Need Help with Prime95 Stability

I just built a new Computer and can't get Prime95 Blend to run over 8 hours. Always at around 8 hours I get a message Prime95 has stopped and needs to close.
Prime95 Blend fails after 8 hours at stock settings.
I even bumped the CPU V to 1.36 and it still fails at the same time. Max temp is about 60C on the hottest core.
IntelBurntestV2 runs fine even on Maximum setting for 10 loops, Max temp 63C on the hottest core.

I tested my Ram with Memtest86 and it ran fine for 10 hours at its rated DDR3200

I don't know what to do, I never had a computer fail Prime95 Blend on stock settings.

Everything seems to run fine, but I don't have confidence in it not being able to run Prime95 for extended period of time.

I tried overclocking to 4200 and Prime95 Blend fails at the same 8 hour time area.

Intel Processor Diagnostics Tool 3.0.0.25 tested and the CPU passed everything.



6850K CPU
ASRock Fatalty X99 Pro Motherboard With newest Bios 1.40
CORSAIR Vengeance LED 32GB (4 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) Desktop Memory Model CMU32GX4M4C3200C16R
Plextor M8Pe M.2 2280 1TB NVMe PCI-Express 3.0 x4 MLC Internal Solid State Drive (SSD) PX-1TM8PeG
EKWB EK-Kit 280 Water Cooling
Windows 10 Pro

Please enable round off checking!

I'll give that a try, thanks for the information! I'll let you know tomorrow how it goes.

H2mike said:

I just built a new Computer and can't get Prime95 Blend to run over 8 hours. Always at around 8 hours I get a message Prime95 has stopped and needs to close.
Prime95 Blend fails after 8 hours at stock settings.
I even bumped the CPU V to 1.36 and it still fails at the same time. Max temp is about 60C on the hottest core.
IntelBurntestV2 runs fine even on Maximum setting for 10 loops, Max temp 63C on the hottest core.

I tested my Ram with Memtest86 and it ran fine for 10 hours at its rated DDR3200

I don't know what to do, I never had a computer fail Prime95 Blend on stock settings.

Everything seems to run fine, but I don't have confidence in it not being able to run Prime95 for extended period of time.

I tried overclocking to 4200 and Prime95 Blend fails at the same 8 hour time area.

Intel Processor Diagnostics Tool 3.0.0.25 tested and the CPU passed everything.



6850K CPU
ASRock Fatalty X99 Pro Motherboard With newest Bios 1.40
CORSAIR Vengeance LED 32GB (4 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) Desktop Memory Model CMU32GX4M4C3200C16R
Plextor M8Pe M.2 2280 1TB NVMe PCI-Express 3.0 x4 MLC Internal Solid State Drive (SSD) PX-1TM8PeG
EKWB EK-Kit 280 Water Cooling
Windows 10 Pro

Click to expand...

Did you try running the memory at stock speed with prime95 Blend?

With Prime95 stress testing, it mostly tests the FPU (floating point unit) and a little bit of integer calculations so the could be a problem with the FPU in the intel X86 CPU, I had that problem once and a new CPU solved it.

Floating-point unit Link: https://en.wikipedia.org/wiki/Floating-point_unit

I ran Prime95 Blend at stock CPU and the ram at XMP 3200 which I thought was stock for that ram. When the computer first booted, the ram was set at 2133 or something like that and I thought that was wrong, so I set it to DDR3200 16-18-18-36 2T

When setting ram to XMP that is overclocking and sometimes the CPU memory controller on the X99 won't work well, or the memory could be the problem at it's rated speed. That is the reasoning behind Intel using JEDEC 2133 speed when it first boots, so folks can see how well it will overclock from stock specification.
DDR4 2400/2133 LINK: http://ark.intel.com/products/94188/Intel-Core-i7-6850K-Processor-15M-Cache-up-to-3_80-GHz

Intel XMP http://www.intel.com/content/www/us/en/gaming/extreme-memory-profile-xmp.html

Warning: Altering PC clock or memory frequency and/or voltage may (i) reduce system stability and use life of the system, memory and processor; (ii) cause the processor and other system components to fail; (iii) cause reductions in system performance; (iv) cause additional heat or other damage; and (v) affect system data integrity. Intel assumes no responsibility that the memory, included if used with altered clock frequencies and/or voltages, will be fit for any particular purpose. Check with memory manufacturer for warranty and additional details. For more information, visit http://www.intel.com/content/www/us/en/gaming/overclocking-intel-processors.html.

Click to expand...

Thanks Wingman, I didn't know setting the ram at its rated speed of 3200 was overclocking. I guess I should set the ram back at 2133 and then try Prime95 Blend again.
Memtest86 works with the memory set at XMP DDR3200 so I thought I was good to go.
If Prime95 Blend works at stock memory 2133 and it fails at XMP 3200, I'll get back with you to see where I should apply voltages to fix the problem. Thanks for the help!

Well heck that didn't work!
Now I am really stumped.
Prime95 Blend just failed again with the memory set to stock 2133 mhz. It failed at about the same place, 8 hours in.
The memory passed Memtest86 even at DDR3200
The CPU passes IntelProcess Diagnostics and IntelBurnTest on Maxium settings.
What now?

Now that you have tested on stock settings and it fails only on Prime95 you have a defective FPU in the CPU. That would mean a CPU RMA is needed.

Floating-point unit https://en.wikipedia.org/wiki/Floating-point_unit

If it was Prime95 unstable, I would expect it to happen randomly. So it failing after 8 hours each and every time is suspicious. I think there is some external influence, but what... are the tests started at similar times e.g. overnight?

I'd break the problem down into steps. First, run only small FFT. This doesn't really affect ram at all, so if this fails it is pointing more at the CPU itself. If that is fine, then run large FFT. This does hit the ram system.

wingman99 said:

Now that you have tested on stock settings and it fails only on Prime95 you have a defective FPU in the CPU. That would mean a CPU RMA is needed.

Click to expand...

Possibly Skylake-gate striking again!

RJARRRPCGP said:

Possibly Skylake-gate striking again!

Click to expand...

Which version of Prime95 were you using? Or were you applying specific settings? The Skylake bug requires downgrading to older instruction set, so if you use a current version without modification, you will not encounter the bug.

mackerel said:

If it was Prime95 unstable, I would expect it to happen randomly. So it failing after 8 hours each and every time is suspicious. I think there is some external influence, but what... are the tests started at similar times e.g. overnight?

I'd break the problem down into steps. First, run only small FFT. This doesn't really affect ram at all, so if this fails it is pointing more at the CPU itself. If that is fine, then run large FFT. This does hit the ram system.

Click to expand...

He is not overclocking so there is no random on transistor switching time. The transistors are switching within lintels specifications. When he runs Prime95 BLEND it runs through small FFT also in place large FFT and system memory automatically, Why do testing separately over again he is not overclocking.

Usually this will take about 21 hours to run through all iterations depending on how fast your core is.
Prime95 will now run a full test in the following order:

448k, 8k, 512k, 12k, 576k, 18k, 672k, 21k, 768k, 25k, 864k, 32k, 960k, 36k, 1120k, 48k, 1200k, 60k, 1344k, 72k, 1536k, 84k, 1728k, 100k, 1920k, 120k, 2240k, 140k, 2400k, 160k, 2688k, 192k, 2880k, 224k, 3200k, 256k, 3456k, 288k, 3840k, 336k, 400k, 480k, 10k, 560k, 16k, 640k, 20k, 720k, 24k, 800k, 28k, 896k, 35k, 1024k, 40k, 1152k, 50k, 1280k, 64k, 1440k, 80k, 1600k, 96k, 1792k, 112k, 2048k, 128k, 2304k, 144k, 2560k, 168k, 2800k, 200k, 3072k, 240k, 3360k, 280k, 3584k, 320k, 4000k, 384k, 4096k

The thought did cross my mind that maybe it was doing a certain thing at a certain time. If it is a certain FFT size that in particular fails where others doesn't, that would still be interesting in itself, as you could run just that FFT to quickly test.

My thinking of running the two types separately was to gain insight into the type of workload that caused the failure. Overclocking or not, I'm not making any assumptions when figuring out what is wrong.

Edit: thinking more, prime95 keeps a log of results so it may be possible to review what FFT size previous runs failed on.

The Version of Prime95 I'm using is V25.6 Build 6. The only thing I am doing now different is that I have a Check mark in (Round off Checking).
I have run this at all times of the day and night, It fails no matter what time of day.
I'm running again now and it should fail in about 4 hours, I'll look at the log and see if I can read it.
On the log, the last one that passed this morning was Self-test 1536K passed. Nothing after that. I guess when it fails it doesn't log that. I'll see in a few hours if it fail at the same place.

I missed that detail. The "normal" error when things are unstable is it will detect a computation error and stop. An app crashing error is different...

Then again 25.6 is really old and I'm not familiar with its operation. It would pre-date FMA which improves performance massively on CPUs Haswell onwards. Any reason a newer version isn't used?

I'll download and use the newer version. I just had the old one and didn't know their was a difference.
Yes the program crashes at 8 hours. I'm coming up on 7 hours now. I'll be ready with the new Prime95.

mackerel said:

The thought did cross my mind that maybe it was doing a certain thing at a certain time. If it is a certain FFT size that in particular fails where others doesn't, that would still be interesting in itself, as you could run just that FFT to quickly test.

My thinking of running the two types separately was to gain insight into the type of workload that caused the failure. Overclocking or not, I'm not making any assumptions when figuring out what is wrong.

Edit: thinking more, prime95 keeps a log of results so it may be possible to review what FFT size previous runs failed on.

Click to expand...

Yes Prime95 tests with different calculations on each iteration, then it does rounding and fatal error crosscheck and keeps moving on. On an absolutely stable system, Prime95 would run indefinitely.

It does not mater what calculations are run first. The small FFT calculations fit in the L1 L2 L3 cache not much ram needed. Small FFTs run fast because of the faster cache speed and smaller calculations. The large FFT calculations run L1 L2 L3 some memory. Blend switches between both calculation sets every 3 minutes with system memory 2GB.

It really does not matter what part of the CPU is failing with large number calculations or small number calculations. The failure could be any one CPU transistor out of billions in any Cores with L1, L2, also the L3 IMC memory controller could be the problem.
I have seen prime95 fail or the OS running it. It makes no difference where the CPU Error comes from and from what I explained how can a person tell where the CPU failure came from, with running software.

I have seen other people run into this problem inducing me with a Q9550. replacement fixed the error.

I just built a skylake 3 months ago and it would have infrequent freezing problem every 1-3 weeks and it passed all stress testing software. Then I RMA'ed the CPU no more problems.

- - - Updated - - -

H2mike said:

I'll download and use the newer version. I just had the old one and didn't know their was a difference.
Yes the program crashes at 8 hours. I'm coming up on 7 hours now. I'll be ready with the new Prime95.

Click to expand...

I hope it was a bug in the old Prime95 have not had time to look at the update DOC.

It just Failed again with everything on stock settings.
8 Hours 9 minutes it said (Prime95 Application has stopped working)
A problem cause the program to stop working correctly. Windows will close the program and notify you if a solution is available.

It stopped on the first test 2100 Lucas-Lehmer iterations of M34230145 using FFT length 1792K

Windows didn't notify me!
Should I run the new Prime95 V28.9 or just RMA my CPU?

I'm running the new Prime95 V28.9 Build 2. I'll run blend all night and if it fails also, I'll RMA the CPU tomorrow.