News Zen 6 Ryzen spotted in AIDA64 — latest software beta adds 'preliminary support' for next-gen AMD desktop processors

[Admin]

On Monday, Finalwire added 'preliminary support for next-generation AMD desktop, server and mobile processors' to AIDA64 Extreme, indicating readiness for Ryzen 10000 series chips.

Zen 6 Ryzen spotted in AIDA64 — latest software beta adds 'preliminary support' for next-gen AMD desktop processors : Read more

 

[usertests]

I'd like to see 10-cores become the "low end".

 

[greenreaper]

64-cores ought to be enough for everyone... right?

 

[dalek1234]

I'd like to see 10-cores become the "low end".

I think it will be 8 cores.

6-core Zen-5 CPU's actually have 8 cores but 2 are disabled most likely because they are defective. That would mean they are disabling 25% of original cores.

Zen-6 will have a 12-core CCDs, so if you use the same math and disable 25% of those, you'll be disabling 3 cores. Can they disable 3 cores, or does it have to be an even number? If they have to disable 4 cores, then they will be left with 8 functional cores.

With that in mind, I think that the lowest core count CPU will be eight cores.

 

[usertests]

6-core Zen-5 CPU's actually have 8 cores but 2 are disabled most likely because they are defective. That would mean they are disabling 25% of original cores.

Yields are so high on these chiplets, that it's likely that "good" 8-core chiplets have become 6-core products. Or maybe only one core is bad, or the cores don't reach the highest clocks.

So we've seen 75% enabled (2 cores disabled) for 6-core products. With Zen 6, the 10-cores will be ~83% enabled, also with 2 cores disabled. An 8-core would be ~67% enabled, 4 cores disabled. I think it's plausible that AMD does not bother making the 8-core because N2X yields for ~75mm^2 chiplets won't justify disabling four of the cores. But they may go for it to provide something for the budget users.

AMD can disable an odd number of cores. Some Epyc products enable only 1 core per chiplet for high clocks and easy management of the L3 cache (the core gets to use the whole 32 MiB without ever sharing). But they aren't likely to make a 9-core consumer product.

In MLID's leak, he claims the I/O die could house two new "LP" cores, with hyperthreading. So we may see 8+2/20T anyway... but it will perform kind of like a 9-core since the LP cores will be nearly half as fast.

 

[dalek1234]

Yields are so high on these chiplets, ....

I missed the in I/O-die LP core variant in MILD's video. That would make for a great laptop chip for idle or low-level tasks.

 

[usertests]

I missed the in I/O-die LP core variant in MILD's video. That would make for a great laptop chip for idle or low-level tasks.

It would even be nice to get power consumption down during idling, video playback, etc. on desktop.

Maybe there could be scheduling issues, but hopefully it can leverage the same work done to support the 2 LP-E-cores in Meteor Lake-H/U and Arrow Lake-H/U. Intel should be bringing these to desktop eventually, and I'm surprised they haven't already.