AMD Preparing BIOS Update to Fix FMA3 Freezes on Ryzen CPU Family

[R-T-B]

AMD has acknowledged an issue in which applications utilizing FMA3 code (basically compute and floating point heavy applications) can freeze Ryzen-based desktops. According to AMD, a fix is already on the way in the form of a basic bios update that will be issued to motherboard vendors, who will then most assuredly update their boards with the fix. If you want to be sure your Ryzen based system is not affected by this or numerous other teething issues, making sure you are running the latest BIOS will go a long way towards easing your experience with your new platform.




AMD has not commented on the specifics of the bug itself, only saying they have "identified the root cause." Wherever the bug originates, it is likely a simple microcode patch that fixes the issue. Microcode has existed as a method to "patch" processor bugs since the infamous Pentium FDIV bug, when Intel decided it might just be smart to have a way to update the processor's running firmware with workarounds for newly discovered errata rather than just assuming it perfect on launch day. AMD and Intel both have microcode update methods now, and microcode can be updated via the bios. If this reported had to guess (and it might be worth noting that this reported has dabbled in bios-modding, so he's not clueless here), that would be my guess as to how they are addressing it.

View at TechPowerUp Main Site

 

[bug]

Great to hear it's not something needing a recall (like the old intel's fdiv bug).

 

[eidairaman1]

Newer steppings should have the issue resolved in the cpu itself

 

[Imsochobo]

Newer steppings should have the issue resolved in the cpu itself

Sometimes a stepping cannot solve issues.
Especially which resides in software.

even if it was hardware related:
+ chips have been sold, so one would have to replace the chips.
that is the hard way and have it Next year instead of now through software.
Steppings take time, usually 12 months for a zero defect capable chip.

 

[ssdpro]

Great to hear it's not something needing a recall (like the old intel's fdiv bug).

You mean the old Pentium one from the 90s? And I agree, a firmware patch is much better. The bummer part is all the sites will run benchmarks to check for performance changes.

 

[R-T-B]

You mean the old Pentium one from the 90s?

Yep, that's the one. It's actually the reason CPU vendors invented microcode updates.

 

[bug]

You mean the old Pentium one from the 90s? And I agree, a firmware patch is much better. The bummer part is all the sites will run benchmarks to check for performance changes.

Yep, that's the one. It's actually the reason CPU vendors invented microcode updates.

Yeah, I read the title and posted. And then I read the article and saw it was mentioned in there already...

 

[Mussels]

We are Pentium of Borg, precision is futile


How i first heard of that intel bug, maaaaany years ago.

 

[eidairaman1]

Sometimes a stepping cannot solve issues.
Especially which resides in software.

even if it was hardware related:
+ chips have been sold, so one would have to replace the chips.
that is the hard way and have it Next year instead of now through software.
Steppings take time, usually 12 months for a zero defect capable chip.

I totally understand all of that however after the initial batches etc definitely fix the. Microcodes are good as a bandage for initial batches.