News AMD Ryzen CPUs Getting New Microcode With Over 100 Improvements Next Month

artk2219

Distinguished
Jun 30, 2010
754
188
19,240
40
So... Ryzen 3k launched about three months before it was actually ready assuming 1004 will be the update that finally works for everyone rather than fix stuff for some and break stuff for others as previous updates have.
Pretty much, but it fits in with how just about everything has been released recently (unfortunately), launch early, patch often.
 

nitrium

Distinguished
So... Ryzen 3k launched about three months before it was actually ready assuming 1004 will be the update that finally works for everyone rather than fix stuff for some and break stuff for others as previous updates have.
Well the silicon was clearly ready, but the microcode was not. I guess the thinking was that the microcode is something they can fix post-launch (which is true, but not exactly ideal).
 

InvalidError

Titan
Moderator
Well the silicon was clearly ready, but the microcode was not.
I wouldn't be ready to say that, you need to remember why microcode exists in the first place. There are two parts to microcode: breaking down complex instructions into simpler steps for the RISC-like execution back-end and providing opportunities to correct or work around hardware flaws.

If the microcode isn't ready at launch, it could be taken as an indication that there were more hardware flaws, more severe flaw, other complications and unexpected misunderstandings of how the architecture worked than expected. The hardware may not have been quite ready but there wasn't time for a re-spin to address it at the source, so they take a gamble on field-fixing flaws with micro-code updates later if they are sufficiently confident about it and cross their fingers that they really do have access to all of the bits they need in micro-code to work around any known and additional issues they find.

Intel promised micro-code fixes for its Puma-6 modem chipset and several months of research later, it was forced to admit that the architecture simply did not have micro-code access to the bits it needed to change to actually fix it. With no software-based work-around forthcoming (at least not until 18 months later), it scored itself a class-action.

With hardware that requires day-0 micro-code updates for normal operation, sometimes you win, sometimes you lose.
 

InvalidError

Titan
Moderator
AGESA isn't microcode.
AGESA is code for the chipset, microcode is hard set inside the CPU.
Intel has had at least partially programmable micro-code in its CPUs since the Pentium. That's why some CPU bugs can be fixed or at least partially mitigated (Spectre, Meltdown and friends) by installing BIOS updates containing those micro-code patches.
 
Reactions: DMAN999

ASK THE COMMUNITY