• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Skylake CPUs can freeze running Prime95

Dave2150 said:
Funny how no Skylake user complained so far about crashes when doing encryption, video rendering, or any other task apart from this one guy who works with prime 95.
Click to expand...

Mines working absolutely fine with these kind of tasks also but Prime is(was) a total no go... Interesting that its been acknowledged as a problem though, despite it not actually effecting the vast majority performing the normal tasks. I can actually see a BIOS release from Asus for my Z170 deluxe dated today 'Stability Fixes' so might see if that resolves it, purely out of curiosity :)
 
Last edited:
Dave2150 said:
Funny how no Skylake user complained so far about crashes when doing encryption, video rendering, or any other task apart from this one guy who works with prime 95.
Click to expand...

It seems relatively obscure and benchmarking will considerably increase the chance of running into bugs like that (often a benchmark will effectively simulate days and days of use inside an hour) - as mentioned sometimes it would freeze the system almost immediately other times take hours - likewise with encryption where prime factorisation is a fundamental part of the process (hence why I mentioned it and not video rendering where its less commonly part of the algorithms used) something might set it off on the first instruction but as it seems to be highly specific chances are you could run a system for a long time before encountering it at all in every day use. That doesn't mean it isn't potentially a serious problem.

This isn't about an issue that affects a processor's stability under heavy load - though heavy loading will make it more likely you will run into it - it could happen at any time.
 
Last edited:
8 Pack said:
Even companies who have the uses pjbelf mentions. They are seeing no issues. Some are my customers. This is why I say Prime no need. Your actual uses are whats needed for testing purposes.
Click to expand...

Frankly I agree with this. This is a non issue for the largest majority of users.

Intel have rolled out two additional sets of microcode already for this regardless. So most vendors should have this resolved shortly, some already.

There are plenty of other ways to test stability on SKL.


Overlag said:
If its not prime stable, its not a stable overclock... But i guess i can post pictures of cpuz showing 5ghz and say woohoo i got 5ghz when even the smallest amount of load would crash it.

Just because prime95 will cause huge temperatures doesn't mean it useless today... If anything thats perfect to see if you able to cool your beasty overclocks. Whats the point of having a huge overclock if its unstable at 100% load?
Click to expand...


Prime's AVX2 routines at this moment in time still don't reflect any real world workload. It's up to the individual user if they feel the drop in multipler to maintain stability simply to run these routines is worth it. Just as I'll leave it up to you what you deem to be sensible.
 
Last edited:
Smffy said:
There are plenty of stress testing programs which run my 6700k at 100% with no problems for hours at great temps, Prime95 crashes in minutes, or at least it did before this 'update' i've not checked since (don't care). The temps are not really an issue as they sit mostly under 80c but it still blue screens. I'm sorry but nobody here is stupid enough to simply look at the speed listed in CPU-Z and think its stable, thats just silly talk we are all aware of putting the CPU under more than 2% load.

Prime is not really a comparative benchmark of a 'stable' overclock, the fact my overclock 4.6ghz has been running perfectly fine with heavy gaming, encoding, editing for weeks tells me its stable... So really what do I care if a program for which I have no use crashes?

I don't think there is any argument here that Prime is THE platform for stress testing, more people rightly pointing out that it has other functions.
Click to expand...


BSOD "but im still stable" :o
 
Rroff said:
It seems relatively obscure and benchmarking will considerably increase the chance of running into bugs like that (often a benchmark will effectively simulate days and days of use inside an hour) - as mentioned sometimes it would freeze the system almost immediately other times take hours - likewise with encryption where prime factorisation is a fundamental part of the process (hence why I mentioned it and not video rendering where its less commonly part of the algorithms used) something might set it off on the first instruction but as it seems to be highly specific chances are you could run a system for a long time before encountering it at all in every day use. That doesn't mean it isn't potentially a serious problem.

This isn't about an issue that affects a processor's stability under heavy load - though heavy loading will make it more likely you will run into it - it could happen at any time.
Click to expand...

We'd know by now if Skylake had a problem with anything outside Prime95.

People have owned Skylake CPU's for almost 6 months, and they've sold millions of them worldwide.

Just as an example, I stress tested my 6700k @ 4.8Ghz (my 24/7 overclock) using a variety of tools including Asus Realbench for over 24 hours.

My current uptime is 28 days...

Skylake is completely stable, only Prime95 users were affected by this bug.

Did you even read the article? :

Prime95 uses Fast Fourier Transforms to multiply extremely large numbers, and a particular exponent size (14,942,209) leads to the crash after “minutes or hours”.
Click to expand...

So, how many of you are using applications which use such functions?....

Besides, Intel have already released the microcode updates that fix this issue. You can actually modify your BIOS if you want the fix right now, or wait for the official BIOS releases from the motherboard manufacturers.
 
Last edited:
Smffy said:
Prime is not really a comparative benchmark of a 'stable' overclock, the fact my overclock 4.6ghz has been running perfectly fine with heavy gaming, encoding, editing for weeks tells me its stable... So really what do I care if a program for which I have no use crashes?
Click to expand...

Maybe because none of those uses have any error checking.

Prime95 checks that the CPU arrives at the same answer over and over again, many thousands/millions of times over. It's not unheard of for it to fail after many hours, which is an incredibly small error rate (considering the chip is doing millions of operations per second). Nevertheless, that's an error you've detected, not one that's potentially corrupted some system file somewhere.

Maybe most errors during gaming or encoding aren't a problem, but you just don't know. A crash/BSOD means something has erred in a catastrophic way - hardly a sensitive measure of stability.
 
8 Pack said:
Even companies who have the uses pjbelf mentions. They are seeing no issues. Some are my customers. This is why I say Prime no need. Your actual uses are whats needed for testing purposes.
Click to expand...

While you're on 8pack, what do you recommend for stability tests? Im stress testing right now, and I've noticed prime needs a fair bit more voltage than ROG realbench. Then again, it is a 6700k, so it could be the BUG!
 
Dave2150 said:
We'd know by now if Skylake had a problem with anything outside Prime95.

People have owned Skylake CPU's for almost 6 months, and they've sold millions of them worldwide.

Just as an example, I stress tested my 6700k @ 4.8Ghz (my 24/7 overclock) using a variety of tools including Asus Realbench for over 24 hours.

My current uptime is 28 days...

Skylake is completely stable, only Prime95 users were affected by this bug.

Did you even read the article? :



So, how many of you are using applications which use such functions?....

Besides, Intel have already released the microcode updates that fix this issue. You can actually modify your BIOS if you want the fix right now, or wait for the official BIOS releases from the motherboard manufacturers.
Click to expand...

You are mis-understanding the nature of the problem - such is the nature of it you could potentially use your system every day for years without crashing due to it or you could be suffering from crashes on a regular daily basis depending on your usage without ever going anywhere near Prime95 - Prime95 just makes it incredibly more likely you'd run into the problem within any given period of time. (It also has no relevance to how busy the CPU is - it could crash due to this problem when at very low utilisation).

There is nothing unique to Prime95 about Fast Fourier Transforms (they are commonly used in translating, rotating and scaling data in a lot of programs) but the data that in combination causes problems is relatively unique but again not unique to Prime95.

bfar said:
While you're on 8pack, what do you recommend for stability tests? Im stress testing right now, and I've noticed prime needs a fair bit more voltage than ROG realbench. Then again, it is a 6700k, so it could be the BUG!
Click to expand...

Give it a couple of hours of something like Intel XTU and then if it passes that drop the voltage til it starts not passing that (or if you want to just do it quick and dirty bump the voltage up one notch when it passes) then start using stuff you'd normally use, maybe throw in some CPU heavy games or video encoding, maybe one of the video encoding based stress tests.

EDIT: Running Prime95 for say 24 hours with core voltage over 1.35v with Skylake would do little more than put months and months worth of wear (electromigration type effects) on it for little reason.
 
Last edited:
My understanding is the bug is related to new 256bit FMA (Fuse Multiply Add) AVX2 instructions. These are very specific instructions added by intel to speed up certain types of calulcations and they were new in haswell. To even get the instructions in your code you either have to hand crank some assembler or use the very latest c++ compilers with specific flags.

Most general purpose code is unlikely to have them. Video processing, cryptogrophy and alike would be the most be the most likely candidates where you could see them on the desktop, but there newness means the reality is that there probably isn't much code using them.

This bug sounds like it will cause minimal issues on the desktop and xeon e3 platforms where we currently have skylake chips. It would probably be more of an issue if it made its way into the Xeon E5 and E7 platforms where you are much more likely to see the kind of workload and specifically optimized code that would trigger the problem. Enterprise customers can be very unforgiving.
 
Most compilers have support for FMA3/4 for upto what 5 years or so?, but yeah its more likely to be encountered in enterprise environments with more aggressive optimization against target hardware rather than programs created with general purpose compatibility in mind.
 
If your overclock is stable on everyday use gaming, watching videos, emailing what ever the hell ya do day to day your overclock is stable...
Best way to test if it's stable is use it as normal.

Hell if it's not stable in prime95 but its stable in everything you use your pc for, who the hell cares. Just leave it be.
 
r7slayer said:
If your overclock is stable on everyday use gaming, watching videos, emailing what ever the hell ya do day to day your overclock is stable...
Best way to test if it's stable is use it as normal.

Hell if it's not stable in prime95 but its stable in everything you use your pc for, who the hell cares. Just leave it be.
Click to expand...

slowly corrupting your data without even knowning it;)
 
r7slayer said:
If your overclock is stable on everyday use gaming, watching videos, emailing what ever the hell ya do day to day your overclock is stable...
Best way to test if it's stable is use it as normal.

Hell if it's not stable in prime95 but its stable in everything you use your pc for, who the hell cares. Just leave it be.
Click to expand...

You'll never achieve any kind of decent overclock like that. You need to know it's limits, and you won't find them like that.
 
Yea, but between skylake bugs and an untealistic weighting towards avx2 instructions, maybe prime 95 isn't the best tool to do that?

That said I can't find anything else that puts a CPU through the wringer quite like p95. On my CPU it needs a fair bit more voltage than realbench, xtu and ibt. The temperatures, however, aren't as bad as I was led to believe.
 
You must log in or register to reply here.
Back
Top Bottom