prime95 crashed how?

banker

Respectable
May 21, 2016
309
0
1,810
i'm overclocking on the M5A97-R2.0 board with an fx-8320 @4019Mhz so far, I was running prime 95 to make sure it was stable but after 13minutes it stopped working, the computer didn't crash though, does that mean i'm unstable or not? sorry kinda of a noob when it comes to overclocking
 
Solution
Personally, I agree with Fly, I think you should reset the overclock as your overclock is effectively nothing. An 8320 runs at 4GHz turbo out of the box. You're gaining .019GHz. I would either shoot for something higher or reset to default.

You're also likely going to have some issues overclocking the 8320 on that board. It's not super great for overclocking.

If you reset your BIOS to default, does the same thing happen? Have you tried uninstalling Prime altogether and then reinstalling it?

Any stress test is going to cause the core voltage and frequency of the processor to increase. Any of them. So are you suggesting then, by said logic, that you should overclock and not test it to make sure it's safe?

AIDA will do this...
Hi my friend.
Prime 95 is such an extreme stress test so I don't recommend it.
It could easily destroy everything.

Actually I have a problem with this. Why on earth is people thinking they need to run a stress test at all?
This must be a myth,

If You must?
Instead, use Intel Linpack and while this is also a power tool , it will not damage Your computer.
Especially when stress-test the memories
https://software.intel.com/en-us/articles/intel-linpack-benchmark-download-license-agreement

Use LinX a derivat from Intel stress test, it will also work on AMD
http://www.xtremesystems.org/forums/showthread.php?201670-LinX-A-simple-Linpack-interface
Best regards from Sweden
 


Yeah this is some strange thing, what exactly do You mean.
Yes of corse it will.
Prime95 will cause such a stress to the professor and therefore rise its temperature above all limits.
The reason is that it will increase the professor core-voltage and frequency.

Do You admit that?
Best Regards from Sweden
 


I got the windows something has stopped working tab pop up
 


Hi my friend.
I suggest that you back down and restart > get in to BIOS Setup > do a Optimized Default or Default settings > Save & Exit
'Best regards from Sweden

 


why are you against overclocking?
 
Not stable means not stable. If the errors are so numerous that Prime95 actually crashes, that means really not stable.

I would have no problem at all running Prime95 24/7 365 days a year because that's what it's for. After all, Prime95 is primarily a distributed computing method to find new prime numbers, and stress testing is only a secondary use.

I have never heard of Linpack reaching lower temperatures than Prime95. It puts the processor under so much stress that it's actually used by Intel to bin processors. That is something I would definitely not use 24/7 as it would cause electromigration degradation if used excessively. I do like it because it's the quickest way to check if an overclock is stable and if your thermal solution is up to snuff.
 
Personally, I agree with Fly, I think you should reset the overclock as your overclock is effectively nothing. An 8320 runs at 4GHz turbo out of the box. You're gaining .019GHz. I would either shoot for something higher or reset to default.

You're also likely going to have some issues overclocking the 8320 on that board. It's not super great for overclocking.

If you reset your BIOS to default, does the same thing happen? Have you tried uninstalling Prime altogether and then reinstalling it?

Any stress test is going to cause the core voltage and frequency of the processor to increase. Any of them. So are you suggesting then, by said logic, that you should overclock and not test it to make sure it's safe?

AIDA will do this. Prime 95 will do this. IBT will do this. Are you suggesting that no testing under stressed conditions should be run to ensure that the overclock is stable?

The core voltage and frequency will increase to run the processor at it's speed whenever it needs it to. Stress tests will increase thermals to a level that one will likely never see in actual use, but as long as the overclocking is done in a safe manner you aren't going to damage squat.

Are you telling me I shouldn't play games because games like say, Fallout 4 are CPU intensive and will make my voltage, frequency, and thermals rise? If I were a content creator, I shouldn't render video because that would make my voltage, frequency, and thermals rise?

Stress tests don't only give you an idea of whether or not your CPU is stable at the voltage you're running, they also give you an idea of the maximum thermal stress you are going to place on your processor. If the maximum is too high during these tests, you need to tune it back.
 
Solution


I would have no problem at all running Prime95 24/7 365 days a year
Say what? That is really not get to me as a serious statement.
Since I'm also is working in the electronic manufactory industry I do have have some basic knowledge about what happens to solder joints and intermalic layer-buildup and so on.
Intel Linpack btw (use LinX) is a very good tool for stress-testing the memory's and moderate professor checking stress.
Run it approx 20 minutes of max ram available and do not touch the mice nor the k-board during the test.
If You have never heard of it before, now this is the time.

But You can enlighten me to the reason one would have to use Prime95 at all?
Ok, if one has to do an extreme build, but otherwise for common people even if many have some moderate OC, I can not see any reason at all.
Personally I have been building and overclocked stabile computers since mid-90s.
Best regards from Sweden 😉
 
Prime95 is not intended for stress testing. It's for discovering new prime numbers 24/7 365 days a year during idle time of >1million computers worldwide, making it the 5th largest distributed computing project. Because many prime numbers are known, these known answers can be compared with what your computer computes. If your computer computes a wrong answer, it is by definition unstable and they don't want you contributing to the search because it is lying. Is that serious enough for you? Perhaps you should tell those thousands of Universities they are destroying all of their machines because you know better than them?

Linpack loads every execution unit on the CPU simultaneously, something not even the most optimized software can do (because it cannot get the RAM bandwidth to do so*). As a result, things get far hotter than any real software like Prime95. It's an artificial test that's useful as it ensures you have more than enough cooling because heatsinks eventually get dusty, and is better for testing CPU stability than the RAM. A convenient package that automates running Linpack is IntelBurnTest, and it's not unusual to see 10-20C hotter than Prime95.

* The exception is in supercomputers which are designed with far more bandwidth per core than any PC, and need better cooling too
 
To be fair, some of the earliest Prime95 to support AVX did locally overheat processors, but those days are long gone.

Linpack using the Intel binaries is nearly all floating point so does not test such features at all even though it is the most severe thermal test. So the moral is to test using as many softwares as possible unless you are sure you will never use those features. After all if accelerated transcoding becomes unstable at an extreme overclock, why would you care if you don't use Quick Sync?
 
Intel Linpack is essentially a benchmark tool for calculating (T)flops and so on, which is also a tool that really stress test the memories and its memory controller and of course the professor itself.

In my different testings scenarios I have discovered that Prime95 have a huge CPU load not comparable to any real user s/w use and Linpack has less professor core load, stll it will maximizes each core to its full extent which I also looked over via Core Temp Gadget on my second screen. (yes this is possible due to 8-windowsgadget s/w)

When I'm looking at the temps from everything in real time, I can clearly see that Prime95 really puts more stress to the professor.
Second, I have an instrument attached for measure the actual power draw at real time, this also shows more W(att) from Prime95 than Linpack.

Best regards from Sweden
 
That is really odd, as both of them mostly hammer FPU yet I have never seen Prime95 hotter than IntelBurnTest. Most people use the largest "Blend" setting to more test RAM but the smaller FFT may be more like Linpack.

The only things I can think of is the AVX, which is more efficient but does more work, or the Skylake bug. Would you be willing to try old version 26.6 from 2011 which doesn't support AVX?

Prime95 is currently running on 1.45 million CPUs, most of them OEM machines with stock heatsinks, and if they were all hotter than IBT I'd think we would've heard something about it by now.
 


Yes why not? I'm willing to do whatever necessary to get an objective and repeatable equal measurement as possible.

But I can't do this today 3:32 clock PM here now and I have other things to do, It must be during the weekdays Tuesday or Wednesday or next weekend?
Perhaps we shall start a new thread about this topic?
Just send me a PM.
Best Regards from Sweden
 

Hi,
This sounds that the limit is not the AMD FX-8320 more probably to the mobo M5A97-R2?
Make sure that AMD FX CPU does not reach temperatures more than 70°C then the professor will down clock itself and start throttling.

I can't say for sure in this case.
Best regards from Sweden