Processors

[gegunio]

Hi

My platform:

- MOTHERBOARD Crosshair VIII Extreme (AGESA 1.2.0.3 patch c)
- CPU Ryzen 5950X
- GPU Radeon 5700XT Nitro +
- PSU be quiet Dark Power Pro 12 1500W
- RAM G.Skill Tridentz Neo 32GB (2x16GB) 4000mhz CL14
- AIO Arctic Liquid Freezer II 420 push-pull
- SYSTEM Windows 10 (up to date)

I've been having a strange problem with my PC for a while now (previously this didn't happen after OC with my previous 5900X on this same system).

Well, sometimes after running Cinebench R23 or Y-Cruncher 2.5b my PC shuts down. It happens only with PBO enabled (it can be PBO + CO + boost, it can be even PBO alone without CO and boost with just power limits enabled).

I thought it could be RAM-related - I've tested my RAM in TM5 for many hours with 1usmus v3 preset - zero errors even after 25 cycles run. I've tested the whole system in the AIDA64 Extreme stress test. Also tried every 1-hour OCCT test. Every. PBO2 and Curve Optimizer are matched well and tested also in Core Cycler 720-720 - no single error. So it should be rock stable.

I don't know why this is happening. Sometimes just when I turn on C23 multithreaded or Y-Cruncher the PC shuts down after a few seconds. The event viewer doesn't show any WHEA error.

This happens even if PBO alone is enabled with changed power limits (270-160-200), with the curve optimizer disabled. Could it mean that I have a faulty/broken 5950X? Such a problem didn't happen with my previous 5900X on this same system.

Does 5950X with PBO not working and causing PC to crash qualify for RMA at the dealer?

[elstaci]

I would open a AMD Support -  Warranty ticket and ask them if your processor needs to be RMAed  to be checked or replaced from here: https://www.amd.com/en/support/contact-email-form

I am little bit perplexed that the processor past all the other Stress testing programs except Cinebench 23. 

Even though your RAM passed all physical tests for checking for defective RAM modules those test doesn't check for compatibility.

If you are running the RAM at 4000Mhz try running the RAM at its SPD Speed and see if it passes Cinebench 23 without crashing with PBO enabled.

I went to your motherboard's Support QVL LIst for the 5000 series CPU and there aren't any compatible RAM at 4000Mhz as 2x16 RAM Modules set: https://rog.asus.com/motherboards/rog-crosshair/rog-crosshair-viii-extreme-model/helpdesk_qvl_memory

The only set I found that uses a set of 2x16 RAM Modules was at 3200Mhz speed:

So it seems like your motherboard doesn't support speeds of 3333Mhz and above with 16GB Density RAM Modules, at least, for the Vendor G-SKILL RAM.

[gegunio]

Thanks for the reply and for the suggestion, I will get back to AMD and will open a warranty ticket.

As for my ram - it runs at 3800mhz because my CPU's IF doesn't allow more. This ram is the Tridentz Neo series, dedicated to working with Ryzen 5000 and tested at G.Skill HQ with these CPUs. It's also a fairly new kit on the market, perhaps that's the reason it's not on the CH8E QVL list.

[elstaci]

If it is a new model then that is why it isn't listed since it hasn't been tested yet by Asus for compatibility.

But I would still run your RAM at its SPD speed and see if the problem occurs. Since I didn't see any RAM modules of a set of 2x16 listed above 3200Mhz as being compatible for G-Skill RAM.

It may not be a processor issue but a Motherboard compatibility issue at higher RAM speeds with G-SKILL.

I also have G-SKILL Trident-Z Neo Ram of 2x16 set with my Ryzen 7 3700 processor at its native speed of 3600Mhz.

[gegunio]

I tried SPD RAM speed also with PBO (in every config) and the shutdown still happens.

[elstaci]

Thanks for the update. Then lets see if AMD Support believes your processor needs to be checked via RMA process.

I know that several Users here at AMD Forums in the past eventually had their 5000 series processors replaced to fix all the issues they were having.

Lets see if you are in the same boat as those other Users.

I was thinking that maybe Cinebench 23 used more RAM Memory than the other stress testing programs when the program was running its test.

When a computer shuts down it is mainly due to Overheating, Power issues with the PSU, Overclocking, Defective hardware, incompatibility, and faulty drivers. 

OCCT can show you the PSU outputs during it PSU Stress test and temperatures and fan speeds etc. If everything check good or normal than most likely it is a hardware defect or incompatibility.

It is strange that no errors pop up when the PC shutdown.

See what programs or files are having issues by running DXDIAG.exe and saving the file. The very last category at the bottom of all that data shows all the files that are having issues in your PC currently or in the past. That might give you an idea on where to concentrate your troubleshooting.

Of course the easiest way to check your Processor is by installing it on another compatible PC and see if the exact same thing occurs. If it does then I would open a AMD Warranty to RMA the processor.

AMD Warranty starts as soon as your Retailer/Vendors Return Date or Warranty expires.

[gegunio]

Update.

I ran many more Y-Cruncher 2.5b tests on stock CPU settings and started getting a "coefficient is too large" error, suggesting possibly unstable OC, maybe some AVX problem. Except that the stock CPU is ... without OC

error

[elstaci]

I believe that PBO is generally on "Auto" in BIOS so if PBO is automatically enabled during your test it is being overclocked.

PBO is AMD way of safely overclocking its processors.

Just for troubleshooting purposes can you disable 50% of your CPU Cores, you can use Ryzen Master or BIOS, and then run Cinebench 23 and Y-Cruncher again and see if it crashes or shuts down?

In the past I noticed that some programs are not compatible with processors with so many cores enabled and thus when you disable half the cores it runs stably.

NOTE: Here is Y-Cruncher FAQ which explains many of the errors you are getting: http://www.numberworld.org/y-cruncher/faq.html

[gegunio]

Hey, thanks for reply. Now I have PBO disabled in BIOS.