AMD Confirms Linux Performance Marginality Problem Affecting Some, Doesn't Affect Epyc / TR

[BillBroadley]

For those who haven't bought ryzen yet... is there a way to buy a known good ryzen?

[duby229]

For those who haven't bought ryzen yet... is there a way to buy a known good ryzen?

Well, the good ones are week 25 of this year and newer, but I don't know if you can tell about that before you buy?

[kaseki]

Well, the good ones are week 25 of this year and newer, but I don't know if you can tell about that before you buy?

Based on the community.amd.com segfault thread, I would offer instead an assertion that week 25 units are being supplied under RMA after being checked for functionality by AMD on a comparable configuration to the user's configuration. We don't know for sure that all samples from week 25 are free of the segfault problem. If AMD knows that, it could save itself a lot of RMA testing.

[timon37]

We don't know for sure that all samples from week 25 are free of the segfault problem. If AMD knows that, it could save itself a lot of RMA testing.

Actually we know that at least one from week 25 didn't work. It was supplied as first RMA replacement, before they started doing the whole "we'll test with your exact mb and ram before we ship".

[janweb]

Actually we know that at least one from week 25 didn't work. It was supplied as first RMA replacement, before they started doing the whole "we'll test with your exact mb and ram before we ship".

Is there a possiblity to know what chip number i have with a commandline in linux?
I know that i'm affected, yesterday i try to compile mesa and it stopped with segfault after 1minutes my temperature is okay.

The AMD Support don't write back after i sent them some photos TWO WEEKS ago....

[kaseki]

I have read somewhere that the date code is only on the thermal heat spreader lid and not written to the CPU.

[bridgman]

Is there a possiblity to know what chip number i have with a commandline in linux?
I know that i'm affected, yesterday i try to compile mesa and it stopped with segfault after 1minutes my temperature is okay.

The AMD Support don't write back after i sent them some photos TWO WEEKS ago....

If it's been two weeks please ping them again - response can be a few days when they are very busy but this sounds more like email got missed/misfiled.

[JC_Yang]

Do you guys know this repo

https://github.com/hayamdk/ryzen_segv_test

Is it the same bug as the heavy compilation one?

[JC_Yang]

Do you guys know this repo?

https://github.com/hayamdk/ryzen_segv_test

is it the same issue as this one?
I'm a developers and just another victim of this issue, 1700x user here.

[keantoken]

I was getting segfaults with the kill-ryzen chip within 3 minutes. But with Vsoc set to 1.05V and Vcore offset set to +0.05V, I no longer get them (I don't know whether upping Vsoc is important or not). These are the only changes I made to the most recent BIOS. Which is interesting considering everyone ignored the people saying to up your Vcore in the last few threads. If I had just tried that I would have had the problem solved. But no, people said that was stupid so I didn't try it (so who's stupid, really?).

At any rate, it still isn't really fair to expect a user to manually adjust their voltages in order to get a stable system, so I will push for an RMA. For me power dissipation is important, so I don't want to up the voltages if I don't need to. Furthermore, if I can lower power draw by undervolting, that might also help.