Overclock.net banner

Prime95 Instant Shutdown

7K views 26 replies 10 participants last post by  Undervolter 
#1 ·
Hey all, I just finished my new build and I'm running into some stability problems. I threw it together and it's been working great for a week or so, and I decided to download Prime95 just to check everything was okay.

As soon as I start a small FFT or blend test, the computer resets with no BSOD. I don't believe it's my power supply so I'm trying to look elsewhere before narrowing it to that...

i7 4790K
Gigabyte Z97MX-Gaming 5
4x8GB DDR3 2400mhz
Visiontek 7970
Seasonic X650

I've tried fooling around with a few settings in the bios on this board (while always returning to stock after) to no avail. I tried boosting the CPU voltage thinking it might be too low, and I also tried boosting the memory voltage since I thought that might be the issue. I have no desire to overclock this rig, but I really just want it running stably from here on out. It's been a long time since I've built a decent rig, so I'd appreciate any help and direction at this point.

Thanks guys!
 
See less See more
#3 ·
Quote:
Originally Posted by The Prodigy View Post

I actually just put together a 4970k build myself. For some reason P95 gives HUGE temperature readings (95+). Only happens with P95 though, not any other stress programs like AIDA or OCCT.

Might be that it's getting too hot and resetting.
It's not fast enough to generate temperatures. It's within half a second of clicking the button to start testing.
 
#5 ·
#7 ·
So I checked the temps and sure enough, Prime 95 caused an instant spike to 95+C and then the shutdown. I then researched the issue and my H100i backplate was a little loose, which is a known issue. I corrected the problem and reseated the cooler after a nice cleaning. All is well right? No wiggle with the backplate, it mounted perfectly.

I ran IBT for about 6 seconds and the temps are okay, slowly going up to 50, 60... They settled at 65 or so. Then my rig still reset. Prime is still immediate... This time with no change in temps.

Any ideas from here guys? I'm unfamilliar with the UEFI on my mobo, so if you could point me into the right direction if it's a setting there I would appreciate it.
 
#8 ·
Looks like its a MB and CPU issue. there are a lot of other people who are running into the same situation. it seems like if the CPU runs above 30% usage the temps will go above 80C and 90C in some cases.

This would explain why prime instantly shuts down cause the system is shutting off to protect against thermal issues.

"Thanks for the update boolybooly. An update for me, I updated my BIOS to F5 for my mobo and found cpu usage has decreased a little bit resulting in temps lowered by maybe 10-15 degrees celcius. However, should cpu usage increase beyond say 35%, temps increase past 80 going into 90s degrees celsius. I see that we are the guinea pigs on this new processor. Very poor testing standards by Intel and Gigabyte."

you can read about some of this from this link:

http://www.google.com/url?sa=t&rct=j&q=gigabyte%20z97mx-gaming%205%20shutting%20down%20issues&source=web&cd=9&cad=rja&uact=8&ved=0CFwQFjAI&url=https%3A%2F%2Fcommunities.intel.com%2Fthread%2F54032&ei=XfoYVKbOL8PkoATk2YLYDA&usg=AFQjCNGF5PDRCOKPqP2UMxU5VJ62i9YQSg

seems like the only longterm solution is to delid the cpu and add enough TIM so that the heat transfers properly.

either way good luck with you situation.

edit: here is another link of people in your situation some tried delidding.
http://www.google.com/url?sa=t&rct=j&q=ngptim%20issues&source=web&cd=2&cad=rja&uact=8&ved=0CCoQFjAB&url=http%3A%2F%2Flinustechtips.com%2Fmain%2Ftopic%2F179597-intel-%25E2%2580%2598devil%25E2%2580%2599s-canyon%25E2%2580%2599-chips%25E2%2580%2599-ngptim-is-still-not-efficient%2Fpage-2&ei=MP4YVIuUCI7koAS-qYDABA&usg=AFQjCNGCXsnrziL2khkEoI26yP0dFz9a5w&bvm=bv.75558745,d.cGU
 
#9 ·
Quote:
Originally Posted by wooshna View Post

Looks like its a MB and CPU issue. there are a lot of other people who are running into the same situation. it seems like if the CPU runs above 30% usage the temps will go above 80C and 90C in some cases.

This would explain why prime instantly shuts down cause the system is shutting off to protect against thermal issues.

"Thanks for the update boolybooly. An update for me, I updated my BIOS to F5 for my mobo and found cpu usage has decreased a little bit resulting in temps lowered by maybe 10-15 degrees celcius. However, should cpu usage increase beyond say 35%, temps increase past 80 going into 90s degrees celsius. I see that we are the guinea pigs on this new processor. Very poor testing standards by Intel and Gigabyte."

you can read about some of this from this link:

http://www.google.com/url?sa=t&rct=j&q=gigabyte%20z97mx-gaming%205%20shutting%20down%20issues&source=web&cd=9&cad=rja&uact=8&ved=0CFwQFjAI&url=https%3A%2F%2Fcommunities.intel.com%2Fthread%2F54032&ei=XfoYVKbOL8PkoATk2YLYDA&usg=AFQjCNGF5PDRCOKPqP2UMxU5VJ62i9YQSg

seems like the only longterm solution is to delid the cpu and add enough TIM so that the heat transfers properly.

either way good luck with you situation.

edit: here is another link of people in your situation some tried delidding.
http://www.google.com/url?sa=t&rct=j&q=ngptim%20issues&source=web&cd=2&cad=rja&uact=8&ved=0CCoQFjAB&url=http%3A%2F%2Flinustechtips.com%2Fmain%2Ftopic%2F179597-intel-%25E2%2580%2598devil%25E2%2580%2599s-canyon%25E2%2580%2599-chips%25E2%2580%2599-ngptim-is-still-not-efficient%2Fpage-2&ei=MP4YVIuUCI7koAS-qYDABA&usg=AFQjCNGCXsnrziL2khkEoI26yP0dFz9a5w&bvm=bv.75558745,d.cGU
Thanks for the reference links. I managed to tweak my setup a little and I can pass 5 tests of IBT with all 32GB of my RAM now. Prime95 is still an instant reboot but I really don't care. I undervolted a little and I'm in good shape, maxing out at 85C.
 
#10 ·
What version of IBT are you using? Have you updated to the latest Intel Math Kernel Library?

Prime95 instantly rebooting is indicative of a major issue somewhere. Even if you don't see any problems in anything else, yet, it would still be a good idea to figure out exactly what's going on so you don't run into any unpleasant surprises down the line.

First thing I would do is redownload and reextract Prime95, incase the program has been corrupted somehow. Then I'd see if the system can run it at stock without issue.
 
#11 ·
#12 ·
Quote:
Originally Posted by abctoz View Post

this is good advice, many find x264 harder to pass than prime95, and it runs COOL.
LOL! x264 is cooler because it stresses a LOT less the CPU! Of course, if you run Prime for just 30 minutes, it may be easier to "pass".
biggrin.gif
 
#14 ·
If it's hotter, it's because the CPU is doing more work, thus drawing more current, and producing more heat. Doesn't always mean it will fail sooner than another test, especially if the other test utilizes the CPU in a different way, but by most definitions, it does mean it's more stressful.
 
#15 ·
Quote:
Originally Posted by abctoz View Post

not true, my cpu requires +0.03v for stability in x264 as compared to prime95, others report the same.
Ok, so you input more volts, but you get lower temperature. I see... Yes, that makes all sense now.
rolleyes.gif


P.S.: x264 is an integer operation. Using that as the only means of "stress testing", in comparison to Prime95, is ridiculous. As for its power consumption, after having encoded 2000+ files and having a watt meter, i think i 'd know something.

Whatever floats your boat...
 
#16 ·
Quote:
Originally Posted by Undervolter View Post

Ok, so you input more volts, but you get lower temperature. I see... Yes, that makes all sense now.
rolleyes.gif


P.S.: x264 is an integer operation. Using that as the only means of "stress testing", in comparison to Prime95, is ridiculous. As for its power consumption, after having encoded 2000+ files and having a watt meter, i think i 'd know something.

Whatever floats your boat...
so encoding 2000 files with a watt meter somehow gives you knowledge of overclocking in haswell.. ok. what i was saying is that h264 is a better test of stability in haswell, and not a test of how much heat can be generated, i don't think you properly understood what i was trying to say.
 
#17 ·
Quote:
Originally Posted by abctoz View Post

so encoding 2000 files with a watt meter somehow gives you knowledge of overclocking in haswell.. ok. what i was saying is that h264 is a better test of stability in haswell, and not a test of how much heat can be generated, i don't think you properly understood what i was trying to say.
Yes, something that stresses only the integer part of the CPU and only moderately the RAM, is better than Prime95 and not only that, but also, you put more volts and get cooler temperatures! It goes against logic and the rules of nature, but if it works for you...

Doesn't work for others:



grey= Prime95
green = x264

Quote:
There's a huge power cost to ramping the chip up this high. Some of the jump is due to heavy RAM usage in Prime95 - and that's why we break things out into Peak vs. Load - but it's worth noting that our Noctua DH15 has no trouble keeping the 5960X below 60 degrees Celsius at stock when encoding video but strains to keep it below 80C at 4.2GHz.

In Prime95, even at stock frequency, the chip tops out at 100C. At 4.2GHz, the core hits 105C and destabilizes after about five minutes. Prime95 is an extremely difficult workload - it puts enormous stress on the CPU - but even allowing for other system components it's clear that increasing frequency drastically increases the CPU's power draw.
http://www.extremetech.com/computing/188911-intel-haswell-e-review-the-best-consumer-performance-chip-you-can-buy-with-some-caveats/2

Oh, by the way, please explain, how do your stress test the FPU in a process that uses only integer, as x264? Mysteries of the universe...

Like i said, whatever floats your boat. I 've been around fora long enough and have basic knowledge of human psychology, to not waste my time further. Great, x264 is better stress test than Prime95 and cooler too with more volts. You should make a club about that.

ADDITION:
Quote:
for those of you that don't know, the lead developer of x264 goes by the online name dark shikari. he recently started his own forum, named doom10, after he was banned by the mod from doom9 for disagreeing with him. in an ironic twist, i was banned from doom10 for challenging some of the idiotic claims made by his darkness.

anyway, a poster over at doom10 asked if x264 will support the new avx instruction set, here is what he said:

http://freeproxyserver.net/index.php?q=aHR0cDovL2Rvb20xMC5vcmcvaW5kZXgucGhwP3RvcGljPTUxNC4w

What about 256bit AVX? Each module can process only one at time (AFAIK) and x264 will surely support these new instructions (I hope eheh)

Float-only, thus a useless pile of tripe.

http://forum.videohelp.com/threads/327955-x264-will-not-be-using-the-new-256-bit-avx-instruction-set
x264 is integer operation. Which is also why FX CPUs close the gap better compared to Intels in x264 benchmarks (they don't suffer the shared FPU penalty from the module architecture). When you use x264 as stress test, you only stress the integer part of the CPU. The FPU (widely used in games for instance), is not.
 
#18 ·
Quote:
Originally Posted by Undervolter View Post

Oh, by the way, please explain, how do your stress test the FPU in a process that uses only integer, as x264? Mysteries of the universe...

Like i said, whatever floats your boat. I 've been around fora long enough and have basic knowledge of human psychology, to not waste my time further. Great, x264 is better stress test than Prime95 and cooler too with more volts. You should make a club about that.
as i was quoting earlier the link
http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics
actually uses x264 as the recommended test for good reasons
Quote:
ADDITION:
x264 is integer operation. Which is also why FX CPUs close the gap better compared to Intels in x264 benchmarks (they don't suffer the shared FPU penalty from the module architecture). When you use x264 as stress test, you only stress the integer part of the CPU. The FPU (widely used in games for instance), is not.
[/quote]
you seem to enjoy theory crafting, I'm no expert on x264 nor prime95, but it seems to me that prime95 is also heavily integer based:
http://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test
i'm going on a limb here but this shouldn't even matter apart from speed as performing operations on integer or floating point numbers follow the same rules of arithmetic in binary

and even if what i said in the last paragraph is completely false, there is still the results of many people who overclock haswell and experience that x264 encoding finds BSOD faster
 
#19 ·
I've had my Phenom II pass 20+ ITB under full RAM usage. Then when I fired up prime95 (blend), instant BSOD or reset. Increasing the CPU voltage and LLC and Current Capability fixed it along with lowering RAM OC settings.
 
#20 ·
Quote:
Originally Posted by abctoz View Post

as i was quoting earlier the link
http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics
actually uses x264 as the recommended test for good reasons
A recommendation made by John Doe, made to avoid high temperatures, doesn't mean that x264 is more stressful. Also, if you remember from school, P = V xI. If you raise V, P will rise. If P rises, temperature will. Also, something that stresses MORE the CPU, will also produce more heat, because more consumption means more heat. The graph in that link should hint that.
Quote:
you seem to enjoy theory crafting, I'm no expert on x264 nor prime95, but it seems to me that prime95 is also heavily integer based:
http://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test
i'm going on a limb here but this shouldn't even matter apart from speed as performing operations on integer or floating point numbers follow the same rules of arithmetic in binary

and even if what i said in the last paragraph is completely false, there is still the results of many people who overclock haswell and experience that x264 encoding finds BSOD faster
The principle of conservation of energy, is not theory crafting, is the base of physics. Good effort on wikipedia, but wrong article:
Quote:
Over the years, Prime95 has become extremely popular among PC enthusiasts and overclockers as a stability testing utility. It includes a "Torture Test" mode designed specifically for testing PC subsystems for errors in order to help ensure the correct operation of Prime95 on that system. This is important because each iteration of the Lucas-Lehmer depends on the previous one; if one iteration is incorrect, so will be the entire primality test.
The stress-test feature in Prime95 can be configured to better test various components of the computer by changing the fast fourier transform (FFT) size. Three pre-set configurations are available: Small FFTs and In-place FFTs, and Blend. Small and In-place modes primarily test the FPU and the caches of the CPU, whereas the Blend mode tests everything, including the memory.

http://en.wikipedia.org/wiki/Prime95
Back to the lowly principle of conservation of energy, Prime95 consumes MORE, because it does MORE. Prime95 is based on arithmetics, but for years, the dev develops it as stress test. x264 was NEVER intended to be a stress test. It is a video encoding library, nothing more. That some people in Intel forum, to avoid the extreme heat of Prime95 decided to baptize an integer ONLY operation as the recommended stress test for Haswells, is another story and IN DEED, this is THEORY CRAFTING, based only on their subjective opinion and those who did it, didn't even know that it was integer only. Let alone that x264 was never programmed to do any weird stress on IMC or RAM...

As for Prime passing when x264 fails, it only means you didn't let Prime run long enough. Nowdays, the "30 min, 1h, 2h" horde, is very big. Click my signature to see why old overclockers were priming for 24h.

Regards

P.S: Another small detail, x264 was never heavy on tbe cache, especially L3 cache. You can google and see that L3 cache on Phenom II made no difference compared to Athlon of the same clock, because it wasn't using it. I guess doesn't matter to test the cache either.
 
#21 ·
Quote:
Originally Posted by Undervolter View Post

A recommendation made by John Doe, made to avoid high temperatures, doesn't mean that x264 is more stressful.
I think you need to read up on haswell and avx, as per the article YOU linked me, the author got 105c on STOCK with prime95
Quote:
Originally Posted by Undervolter View Post

Good effort on wikipedia, but wrong article
i linked that article to show you that the prime95 algorithm is integer based..
Quote:
Originally Posted by Undervolter View Post

As for Prime passing when x264 fails, it only means you didn't let Prime run long enough. Nowdays, the "30 min, 1h, 2h" horde, is very big. Click my signature to see why old overclockers were priming for 24h.
a stress test that fails faster is more useful in my books
 
#22 ·
Quote:
Originally Posted by abctoz View Post

I think you need to read up on haswell and avx, as per the article YOU linked me, the author got 105c on STOCK with prime95
i linked that article to show you that the prime95 algorithm is integer based..
a stress test that fails faster is more useful in my books
Allow me to illustrate, again, how all this started:

http://www.overclock.net/t/1513462/prime95-instant-shutdown/10#post_22862431

And i linked an article, on how Prime95 implements that.

Why don't you go here and ask the dev himself? It will be an englithening experience:

http://www.mersenneforum.org/forumdisplay.php?s=93422047fe6639c7a04cdd0a4984c976&f=11

You could ask "Dear sir, is it true, that your Prime95 will stress my FPU, cache, IMC, RAM? Is it true you do that on purpose?"

Then you can go here and ask the x264 developer, if he 'd ever developed his library as stress test (Dark Shikari is his name). Or ask "Dear Mr. Shikari, is it true than your library isn't as stressful as Prime95?". He has a bit of short fuse and often doesn't answer to certain questions, but someone else there will.

http://doom10.org/

In my books, x264 is a stress test only if one wants to believe it is, which is another story. They picked it, just because it almost loads 100% the cores at all times (note the "almost). They could have taken Folding at Home, BOINC, Seti etc. Still, none of those were ever developed to be stress tests.

But, that has little relevance, as i said, i thought it was good to remember how this started and it certainly wasn't about which is quicker.

P.S.: Yes, that author got 105C. Because, physics for him works as usual. Prime stresses more, so heats more. x264 is so far behind, because does so much less. Guess which of the 2 was the most stressful. One thing is what is more stressful, another which one is cooler or "faster". I can say Battlefield 4 is a good stress test. It loads all cores, after all, doesn't it? Is it a real stress test? No, but if i believe it is, then it is (crafting theory). x264 was "chosen", exactly to AVOID heat. Not on any programming merits that were geared towards CPU stresstesting.
 
#23 ·
Quote:
Originally Posted by abctoz View Post

I think you need to read up on haswell and avx, as per the article YOU linked me, the author got 105c on STOCK with prime95
i linked that article to show you that the prime95 algorithm is integer based..
a stress test that fails faster is more useful in my books
Think about it like this:

You have a cpu, it does stuff.

Integer calculations, floating point operations, has integrated cache, integrated memory controller.

Now you run a 264x encoder, you do a lot of stuff with the interger calculations then, while the other parts are maybe 5% utilized.

Now if you were to utilize all of the parts at once at 100%, it would produce a lot of heat. That's what prime does.

Nothing more, nothing less.

Games tend to use a lot more floating point operations than 5%, more memory access than that, more cache usage.

Surely you could run tests for all the parts your cpu has individually, but I don't know many tests that stress specifically the parts that x264 does not stress, without stressing everything of the cpu at once. But testing only x264 on its own might make you see game/OS crashes when you encounter a game/install a new OS, that heavily uses something that you did not stress test, of the cpu.

P.S. intel burn test might as well be more useful for quick tests, but make sure you get a version that supports AVX2 as well c: It will also run as hot as Prime95, though. Or you could use an older version of prime/ibt that only has regular AVX support, to avoid the additional stress from AVX2, since an up to date x264 encoder (e.g. not the one in handbrake) would be using AVX2 already, so you can test that with that.
 
#24 ·
Quote:
Originally Posted by Undervolter View Post

You could ask "Dear sir, is it true, that your Prime95 will stress my FPU, cache, IMC, RAM? Is it true you do that on purpose?"

Then you can go here and ask the x264 developer, if he 'd ever developed his library as stress test (Dark Shikari is his name). Or ask "Dear Mr. Shikari, is it true than your library isn't as stressful as Prime95?". He has a bit of short fuse and often doesn't answer to certain questions, but someone else there will.
stahp
lachen.gif

i see it does use floating point operations:
http://www.mersenne.org/various/math.php
interesting, and no the developers didn't do it on purpose, the project is aimed to find mersenne primes.
Quote:
Originally Posted by Undervolter View Post

But, that has little relevance, as i said, i thought it was good to remember how this started and it certainly wasn't about which is quicker.
the goal of the stress test is to find instability, the one that finds it quicker is better.
Quote:
Originally Posted by Undervolter View Post

P.S.: Yes, that author got 105C. Because, physics for him works as usual. Prime stresses more, so heats more. x264 is so far behind, because does so much less.
... so how is a stress test that returns 105C at stock useful?
 
#25 ·
Quote:
Originally Posted by Tivan View Post

Now if you were to utilize all of the parts at once at 100%, it would produce a lot of heat. That's what prime does.
yes i know it runs hot, about 8-10c higher than h264. but anecdotally with my tests on the g3258 i could run prime blend for hours only to have x264 fail in the first pass, so its more useful as a stress test for dialing in vrin/vcore, i used prime95 for weeks but after discovering h264 i was amazed how much faster it was at finding bsod in haswell.
 
#26 ·
Quote:
Originally Posted by abctoz View Post

stahp
lachen.gif

i see it does use floating point operations:
http://www.mersenne.org/various/math.php
interesting, and no the developers didn't do it on purpose, the project is aimed to find mersenne primes.
The developers, my dear friend, are very well aware of the potential of their program as stress tests, for many years now. It may have not started as stress test, but the development is aware of that use and updates regularly for new CPUs. They even bothered to make a special wiki for it:

http://www.mersennewiki.org/index.php/GIMPS_overclocking_and_Prime95#Why_is_Prime95_good_for_testing_an_overclocked_machine.3F

Hint: in the interface, instead of simply saying "Find Prime", they have put 3 different "torture" tests (emphasis on torture), each of which, stresses more something different. To accomodate overclockers.

Quote:
the goal of the stress test is to find instability, the one that finds it quicker is better.
... so how is a stress test that returns 105C at stock useful?
But x264 does find instability at the end? All it finds, is that it's x264 "stable", whatever that means. One finds useful a stress test, that only stress integer, but can't find use for something that stresses everything... The bottom line, is that you don't want to see if you are stable. You want to see if you are x264 stable. That's all. To me, being "x264 stable" has no meaning, because i know x264 means nothing about stability. I 've been encoding for years and if i was to judge stability by x264, i 'd still be troubleshooting my Windows problems. If it was me, i 'd simply OC to the point i can run a real stress test, within "specs". Nobody put the gun at the head of anyone to OC to the point that temp under Prime will reach 105C... It's his choice.

Stability is one thing. As in "i can pass anything i throw at it". I wonder how many that are "x264 stable", would really pass real stress test. Oops, let's not go there, better stay with the "cool test", that allows us higher overclock. I will be stable, if i believe i am. x264 is a placebo stress test.

Finally, the question should be turned around. What kind of CPU, goes "bananas" when you put a 100% AVX load on it? Is it the stress test at fault, or is it a poor CPU design? What will happen if you encounter in the future a game, say Battlefield 5, that loads 100% CPU and uses extensively AVX2? Complain about Battlefield that brings the CPU to 100C?
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top