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

[mlau]

Did you also ask them about mce bug (random freeze / reboots)? Those are actually more annoying than segfaults.

That is what interests me as well: Sometimes tasks just get stuck, and those are far more annoying because you can't just restart the build after a segfault..

[sdack]

Hi Yall,

excellent work Michael & RyzenNewbie/FreeBSD peeps.

This is the strength of the Unix community in action.

Greekgeek :-)

*lol* No. This was a tantrum thrown by young AMD fanboys. The problem doesn't even exist any more according to AMD, at least with newer revisions of the CPU nor with Windows at all. This was an oddity at best, triggered only in a rare case on an OS which doesn't even cover 3% of the PC market and a CPU maker, whose market share is also only about 20% (Intel still holds about 80%). This has little to do with the almost 50-year old UNIX community - that's how old UNIX is.

[bug77]

*lol* No. This was a tantrum thrown by young AMD fanboys. The problem doesn't even exist any more according to AMD, at least with newer revisions of the CPU nor with Windows at all. This was an oddity at best, triggered only in a rare case on an OS which doesn't even cover 3% of the PC market and a CPU maker, whose market share is also only about 20% (Intel still holds about 80%). This has little to do with the almost 50-year old UNIX community - that's how old UNIX is.

And again, the root cause hasn't been found, you don't know if it has been fixed or that it is confined to Linux. It's easily reproducible in Linux, but that doesn't mean it's all it affects. I hope it doesn't affect anything else, but until it's diagnosed properly, you can't just make assumptions about what it affects.

I'm a software developer and I can tell you the bugs that you don't fix, but stop manifesting themselves without explanation are the scariest kind. Because you're never 100% sure they won't bite you again in the future.

[GreekGeek]

Hi Yall & "sdack,"

word of advice, you do realize this issue has been acknowledged by AMD? Ergo, hardly fanboy noise in a vacuum. As to the quality of your posts, on this issue .... Needs to focus, could try harder. ;-)

GreekGeek

[GreatEmerald]

With the exception of most Gentoo users, though admittedly a small percentage.

They pretty much count as developers From the CPU usage perspective, anyway.

[sdack]

And again, the root cause hasn't been found, you don't know if it has been fixed ...

Oh please, spare me the nonsense. You're 17 years too late. The last time anyone cared for computer FUD was when experts proclaimed a world-wide crash of computers by a 99 turning into a 00.

The cause doesn't matter. As a software developer do you not need to know how a CPU works internally. You only need to write your software in conformance to its specification. When you then don't have the full specs should you also not expect to find kindness and forgiveness for your stupidity, but be ready to find problems. Obviously has the Linux kernel not been adjusted properly to the new CPU, but it was assumed to behave like a previous CPU. This doesn't make it a CPU bug. It makes it a Linux bug. All CPUs have oddities, quirks and bugs, and quite many. They all need to be respected and taken into account by OSes. Clearly this hasn't happened yet with Linux as Linux is always a bit behind when it comes to new hardware.

This is/was nothing more than a tantrum thrown by idiots.

[duby229]

Oh please, spare me the nonsense. You're 17 years too late. The last time anyone cared for computer FUD was when experts proclaimed a world-wide crash of computers by a 99 turning into a 00.

The cause doesn't matter. As a software developer do you not need to know how a CPU works internally. You only need to write your software in conformance to its specification. When you then don't have the full specs should you also not expect to find kindness and forgiveness for your stupidity, but be ready to find problems. Obviously has the Linux kernel not been adjusted properly to the new CPU, but it was assumed to behave like a previous CPU. This doesn't make it a CPU bug. It makes it a Linux bug. All CPUs have oddities, quirks and bugs, and quite many. They all need to be respected and taken into account by OSes. Clearly this hasn't happened yet with Linux as Linux is always a bit behind when it comes to new hardware.

This is/was nothing more than a tantrum thrown by idiots.

Dude seriously, that is so wrong. At this point it seems like your just making excuses now. It's plainly obvious your just trying to make a pass. The next time Intel releases a horribly bugged product you want to be able to say, but look AMD does it too.... And that thought process just aint right.

[sdack]

... a horribly bugged product ...

*lol* That's your interpretation of this issue? That Ryzen is a horribly bugged product? Damn, you're pathetic.

[duby229]

*lol* That's your interpretation of this issue? That Ryzen is a horribly bugged product? Damn, you're pathetic.

Obviously. AMD acknowledged that yesterday. It's part of history now. Whats done is done. The only thing pathetic is you weak ass excuses. I can't wait for Intel to release their next horribly bugged product. Then we can listen to all the Intel fanboys like you bitch about how it's OK and perfectly normal.

[bug77]

Oh please, spare me the nonsense. You're 17 years too late. The last time anyone cared for computer FUD was when experts proclaimed a world-wide crash of computers by a 99 turning into a 00.

The cause doesn't matter. As a software developer do you not need to know how a CPU works internally. You only need to write your software in conformance to its specification. When you then don't have the full specs should you also not expect to find kindness and forgiveness for your stupidity, but be ready to find problems. Obviously has the Linux kernel not been adjusted properly to the new CPU, but it was assumed to behave like a previous CPU. This doesn't make it a CPU bug. It makes it a Linux bug. All CPUs have oddities, quirks and bugs, and quite many. They all need to be respected and taken into account by OSes. Clearly this hasn't happened yet with Linux as Linux is always a bit behind when it comes to new hardware.

This is/was nothing more than a tantrum thrown by idiots.

W-T-F?

Edit: Oh, I see now. You're blaming this on gcc. Again, you can't do that just yet, you still need the root cause to sort this out.