News AMD's next-generation Zen 6 "Medusa Point" APUs could feature as many as 22 cores

"It's probable RDNA 4 won't make its way to the APU landscape, since AMD's next-generation graphics architecture, UDNA 1 / RDNA 5, is projected for release during the same timeframe."

If AMD APU's continue to use N-1 Radeon generations (or N-0.5), it would just mean that Medusa Point's successor generation would use RDNA4, or maybe RDNA 4.5 at best if that even becomes a thing. I'd like to see AMD close the gap, but we'll have to see.
 
  • Like
Reactions: bit_user
The 3nm server AMD (Turin) chips appeared already in 2024 but the laptop parts on 3nm planned only for 2027 ???

During this time Intel, which despite of all its problems with access to UV lithography surprisingly all these years still keeps dominating laptop market, will totally bury AMD in this segment
 
Last edited:
  • Like
Reactions: bit_user
It's a weird layout and in direct conflict with the MLID leak, at least for the top R9 version. Maybe they have lots of chiplets to work with and assembling a 20-22 core monstrosity is no big deal. Just toss in the smaller 4+4 chiplet alongside the desktop CCD if there is room.
 
If they are going for MCM, I'm guessing AMD finally figured out how to lower idle power consumption?

The 8CU iGPU would finally give a good distinction from Z series, although I'm guessing they'll sell it alongside a renamed 7840/8840/HX370.
 
This sounds almost like they're adding a regular Zen 6 CCD to a design similar to current APUs. If this is indeed the case then they're likely using TSMC's InFO for connecting everything like Strix Halo does.

Also makes me wonder if they're going to have two APU designs because there's no way they're leaving 8CU as the top end graphics configuration.
 
If they are going for MCM, I'm guessing AMD finally figured out how to lower idle power consumption?
Yes, silicon bridges between the dies, which is why they are touching each other now unlike the Zen 2-5 desktop designs. I guess there are several different names for this.

Also, they may be able to turn off a whole compute chiplet when it's not needed, that should save some power.

Also makes me wonder if they're going to have two APU designs because there's no way they're leaving 8CU as the top end graphics configuration.
There's going to be a lot of consternation about this 8 CUs. I won't say it's impossible for 8 CUs to outperform 16 CUs if it has Infinity Cache and RDNA 4/5 improvements. But it seems unlikely. A generational regression is unacceptable. We already have conflicting leaks for Medusa Point, so there's likely different configs being confused. Or AMD is going to pair smaller graphics with CPU-oriented models that are intended to be paired with discrete Nvidia graphics.