• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Help with 3900x and Prime95 Blend, LargeFFT

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

xielvas

New Member
Joined
Jun 16, 2020
I have tried so many thing to get this system to run prime95 blend/largeFFT that I now want to break the damn system!

I have Asus ROG 570-e, AMD Ryzen 3900x, Corsair CMW32GX4M2C3200C16 Dram Samsung B-die DDR4 (Thaiphoon)

The bios settings have been different. Right now they are set to auto (optimized defaults)

Every time I try to run the damn prime95 I get workers that stop, the program stops and crashes or the display goes wonky. This happened to me when I was using XMP profile 16-18-18-36-54 @1.36v, CPU at 4Ghz AND with optimized defaults. I am going to run OCCT when I go to bed to see what happens but it seems a moot point while it's running defaults settings.
I have also tried running prime95 with a diagnostic startup (msconfig) and Prime95 does the same thing. I have never had such a problem overclocking. Jesus I'd be happy with just running the Corsair memory with the XMP certified profile. To hell with a 200 mhz overclock on the CPU!

Please help this "old" overclocker with an AMD Ryzen.
 
Well, I never run large fft... i stick with small (cpu) and blend (cpu+mem).

That said, you mention mention bios settings have been different, but not sure what specifically.

Have you updated your board to the latest bios?
Is that memory kit on the board's QVL list?
Have you tried raising SOC voltage?
 
What cooler are you using and what is your max CPU temp and max CPU power?

Running Prime95 with Small FFT under PBO, with my 3900X cooled by a Cooler Master 240mm AIO, HWiNFO shows a CPU package power of 168W, with CPU temp maxing at 83C, and all cores at 3,900 MHz.
 
Yes, you don't mention how you are cooling the CPU or what your temps are like when stress testing.

That the RAM is Samsung B die is good but the fact that it is Corsair brand may not be good when paired with Ryzen. G. Skill is usually a better choice with Ryzen. As has already been asked, is your RAM on the QVL list? As EarthDog suggested, try giving the SOC a little voltage bump and maybe even the RAM itself a little voltage bump. You are running a 32gb kit so with that much RAM it may help.

I would also suggest doing a CMOS reset to make sure all settings are returned to bone stock. In some instances, just choosing optimized default may not do this.

What about your other components? Is the PSU stout enough to carry the load? More information about your system would be helpful.
 
If it works with memory at auto/SPD (probably 2133 or 2400) then try manual settings instead of XMP. Usually, motherboards handle it well when popular IC is in use. Simply set manual memory mode (it should be 1st option with something like XMP/XMPII/manual/auto options), set memory clock so 3200, main timings of 16-18-18-36 and memory voltage so 1.35V. Leave everything else at auto.
 
What cooler are you using and what is your max CPU temp and max CPU power?

Running Prime95 with Small FFT under PBO, with my 3900X cooled by a Cooler Master 240mm AIO, HWiNFO shows a CPU package power of 168W, with CPU temp maxing at 83C, and all cores at 3,900 MHz.

Corsair H115 Platinum , 140 watts, 81c

I did a cmos reset (shorted the two pins) previously. I tested last night for 8.5 hours with occt. I had a 4.1ghz cpu overclock, No Avx, 24 threads, Large. Memory set to DOCP/XMP 3200, FLCK 1600, SOC 1.15, CPU VDD 1.3, LLC Level 3, SOC LLC Level 1, CPU Power Phase Optimized and there was NO errors and the temp never hit over 70c. I have also run many loops of 3dmark CPU stress test with out errors and Fire Storm with 98% score. I don’t know what the hell is the matter with prime95.

Yes the memory is on qvl and ryzen optimized. I checked it last night as well. PSU is fine EVGA 850 and none of the voltages monitored drop below spec ever.
cpu: ryzen 3900x
motherboard: asus rog strix 570-e
cooler: Corsair H115 AIO
ram: 2 x 16 Corsair Vengeance RGP Pro
GPU: 2080 XC Ultra
Case: Corsair 680
PSU: EVGA Supernova 850 P2
Storage: Samsung 1TB NVMe SSD 970 Evo Plus, Samsung 1TB SSD 860 Evo, WD Gold 4TB Enterprise Class Internal HDD
OS: Windows 10 Pro 64 bit (Virtualization enabled in BIOS for Hyper-V & Windows Sandbox)

That enough information of the system?
 
Last edited by a moderator:
I don’t know what the hell is the matter with prime95.
Are you 'playing' P95? Are you stable in things you actually do like gaming, etc? Don't worry about passing multiple tests. Settle on one you know that works for your uses.

As I said, trying adding .1V or so of SOC voltage and see if it stabilizes... otherwise, use your system as you normally would and see if it is stable. If it is, who cares about P95. ;)
 
For starters, I wouldn't use blend to start with as that mixes wildly different FFT sizes and it makes it hard to tell what's going on. Small FFT will be a core stress. Large FFT is more of an IMC/ram test than core test. The original post makes it confusing what state the system is running in. Is it stable if the CPU is stock (no overclock, especially no PBO) and the ram is full auto? Then repeat with XMP? What FFT size was running when it fails?
 
Prime95 issues

For starters, I wouldn't use blend to start with as that mixes wildly different FFT sizes and it makes it hard to tell what's going on. Small FFT will be a core stress. Large FFT is more of an IMC/ram test than core test. The original post makes it confusing what state the system is running in. Is it stable if the CPU is stock (no overclock, especially no PBO) and the ram is full auto? Then repeat with XMP? What FFT size was running when it fails?

It did this stock AND overclocked. I don’t have any of this problem with other benchmarks; aida64, realbench and occt v6

Large FFT and Blend were/are the two that would cause glitches.
 
Friend... does it do this in normal activities on your PC? If you are stable there, don't worry about it...... :)
 
So have you run Prime95 with Small FFT?

Do you have any issues running actual games and/or applications?

No issues with doom eternal, far cry 4 or overwatch all set to Ultra settings 4k. Yes I can run small FFT

- - - Auto-Merged Double Post - - -

I agree. Time to disconnect the OCD clock in my brain and actually “trip out” when there IS a problem.
 
No issues with doom eternal, far cry 4 or overwatch all set to Ultra settings 4k. Yes I can run small FFT

Then I'd guess your issue is RAM or IMC related. But since your games run fine, it's likely an edge case and nothing to obsess over.
 
Then I'd guess your issue is RAM or IMC related. But since your games run fine, it's likely an edge case and nothing to obsess over.

Yea makes no sense. Ran memory tests at stock speeds with no errors. I have since tightened up the settings in BIOS, e.g. SOC, VDDCC etc. I ran asus realbench stress 2 hours without an issue, occt small fft with AVX2 - run hot 199 watts 87c without issue and 8.5 hours running large fft overnight no avx. 3dmark never has a hiccup either with fire stress test showing 98%.

I think you guys are right. I need to meditate, breath and just relax. I suppose I can get OCD with this benchmarking crap. I have always been that way though.
 
You're not alone. This crowd tends to be OCD by the nature of what we do.
 
Fairly sure we have another thread running that (at least) one of the comments speculates about Corsair memory in general not playing well with Ryzen ?
 
You are correct, Kenrou. It's a topic of discussion that's been around.
 
It depends on the memory series. In general, Corsair doesn't work well on Gigabyte boards. I had no problems with Vengeance Pro/RGB and Dominators RGB but users were reporting multiple issues with various Vengeance LPX kits.
Ryzen 3000 doesn't really need higher SOC or other voltages up to 4000+ (as long as the motherboard isn't undervolting). I could run a 2x32GB kit at 4133 stable with +0.1V SOC so it's like every "gaming" board will set at auto. The only problem can be with the XMP profile but manual settings should force a different timing table and that's why I suggested doing that. ASUS motherboards usually have no problems with memory kits but sometimes XMP isn't programmed right.

One tip. Run hwinfo64 and when Prime95 stops or crashes, check what is on the bottom of the table (literally the last one). There is info about detected hardware error.
You can also check if something like 7-Zip benchmark is passing let's say 30mins+ test (set at a higher dictionary to fill the whole RAM).
 
Back