AMD Navi 31, 32 And 33 RDNA 3 APUs And GPUs Break Cover In Linux Patch Notes
There are a great many advantages to working with free and open-source software (FOSS
), and your author here is an outspoken supporter of such source-sharing ways. We probably don't need to explain the advantages, but suffice to say that software developed as a community effort can end up more secure and much more interoperable compared to proprietary software.
That's not to say that FOSS doesn't have its downsides, especially if you're a commercial entity like a corporation. We, again, won't go into that topic here, but rather focus on one specific downside—the minute you contribute your changes upstream, the whole world knows about everything in them. That's why companies like AMD tend not to add driver support for their upcoming hardware until pretty late in the game.
A chart listing the new GPU device IDs. Source: Coelacanth's Dream
As you've no doubt guessed by this point (based on the headline, if nothing else), AMD's pushed a patch for the Linux kernel that reveals a little bit more information about its incoming RDNA 3
graphics architecture. That includes confirmation of four device IDs as well as some specifics of implementation for the Phoenix Point
This information was uncovered by Japanese-language hardware blog Coelacanth's Dream
, who points out
that of the four device IDs, three of them—GFX1100, GFX1101, and GFX1102—are discrete GPUs, corresponding respectively to Navi 31, Navi 32, and Navi 33. The next device ID, GFX1103, seems to match the integrated graphics in AMD's future Phoenix Point APUs, which will pair Zen 4 CPU cores with RDNA 3 graphics.
AMD display driver patch setting up DCN 3.1.4 for Phoenix Point APUs. Source: Coelacanth's Dream
Curiously, the display controller revision for Phoenix Point appears to be older than that for the RDNA 3 discrete GPUs
, although there doesn't seem to be any practical difference in the two revisions. All of the mentioned products appear to max out at four display outputs—impressive for an APU, but only average for a high-end discrete GPU.
The data also seems to indicate that contrary to certain rumors, and like the company's current-generation processors with graphics, the Phoenix Point processors will not have Infinity Cache, at least in their initial iteration. That's not really a surprise, but it could help keep people from getting their hopes up.