workloads.
Click to expand...
I used to be in that group, but then I realized that "anything and everything" was an impossible goal, and changed it to "anything and everything that I would typically run."
 
  • Like
Reactions: Aenra
like this
VanGoghComplex said:
I used to be in that group, but then I realized that "anything and everything" was an impossible goal, and changed it to "anything and everything that I would typically run."
Click to expand...
I feel like I can't trust the CPU to do correct calculations even if it doesn't crash, if it can't run something it could while stock.
 
Meeho said:
I feel like I can't trust the CPU to do correct calculations even if it doesn't crash, if it can't run something it could while stock.
Click to expand...
So you'd give up on an overclock in the event that it couldn't reliably run a piece of software that you never need to run? I mean, to use a pretty obvious example, I don't rely on Prime95 for work or play. I'm not gonna abandon my 5.1GHz if Prime95 is the only piece of software that crashes it. If Prime95 and, say, Audacity, a program I actually use, both crash at 5.1, then I'd probably go back to 5.
 
VanGoghComplex said:
So you'd give up on an overclock in the event that it couldn't reliably run a piece of software that you never need to run? I mean, to use a pretty obvious example, I don't rely on Prime95 for work or play. I'm not gonna abandon my 5.1GHz if Prime95 is the only piece of software that crashes it. If Prime95 and, say, Audacity, a program I actually use, both crash at 5.1, then I'd probably go back to 5.
Click to expand...
Yes, I would, because I don't know what new program I'll run tomorrow or if I'll run some new specific workload in my existing programs that will trigger the instability. I care for "stock-like" stability, not "stability as long as I don't deviate from my usual work". If you can't pass all the tests, you're just avoiding triggering your crash prone CPU. The only exception would be thermal constraints. If I'm stable but the temps get too high with specific synthetic tests, I wouldn't necessarily downclock if those kind of temps won't be seen in real life use.
 
Meeho said:
Yes, I would, because I don't know what new program I'll run tomorrow or if I'll run some new specific workload in my existing programs that will trigger the instability. I care for "stock-like" stability, not "stability as long as I don't deviate from my usual work". If you can't pass all the tests, you're just avoiding triggering your crash prone CPU. The only exception would be thermal constraints. If I'm stable but the temps get too high with specific synthetic tests, I wouldn't necessarily downclock if those kind of temps won't be seen in real life use.
Click to expand...

I think van gogh has "stock-like" stability with his 5.1, he can come back though and let us know if he doesn't.
 
Meeho said:
Yes, I would, because I don't know what new program I'll run tomorrow or if I'll run some new specific workload in my existing programs that will trigger the instability. I care for "stock-like" stability, not "stability as long as I don't deviate from my usual work". If you can't pass all the tests, you're just avoiding triggering your crash prone CPU. The only exception would be thermal constraints. If I'm stable but the temps get too high with specific synthetic tests, I wouldn't necessarily downclock if those kind of temps won't be seen in real life use.
Click to expand...
Okay, so what would you do if I wrote a program that would crash your CPU at stock clocks? Let's say for the sake of argument that this program displays an image of a cupcake on your screen, and does nothing else - except crash your PC, of course - but since I've written it, it exists and thus your stock clocked system is now "unstable" by your definition.

I'm not trying to get you to change your practices, BTW - it's your CPU, of course. I just like this discussion. We throw around absolute language regarding stability, when actually verifying 100% stability in all possible cases is practically impossible.
 
VanGoghComplex said:
Okay, so what would you do if I wrote a program that would crash your CPU at stock clocks? Let's say for the sake of argument that this program displays an image of a cupcake on your screen, and does nothing else - except crash your PC, of course - but since I've written it, it exists and thus your stock clocked system is now "unstable" by your definition.

I'm not trying to get you to change your practices, BTW - it's your CPU, of course. I just like this discussion. We throw around absolute language regarding stability, when actually verifying 100% stability in all possible cases is practically impossible.
Click to expand...
No, that is not my definition. I'm ok with it behaving like a stock CPU. Whatever makes a stock CPU crash is ok to be crashing on an OCed one. By "stock-like" I mean that it has to be indistinguishable from a stock CPU, regardless if it's crashing or not crashing. If the cupcake made an OCed CPU crash, but runs fine on a stock one, no matter how useless it is, I wouldn't consider the rig stable enough for my liking.
 
haven't we had this prime vs not prime stable argument for a while now? prime95 pushes a system well past any normal 100% usage scenario. if everything but prime is stable for him, who cares?
 
I've used photogrametry software that would crash a 24/7 Prime95 stable rig.
 
Meeho said:
I've used photogrametry software that would crash a 24/7 Prime95 stable rig.
Click to expand...
ok so something specialized? is it common? can it be used instead of prime? you sure its not the program crashing itself?
 
pendragon1 said:
ok so something specialized? is it common? can it be used instead of prime? you sure its not the program crashing itself?
Click to expand...
It's not common, but it's real life workload. Today it's running x264 fine, tomorrow it could crash doing x265, or x266 the day after. The point is that there are actual uses that stress the system more than Prime, so it's not like Prime is doing something outrageously hard.

It's not the program, they were 0x124 BSODs. More voltage and/or lower OC fixed it.
 
You must log in or register to reply here.
Back
Top