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

[zakhrov]

Can someone tell me why AMD makes such a fuss to send Michael CPU review samples? I mean you gave away 250 of those review kits to tech youtubers. Unless you were going to surprise him anyway :P

[mir3x]

Can someone tell me why AMD makes such a fuss to send Michael CPU review samples? I mean you gave away 250 of those review kits to tech youtubers. Unless you were going to surprise him anyway :P

Oh, it's why I see everywhere on YouTube how ryzen is awesome and intel sucks ...
I thought its's honest

[Brutalix]

Can someone tell me why AMD makes such a fuss to send Michael CPU review samples? I mean you gave away 250 of those review kits to tech youtubers. Unless you were going to surprise him anyway :P

I hope and assume it's to calm fears of the same errors in EPYX and threadripper. The "youtubers" might have more "followers", and is less likely to ask difficult questions that the marketing department cannot answer.

Edit: To early for me detecting cleverly hidden sarcasm.....


Well they at least answered my question if they manage to reproduce the problem, and they do. For me Ryzen works for the intended purpose heavy parallel computing, so I don't mind the wait, but for those running compiling feasts it's a let-down, to run with one or a few cores.


Kind regards.
B.

[leipero]

I'm still having really hard time to grasp how it could be hardware problem if it works well on another software (eg. Windows). However, as I'm sure AMD people know (much) more about their product than me, I will apologize for being wrong and saying it is a software bug, even tho I am still not convinced about it..., Michael should do the same on conftest segfaults once situation is resolved, and I think it was quite clear from comments on that article that is normal behaviour.

[sdack]

So it's not a bug with Ryzen per se as it's not happening with Windows but an issue particular to Linux. Who would have thought ...

[pete910]

How long does the kill-ryzen script take to throw an error? have run it up to an hour without any problems

Currently set it going again about 15mins ago or so to see when/if it does anything

Antergos(updated)
running a 1700x @ 3.9
3200 mem via XMP
Gigabyte k7 board

well to update, hour now. it does show this though every 5 mins.

[KERN] Aug 08 10:05:34 com1 kernel: mce: [Hardware Error]: Machine check events logged
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: Corrected error, no action required.
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: CPU:15 (17:1:1) MC3_STATUS[Over|CE|MiscV|-|-|-|-|SyndV|-]: 0xd820000000000150
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: IPID: 0x000300b000000000, Syndrome: 0x000000002a000503
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: Decode Unit Extended Error Code: 0
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: Decode Unit Error: uop cache tag parity error.
[KERN] Aug 08 10:05:34 com1 kernel: [Hardware Error]: cache level: RESV, tx: INSN, mem-tx: IRD

Is this related?

[bug77]

Remember good old days? http://www.pcworld.com/article/112891/article.html No shame in using commodity hardware if you know what you are doing.

Yes, but that's not commodity hardware. Those are fault tolerant clusters built on top on commodity hardware (i.e. not something a faulty CPU can bring down by itself).

[ernstp]

I definitively get this bug with overclocking, but do I also get it without overclocking... ? Not sure yet...

[GreekGeek]

Hi Yall,

excellent work Michael & RyzenNewbie/FreeBSD peeps.

This is the strength of the Unix community in action.

Greekgeek :-)

[Michael]

Can someone tell me why AMD makes such a fuss to send Michael CPU review samples? I mean you gave away 250 of those review kits to tech youtubers. Unless you were going to surprise him anyway :P

They don't see the Linux consumer market being big enough?