Sunday, February 8th 2026

AMD on FSR 4 for RDNA 3 and Older GPUs: "No Updates to Share at This Time."

AMD's FidelityFX Super Resolution 4 technology, now known simply as FSR 4, is currently supported in many games, but not across all AMD RDNA GPU generations. In response to an inquiry from Hardware Unboxed, AMD mentioned that it is still uncertain whether official FSR 4 support will be extended to the Radeon RX 7000 series and older GPUs, as the company reportedly has "no updates to share at this time." AMD official product separation stems from its RDNA 4 architecture and the support for 8-bit floating point instructions. While the latest RDNA 4 hardware supports Wave Matrix Multiply Accumulate in FP8 format, older RDNA generations like RDNA 3 and RDNA 2 lack this hardware instruction support and can't process 8-bit floating point data in this format.

However, older Radeon GPUs can instead rely on the 8-bit integer (INT8) data formats, which Radeon RX 7000 series fully supports. AMD accidentally leaked FSR 4 INT8 on its AMD GPUOpen platform, showing that FSR 4 on older GPUs is a possibility, which is just kept hidden for now. Later on, ComputerBase tested this leaked library, finding that FSR 4 offers a balance between native image quality and FSR 3.1 performance on both RDNA 3 and RDNA 2 hardware. In tests with Cyberpunk 2077 in 4K on Ultra settings using the AMD Radeon RX 7900 XTX, FSR 4 delivered 11% faster performance than native, but was 16% slower than FSR 3.1. Interestingly, performance may be the reason why AMD is holding these INT8 FPR 4 libraries back, but another point could be product separation.
AMD already segregates its latest FSR "Redstone" suite of technologies, with features like Ray Regeneration and Radiance Caching exclusive to RDNA 4 hardware in the Radeon RX 9000 series of GPUs. Other basic technologies like upscaling and frame generation are supported on older RDNA 3/2/1 generations, but use an FSR 3.1 Fallback, with no FSR 4 support for now. However, since INT8-based FSR 4 exists, it may only be a matter of time before the company releases the capability to older GPUs, but the expected performance might not be there.
Source: Hardware Unboxed
Add your own comment

110 Comments on AMD on FSR 4 for RDNA 3 and Older GPUs: "No Updates to Share at This Time."

#26
Mack4285
LuminescentWho knows what's behind all of this, it is strange, if the tech is there, why not release it, it doesn't cost them anything.
Maybe they are not allowed to do this, they might have an agreement to not do this, Nvidia and AMD do talk, they are enemies in public but privately they can strike deals and agree to milk us dry, AMD should thank Nvidia for what a GPU ended up to be and do.
There's nothing strange. AMD has passed the "goodwill" period toward consumers. They feel big enough now to focus profits and take a dump on consumers. Every public company walks this path.
Posted on Reply
#27
wolf
Better Than Native
Didn't take long for the subject to be attempted to be gatekept, and in some cases by users most vocal in Nvidia threads lol.

Heaven forbid we look beyond TPU for opinion (and well, it's been confirmed here too that RDNA2-3.5 users want it), and want AMD to do better so we can confidently recommend and buy their products and feel like we'll get long term support.
Posted on Reply
#28
melkor_unlimited
Just look around who whines most about not supporting FSR4 ML upscaling on older HW officially
- owners of nVidia cards
- owners of RX 7900XTX

And who should be concerned most?
- owners of APUs where RDNA3/RDNA3.5 will last for a long long time.
Posted on Reply
#29
Dr. Dro
"Get stuffed, there ain't no money in it for us, so we won't do it" - nGreedia Abandon My Drivers, Inc.
Posted on Reply
#30
wNotyarD
As someone who has tried FSR4 on my 7900XTX just for the sake of it in a single game, which has a paltry FSR3.x implementation and even worse TSR with no option to turn it off (thank you UE5), I say it has the potential to be the real deal, even if inferior to FP8.
Will I cry on the internet because AMD doesn't spend time and money to bring it officially? No because I never bought my card with upscaling in mind, but it does leave a bitter taste...
Posted on Reply
#31
Harun8
I think the biggest crime AMD has committed here is not only not open sourcing their code, but blocking attempts others from it. They blocked Valve from enabling FP8 emulation in RDNA3 in Proton github.com/HansKristian-Work/vkd3d-proton/commit/dd0babcf5df15e9e3826c3610759695ce6c8627c

And one benefit of the doubt you could give AMD is that they developed an INT8 variant in the first place

Otherwise, FSR4 doesn't interest me much and I don't think most Radeon owners should care. I bought my 9070 XT because I wanted something cheaper with inferior upscaling. Also the game support is pretty awful. 6 months with a 9070 XT and I've played through 1 game with FSR4. I use XeSS more often than FSR

Also, it doesn't increase FPS much on RDNA2/3 with INT8, and especially FP8 emulation, to extend the life of the card
AusWolfFSR 4 is supported by every game that has FSR 3.1 in it. I don't know how much wider support can get.
DLSS is in ~700 games, FSR3.1 is in ~200.

FSR4 doesn't even support Vulkan, it's unimpressive imo
Posted on Reply
#32
Grimling
If this limitation of FSR4 would only effect OLD products, i could understand.
But the fact that brand new Products from AMD, like Strix Halo, that are based on RDNA3 variants, and therefore limited to FSR3.1 (which is absolute crap), is absolutely unacceptable!
That we are even talking about this is unacceptable!
AMD should have released FSR4 for RDNA3 Variants with the official release of FSR4, even if there are some limitations of it on RDNA3.
Everything is better that FSR3!
Posted on Reply
#33
Dr. Dro
wNotyarDAs someone who has tried FSR4 on my 7900XTX just for the sake of it in a single game, which has a paltry FSR3.x implementation and even worse TSR with no option to turn it off (thank you UE5), I say it has the potential to be the real deal, even if inferior to FP8.
Will I cry on the internet because AMD doesn't spend time and money to bring it officially? No because I never bought my card with upscaling in mind, but it does leave a bitter taste...
This assumption that it's "inferior" is because any and all evidence we have is based on a prototype DLL from an unintended early SDK leak that you were never supposed to have seen. This headache is largely self inflicted, because AMD intended from the start to tell all of you to get bent. The Linux community is resourceful because they can work around pretty much anything. This is not a luxury Windows users who rely on proprietary stacks running on proprietary software have.

AMD could perfectly have spent the engineering time and resources to optimize the INT8 version to look the exact same as the final one, but they didn't. Instead they just retired driver support for RDNA 1 and 2, and then halfheartedly told their upset customers "we promise you'll still get game updates!11one" when the backlash hit, but didn't walk back one step on their plan. Instead they hid the driver versioning from download pages, and release 3 different driver versions under the same name. It would be really amusing if it wasn't sad. I understand why they did it, but I do not necessarily think they are doing right by their customers either. It's a delicate situation.

Anyone that staunchly defends this company suffers from some sort of Nvidia derangement syndrome at this point, my conviction regarding this is solid as steel
Posted on Reply
#34
nguyen
Here we are seeing the reason why Radeon owners have begrudgingly migrated to Nvidia over the year and stay with Nvidia LOL.

RTG has not been catering to PC gamers and tech lovers for the past decade, they just provide ramen level of GPU just so people could barely play game and most importantly participate in online arguments
Posted on Reply
#35
wolf
Better Than Native
RejZoRNot only that, people don't mention that while RTX 2000 and 3000 series support latest models, no one mentions both have severe performance penalties when using newer models.
And yet it's absolutely fantastic that they did it anyway and put the choice in the users hands, rather than making the choice for them. I've been using the new L and M presets on my 3080 rig, as has my best mate too. It's genuinely shocking how good preset L looks in Ultra Performance mode at 4k, or perhaps my new favourite scaling factor, a custom forced 42% render scale fed into preset L (UP is 33% and P is 50%). The performance penalties are absolutely not so severe that they make the new models undesirable, on the whole, for all users of all supported hardware. I want that comparable choice for RDNA2-3.5 users with FSR4.

Nvidia have legitimately paved the way for worse performing but better looking upscaling to be backported, to the joy of it's users, so maybe AMD can now do what they arguably do best, and copy them.
outlw6669the only reason I can see for nVidia users complaining about AMD not supporting more users with FSR 4 is the want for a valid competitor.

Now, if they just want this so that nVidia will lower prices, or if they would seriously consider switching to AMD, is another discussion.
The reasons are multiple, for me at least.

-I'm part of several broader communities, in which many have RDNA2-3.5 products, and of that many, many want FSR4 officially. This was also true on a specific TPU thread created to get those testimonies.
-Multiple of my friends on RDNA2-3.5 products, want FSR4 officially. If all these users want it, I want it for them.
-It speaks volumes to AMD's long term feature support for past, current and even future products, which is an important consideration for personal purchases or advice I would give anyone making a GPU purchase, both new and used.
-As a collector and system builder that's often with second hand parts, factors like these are considered when selecting or recommending parts, as they are a core feature/capability circa 2026.

But in this day and age forum in particular, to some, makes me an AMD hater. The constructive criticism, wanting AMD to be better and hand their existing users a big, easy win (as well as their own brand perception) is met with attempts at gatekeeping and invalidation because of irrelevant criteria.
Posted on Reply
#36
Dirt Chip
Remember AMD saying they will focus on getting grasp on the mid-level market:
AMD Confirms Retreat from the Enthusiast GPU Segment, to Focus on Gaining Market-Share (9.9.2024, TPU).

So here you go and please understand: AMD must priorities the development on next gen RDNA to replace the current (4) in order to finally drop support to the latter and by doing so achieve their multi-year goal. You can't possibly expect tham to invest on already sold product, that obviously not how you up your market shere. You need to allocate all resources to the future, not the past.
Else, how could you drive those sub 10% market consumers to keep buy AMD?


What a shitshow.
Posted on Reply
#37
wNotyarD
Dr. DroThis assumption that it's "inferior" is because any and all evidence we have is based on a prototype DLL from an unintended early SDK leak that you were never supposed to have seen.

AMD could perfectly have spent the engineering time and resources to optimize the INT8 version to look the exact same as the final one
I call it inferior not because of image quality, it could be close enough between INT8 and FP8 FSR4 to the point I'd just shrug it off, but because of performance. Then again, I'd rather opt for no upscaling if I can, but in the example I mentioned (Immortals of Aveum) the hacked FSR4 made the game look better and not by little.
Just goes on to show my early prejudice (caused by DLSS when I had a 3070, mind you) may just be getting in my way as tech currently evolves.
Posted on Reply
#38
Harun8
FSR4 image quality is pretty good. I was impressed with it, being used to mostly DLSS3 on an old 3060 Ti
Posted on Reply
#39
TheinsanegamerN
Hecate91People seem to conveniently forget this when mentioning DLSS4.5.
People always seem to forget the REST of DLSS4/4.5 and focus solely on the one thing that doesnt work on older cards, MFG, which doesnt work on older generations, it only produces black frames on anything over 2x.
Hecate91Perhaps there may be a want for product segmentation, as FSR4 wouldn't really be FSR4 on older cards, quality or performance wouldn't be the same. Although I'd want to see an FSR4 beta anyway, doing so would drop the whole "AMD doesn't have my upscaling they don't care about me" arguments. Even though upscaling hasn't done anything good for modern AAA games, it's turned into a crutch for devs to avoid optimizing game engines.
FSR4 on older cards is still FSR4. It doesnt perform as well as on rDNA4 with FP8, but it still performs better than native and looks WAY better then FSR3.
Hecate91And also Interestingly enough those who seem to care the most about AMD FSR are Nvidia users.
Ah yes, nvidia users and their.....6800xts.
Posted on Reply
#40
FierceRed
Once again, AMD never misses an opportunity to miss an opportunity. :shadedshu:
Posted on Reply
#41
TheinsanegamerN
nguyenHere we are seeing the reason why Radeon owners have begrudgingly migrated to Nvidia over the year and stay with Nvidia LOL.

RTG has not been catering to PC gamers and tech lovers for the past decade, they just provide ramen level of GPU just so people could barely play game and most importantly participate in online arguments
Everything AMd has tried to do to avoid just making good GPUs with good drivers over the last 10 years has destroyed their marketshare. Reminds me of the tweet where someone summed up how disastrous AMDs pro tier card efforts have been in claiming any market since 2012.

Posted on Reply
#42
zenlessyank
If only my 7900XTX needed fake frames. But it doesn't since it can actually do 60+FPS on everything!

*fake tears*
Posted on Reply
#43
Dr. Dro
Harun8I think the biggest crime AMD has committed here is not only not open sourcing their code, but blocking attempts others from it. They blocked Valve from enabling FP8 emulation in RDNA3 in Proton github.com/HansKristian-Work/vkd3d-proton/commit/dd0babcf5df15e9e3826c3610759695ce6c8627c
Just read this.



This is modern AMD in a nutshell. Suits run the show, and the developers are just basically code monkeys with no real agency over their work. This has been an endless source of frustration for me, and if I may be honest, more than half the reason I dropped them after the VII.
zenlessyankIf only my 7900XTX needed fake frames. But it doesn't since it can actually do 60+FPS on everything!

*fake tears*
"Fake frames" aren't supposed to make unplayable games playable, they're supposed to smooth the load at high refresh gaming. And if I may be entirely honest, the 7900 XTX isn't all that's cracked up to be. Could be worse for a 3 year old GPU, but if you put it next to an RTX 5080, you'll find it sorely lacking. 5090 then... it's not even a fair comparison.
Posted on Reply
#44
ThomasK
RejZoRNot only that, people don't mention that while RTX 2000 and 3000 series support latest models, no one mentions both have severe performance penalties when using newer models.
Selective criticism, that's what it is.

But what amuses me the most is the fake empathy coming from people who do not own the hardware in question, but feel like they have to come up here and tell people how they should feel about it, just to stir the pot. If you care enough to pay attention, is the same people, over and over.
Posted on Reply
#45
zenlessyank
Dr. DroJust read this.



This is modern AMD in a nutshell. Suits run the show, and the developers are just basically code monkeys with no real agency over their work. This has been an endless source of frustration for me, and if I may be honest, more than half the reason I dropped them after the VII.



"Fake frames" aren't supposed to make unplayable games playable, they're supposed to smooth the load at high refresh gaming. And if I may be entirely honest, the 7900 XTX isn't all that's cracked up to be. Could be worse for a 3 year old GPU, but if you put it next to an RTX 5080, you'll find it sorely lacking. 5090 then... it's not even a fair comparison.
As they say, opinions are like buttholes. I would never have a 5080 next to my 7900XTX because I am not a reviewer. Nor do I feel the need to support a shitbrand company as nVidia is.

I still don't want/need FSR. See how that works?
Posted on Reply
#46
TSiAhmat
wNotyarDWill I cry on the internet because AMD doesn't spend time and money to bring it officially? No because I never bought my card with upscaling in mind, but it does leave a bitter taste...
This sentence very much expresses how I feel about the situation.

(not that it should really matter to anyone)
Posted on Reply
#47
Dr. Dro
ThomasKSelective criticism, that's what it is.

But what amuses me the most is the fake empathy coming from people who do not own the hardware in question, but feel like they have to come up here and tell people how they should feel about it, just to stir the pot. If you care enough to pay attention, is the same people, over and over.
The constant is you and your Nvidia derangement syndrome.

Looks like His Highness' mighty ignore list isn't ignoring too good.
zenlessyankAs they say, opinions are like buttholes. I would never have a 5080 next to my 7900XTX because I am not a reviewer. Nor do I feel the need to support a shitbrand company as nVidia is.

I still don't want/need FSR. See how that works?
I don't disagree, but I'm just saying it's hardly being above receiving meaningful feature upgrades. Like FSR 4's much improved FG over FSR 3's.
Posted on Reply
#48
Harun8
Dr. DroJust read this.

I couldn't believe it when I first read it. It's absurd
Dr. DroThis is modern AMD in a nutshell. Suits run the show, and the developers are just basically code monkeys with no real agency over their work. This has been an endless source of frustration for me, and if I may be honest, more than half the reason I dropped them after the VII.
I still like Radeon, they are cheaper for more raster performance which is what I need, and their gpu drivers aside from FSR are still open source. But I am not loyal to the brand
Posted on Reply
#49
RejZoR
TSiAhmatThis sentence very much expresses how I feel about the situation.
And that's perfectly valid feeling. But in a way we also need to look at the situation how NVIDIA has money to throw around and they felt despite performance penalty, it won't be an issue. AMD on the other hand doesn't have resources to throw around (at least not graphics division) and they aren't sure if it's worth investing resources into officially supporting a feature that will by default be inferior to implementation on RDNA4. They don't want to outright say no because they might one day implement it, but they don't want to promise anything long term either.

Personally, if they are unsure they should just say it, but leave it open. Instead they seem to flip flop around and not say anything. Like, say it's not planned or in the works right now, but they won't say no either. And that would be it. If they one day drop it, people will be happy. And if some moved on to RDNA4 specifically because of this in the meanwhile, then so be it.

Main issue with AMD is really this flip flopping around like they themselves don't know what they are doing or they just really suck at communicating. In which case, they should maybe hire new PR team. AMD released a lot of really good stuff, but they are just so bad at telling the world what they have and what they are bringing.
Posted on Reply
#50
wolf
Better Than Native
ThomasKthe fake empathy coming from people who do not own the hardware in question, but feel like they have to come up here and tell people how they should feel about it
And yet what is actually happening in this thread (If you care enough to pay attention, because it is the same people over and over) is the attempts to invalidate and gatekeep having an opinion and indeed real empathy on this subject because of irrelevant and arbitrary barriers, so really, it's them telling us how we should feel about it.

The lengths some will go to to minimise, deflect and excuse away, in bad faith, what AMD should do better does not help AMD or their users, past, present and future.

Selective criticism indeed.
Posted on Reply
Add your own comment
Nov 13th, 2026 07:34 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts