AMD's Frank Azor pushes back against claim that FSR 4.1 won't be ported to RDNA 3.5 GPUs — says 'no such decision' has been made

News
By
Published

Azor posted about reports on X.

A Ryzen AI Max chip.
(Image credit: AMD)

During Computex, multiple reports cited that AMD was potentially skipping FSR 4.1 integration on RDNA 3.5 iGPUs, leaving these users with no choice but to use older FSR versions. AMD client and graphics marketing executive Frank Azor posted on X pushing back against this claim, suggesting RDNA 3.5 GPUs will indeed support FSR 4.1 in the future.

The source of all this controversy reportedly originated from AMD’s VP and GM of Ryzen and Radeon products, David McAfee. Azor admitted in his X post that he wasn’t able to hear McAfee’s thoughts about FSR 4.1 support on RDNA 3.5 in person, however, he clarified that no decision to drop RDNA 3.5 support from AMD’s latest FSR implementation had been made.

Azor's post is carefully worded, and doesn't make any promises one way or the other. But it does suggest that support for FSR 4.1 on RDNA 3.5 hasn't been dropped just yet. Even though RDNA 3.5 exclusively revolves around integrated GPUs and not discrete GPUs or graphics cards, many gaming-capable handheld devices and laptops come with RDNA 3.5-powered hardware, including Strix Halo-powered laptops featuring GPU hardware equivalent to an RTX 4060.

Latest Videos From

All of these devices make RDNA 3.5 a highly valuable user base to keep supporting with FSR updates. There is also no architectural hardware difference that we are aware of that would prevent AMD from porting FSR 4 to RDNA 3.5. RDNA 3.5 is a incremental update over RDNA 3, primarily geared towards improving power efficiency.

AMD finally announced official FSR 4.1 upscaling support for RDNA 3 and RDNA 2 GPUs last month. The move comes almost a year after AMD accidentally leaked its own FSR 4 version online capable of running on non-RDNA 4 hardware using INT8 instructions. FSR 4.1 is the latest iteration of the FSR feature set, including an improved upscaler that better preserves detail in motion and Ray Regeneration 1.1.

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Aaron Klotz
Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

7 Comments Comment from the forums
  • Notton
    Videocardz vs. Frank Azor?
    These info sources are awful.

    You might as well ask a magic 8 ball.
    Reply
  • usertests
    Notton said:
    Videocardz vs. Frank Azor?
    These info sources are awful.

    You might as well ask a magic 8 ball.
    The source was Hardware Luxx, a German outlet, and they were told it directly from AMD's David McAfee.

    AMD is the awful source in the end.
    Reply
  • DS426
    AMD needs to simplify their support policies: hardware that is in support and software that is in support are supported together, with VERY few exceptions. If something has a huge perf penalty like FSR4 on RDNA2, that caution should be said upfront and then left upon the user. Instead, AMD is trying too hard by determining what has value to who and when, and that's obviously just getting too messy; two different key spokespeople over a product aren't on the same page, so...
    Reply
  • psyconz
    DS426 said:
    AMD needs to simplify their support policies: hardware that is in support and software that is in support are supported together, with VERY few exceptions. If something has a huge perf penalty like FSR4 on RDNA2, that caution should be said upfront and then left upon the user. Instead, AMD is trying too hard by determining what has value to who and when, and that's obviously just getting too messy; two different key spokespeople over a product aren't on the same page, so...
    I get the impression with a lot of these large organisations, that the left hand often doesn't know what the right hand is doing. It's definitely a matter of, between organisations, they might sue each other over one thing, meanwhile, at exactly the same time, be collaborating on things in other areas. It sure is an interesting world we live in. 🤣
    Reply
  • Flyfisherman
    psyconz said:
    I get the impression with a lot of these large organisations, that the left hand often doesn't know what the right hand is doing. It's definitely a matter of, between organisations, they might sue each other over one thing, meanwhile, at exactly the same time, be collaborating on things in other areas. It sure is an interesting world we live in. 🤣
    And sometimes the right hand doesn't know what it's doing either...
    Best regards from Sweden
    Reply
  • psyconz
    Flyfisherman said:
    And sometimes the right hand doesn't know what it's doing either...
    Best regards from Sweden
    🤣🤣🤣 now that's funny =)
    Reply
  • dva852
    Notton said:
    Videocardz vs. Frank Azor?
    These info sources are awful.
    The sources aren't conflicting. It's just the usual distortion you get from sensationalized headlines to get clicks.

    HardwareLuxx cites AMD exec David McAfee, General Manager of AMD Client Division, who said FSR 4.1 "isn't currently planned for RDNA 3.5...Currently the decision seems to be leaning towards 'no'."

    Videocardz re-reported the convo, but its piece title says "AMD FSR 4.1 to skip RDNA 3.5", which oversells what McAfee said. "To skip" makes for a more compelling click than "may skip." I understand and empathize with the clickbait practice, as all "free" sites including this one live on clicks. You recognize it as the cost of "free" and move on.

    Frank Azor then came out with a clarification, essentially backing up McAfee but leaning toward a more neutral stance, to say "no decision has been made." To its credit, Videocardz did update above piece with Azor's statement.

    Videocardz has a follow-up piece, re-reporting from Tom's Guide, with Jack Huynh's statement that "I did not say it's coming " with the usual corporatese about wanting to maintain high-quality bar, yada yada. So, somewhere between Azor and McAfee's blurbs.

    The most informative tidbit is a reporter's (Andreas Schilling) post within the Videocardz piece, stating that "technical aspects were then explained, such as the lower computing power of the integrated solutions and the memory bandwidth, which has implications for FSR 4.1."

    Benchmarks of the leaked INT8 FSR4 have shown it has high overhead, with marginal speedup over native res for RDNA 3. These were ran on dedicated GPUs with ample power. On iGPUs with much lower power envelope, and especially handhelds where battery life is key, the overhead is likely to be prohibitive.

    So, AMD's first statement (McAfee's) is on the nose, and 4.1 isn't likely to happen on iGPUs because they simply aren't fast enough to cover the overhead. No amount of pleading and petitioning from gamers is likely to change that immutable fact.

    Stepping back and viewing it from a larger perspective, gaming isn't a high priority for AMD. AI, and now local AI, are more important. Yet, on local AI, Gorgon Halo is still stuck at RDNA 3.5, with RDNA 5 only coming in Medusa Halo in '27. It indicates that AMD is resource-constrained, and the client division gets less resources than the bread-winner enterprise division.
    Reply