3950x + X370 K4, prime95 errors

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

   Dear ASRock experts,

last month I got a Ryzen 3950x for my workstation (which I use for software development + scientific computing), as an upgrade for my aging Ryzen 1700. The motherboard is an ASRock Fatal1ty X370 Gaming K4 paired to 32GB of Corsair RAM. The CPU is cooled by a Noctua NH-D15S, I am using linux, and everything is running at stock speed/default configuration on the latest beta BIOS (5.70).

I was very happy with the purchase, the CPU was chewing flawlessly through hours of multithreaded simulations, parallel compilations, etc. without a hitch. A couple of weeks ago however I was reading about prime95 and CPU stress testing, and thought to give it a try to see how the CPU would behave. And (I guess you can see where this is going), lo and behold, almost immediately I started getting errors.

The failing test is the small FFT one, and it fails pretty fast (usually a handful of seconds) complaining about wrong accuracy. Thermals don't seem to be an issue (the CPU stays around 66C even under the most stressful tests), and after a bit of googling it seems like my specific error message might be related to Vdroop issues.

So I started to do a bit of investigations and poking around, and, to keep the story short, the two salient data points/conclusions are:

- if I disable automatic voltage/frequency scaling in the BIOS and fix Vcore and frequency to the default values (1.1V and 3.5GHz), the prime95 test does NOT fail (it went on for about 10 hours without issues);

- with automatic voltage/frequency scaling ON (i.e., the default, failing configuration), voltage drops to 0.98 and frequency goes almost to 3.7Ghz. This frequency is in sharp contrast with what has been reported by other users, who see a frequency of 3.5Ghz in this specific test (which is very AVX2-heavy, hence the downclock to the base speed).

So my feeling at this point is that, for whatever reason, the CPU is clocking higher than it should in this specific test, leading to errors due to the motherboard being unable to supply the necessary current/voltage.

With all that said, what do you think I should do? Do you think this is a motherboard/BIOS issue, or do you think the CPU is at fault here? At this point I would be inclined to wait a bit more for another BIOS update from ASRock before thinking about RMAing the CPU, but I am a bit at a loss on how to proceed from here.

Thanks for reading!

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
bluescarni Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 27 Jan 2020 Status: Offline Points: 4	Post Options Post Reply Quote bluescarni Report Post Thanks(0) Quote Reply Topic: 3950x + X370 K4, prime95 errors Posted: 27 Jan 2020 at 6:02pm
	Dear ASRock experts, last month I got a Ryzen 3950x for my workstation (which I use for software development + scientific computing), as an upgrade for my aging Ryzen 1700. The motherboard is an ASRock Fatal1ty X370 Gaming K4 paired to 32GB of Corsair RAM. The CPU is cooled by a Noctua NH-D15S, I am using linux, and everything is running at stock speed/default configuration on the latest beta BIOS (5.70). I was very happy with the purchase, the CPU was chewing flawlessly through hours of multithreaded simulations, parallel compilations, etc. without a hitch. A couple of weeks ago however I was reading about prime95 and CPU stress testing, and thought to give it a try to see how the CPU would behave. And (I guess you can see where this is going), lo and behold, almost immediately I started getting errors. The failing test is the small FFT one, and it fails pretty fast (usually a handful of seconds) complaining about wrong accuracy. Thermals don't seem to be an issue (the CPU stays around 66C even under the most stressful tests), and after a bit of googling it seems like my specific error message might be related to Vdroop issues. So I started to do a bit of investigations and poking around, and, to keep the story short, the two salient data points/conclusions are: - if I disable automatic voltage/frequency scaling in the BIOS and fix Vcore and frequency to the default values (1.1V and 3.5GHz), the prime95 test does NOT fail (it went on for about 10 hours without issues); - with automatic voltage/frequency scaling ON (i.e., the default, failing configuration), voltage drops to 0.98 and frequency goes almost to 3.7Ghz. This frequency is in sharp contrast with what has been reported by other users, who see a frequency of 3.5Ghz in this specific test (which is very AVX2-heavy, hence the downclock to the base speed). So my feeling at this point is that, for whatever reason, the CPU is clocking higher than it should in this specific test, leading to errors due to the motherboard being unable to supply the necessary current/voltage. With all that said, what do you think I should do? Do you think this is a motherboard/BIOS issue, or do you think the CPU is at fault here? At this point I would be inclined to wait a bit more for another BIOS update from ASRock before thinking about RMAing the CPU, but I am a bit at a loss on how to proceed from here. Thanks for reading!

Ray62 Members Profile Send Private Message Find Members Posts Add to Buddy List Groupie Joined: 01 Jul 2018 Location: Germany, Berlin Status: Offline Points: 706	Post Options Post Reply Quote Ray62 Report Post Thanks(0) Quote Reply Posted: 28 Jan 2020 at 6:32am
	Try it with better cooling of the VRMs (FAN). https://docs.google.com/spreadsheets/d/1d9_E3h8bLp-TXr-0zTJFqqVxdCR9daIVNyMatydkpFA/htmlview?sle=true#gid=639584818
	Asrock Fatal1ty X370 Professional Gaming Offline Under test: MSI X570_MEG_ACE \| AMD Ryzen 9 3900X \| Scythe Mugen5 \| 2x16GB F4-3200C14D-32GTZ@3466 \| ASUS GeForce GTX 1080 Ti FE

bluescarni Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 27 Jan 2020 Status: Offline Points: 4	Post Options Post Reply Quote bluescarni Report Post Thanks(0) Quote Reply Posted: 29 Jan 2020 at 4:35am
	Thanks a lot for the link! I am not sure whether I have VRM temp sensors on my motherboard, but there is at least one temperature that gets pretty hot during the prime95 testing (it's called AUXTIN1 and it got up to 92C in the "stable" configuration). Any suggestion on fans, fan dimensions and ways of attaching them to the VRM protective plate? Thanks a lot again, it was really helpful.