Thursday, May 14th 2026

AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

May 14th, 2026 06:22 Discuss (252 Comments)

AMD has responded to what gamers have been most vocal about: bringing its FSR 4.1 upscaling technology to older RDNA 3 and RDNA 2 GPUs. Today, we learn that this dream is becoming a reality, as AMD is officially introducing its FSR Upscaling 4.1 technology to millions of Radeon RX 7000 Series graphics cards worldwide. Arriving this July, millions of RDNA 3 gamers will enjoy official support for FSR Upscaling 4.1 in over 300 supported games. This means that, just like the latest RDNA 4-based Radeon RX 9000 Series, RDNA 3 gamers will also benefit from the upscaling algorithms based on various machine learning techniques without needing to upgrade to the latest Radeon RX 9000 GPUs if their current GPU performs adequately.

Interestingly, the FSR Upscaling 4.1 will use the same underlying model on RDNA 3 as it does on RDNA 4. However, differences in hardware create some background variations, as RDNA 4 has FP8 processing units and INT8, whereas RDNA 3 only has INT8. AMD explains that the company has developed FSR Upscaling 4.1 with the FP8 format in mind, while the latest ML algorithm has been optimized and ported for the INT8 data format to work with RDNA 3. This difference in hardware support delayed AMD's port of FSR technologies to RDNA GPU IP below version four, but it is now finally making an appearance.

Finally, FSR Upscaling 4.1 is also coming to older RDNA 2 GPUs, but that port will have to wait until early 2027. This indicates that AMD is doing additional optimization for RDNA 2 cards, as it is challenging to port this upscaling algorithm to hardware that isn't optimal for the latest techniques. However, it is promising that it is possible, and RDNA 2 gamers will just have to wait a bit longer.

Source: Jack Huynh on X

Add your own comment

252 Comments on AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

#126

Event Horizon

Tek-CheckWhere does this table with AI TOPS come from? There's something odd about it. Would you, kindly, post a link?

I got the chart from this video around the 11:25 mark:

I also used a search engine to verify that the numbers for PS5, PS5 Pro, Xbox X, and 7900XTX were accurate. In the case of the base PS5, we are comparing FP16 with INT8, but I think that further reinforces the point on why Sony probably won't/can't bring PSSR2 to the base model.

#127

jnv11

Visible NoiseLink?

The announcement doesn’t say a thing about integrated graphics. It specifically calls out 6000 and 7000 series discrete GPUs.

I posted a video showing that iGPUs starting with RDNA 2-based iGPUs will likely get FSR as shown in the post below that I made earlier in this thread that contains the video showing that the iGPUs will get FSR 4.1 at the 8:22 time stamp in the video:

jnv11It looks like AMD's failure to commit to supporting FSR 4 on RDNA 3 cards was causing game developers to not support FSR 4 or at least consider supporting it low priority as seen in the following video:

After the announcement of official FSR 4.1 support on RDNA 2 and 3, game developers are changing their priorities to make FSR 4.1 support in their games a high priority task.

EDIT: Add relevant time stamp

#128

AsRock

TPU addict

Dr. DroWhich is exactly why performative garbage PR posts disgust me.

"As a lifelong gamer I spend a lot of time thinking about..." no, you don't.

We've known it's possible before it even released for RX 9000 GPUs. AMD took every step, technical and legal, to block it until now.

Yes we knew but it would have to be going though so much testing to get out and looks like a lot more for the 6000 range.

If that's the case why is the 6000 range delayed to next year ?.

#129

Visible Noise

jnv11I posted a video showing that iGPUs starting with RDNA 2-based iGPUs will likely get FSR as shown in the post below that I made earlier in this thread that contains the video showing that the iGPUs will get FSR 4.1 at the 8:22 time stamp in the video:

EDIT: Add relevant time stamp

Got anything official? I don’t pay attention to random YouTubers.

#130

wolf

Better Than Native

Tek-Check'Great crime'? Right... The 'crime' is so 'great' that buyers of APUs increased their market share by 6% in the last year alone. Such a 'great crime'. Oh boy... You must admit that millions of buyers were so gullible that they willingly ended up with the worst upscaler on the market.

I'd wager there is a fair disparity in what features and performance levels buyers of APU's and discrete GPU's are looking for. Both sets of products can intersect but also tend to suit vastly different types of buyers/product needs/product expectations. APU is also extremely broad here, from laptops and NUC's not aimed at gaming, to gaming handhelds and mobile productivity beasts. A discreet GPU is a far more focussed product.

Tek-CheckNot really. You said more than that, quote: "These [RDNA 1,2,3] severely lack in features that the most basic, cheapest entry level Turing card from 2018 has". If you want to look at it that way, you could also say that even 5080, and for that matter all previous class 80 cards, severely lack in features, basic hardware features such as enough VRAM for the class of cards and their price.

Well I'll call this a stretch and a half. Calling VRAM a feature for one when every discreet card has VRAM, it's hardly a feature in context to the topic of this thread or discussion, it's an amount of a particular specification. You say severely lack in features, but named one and it's not really a feature. I aldo disagree that 80 class cards don't have enough, it's a personal distinction that you think there isn't enough for the class or price. I don't see the correlation between a 5080 having 16GB, RDNA2-3.5 lacking in features (in context to the topic of course) and Turing as an architecture possessing those features.

#131

Dr. Dro

AsRockYes we knew but it would have to be going though so much testing to get out and looks like a lot more for the 6000 range.

If that's the case why is the 6000 range delayed to next year ?.

Come on, dude. Don't be naïve. You know full well the reason this was delayed so much, and why RX 6000 series are being delayed even further, is not because of testing or optimization.

#132

Tek-Check

wolfI'd wager there is a fair disparity in what features and performance levels buyers of APU's and discrete GPU's are looking for. Both sets of products can intersect but also tend to suit vastly different types of buyers/product needs/product expectations. APU is also extremely broad here, from laptops and NUC's not aimed at gaming, to gaming handhelds and mobile productivity beasts. A discreet GPU is a far more focussed product.

I agree with this.

wolfCalling VRAM a feature for one when every discreet card has VRAM, it's hardly a feature in context to the topic of this thread or discussion, it's an amount of a particular specification.

Here, it's more semantics. Nvidia explicitly lists the amount of VRAM as a feature. They also list it in the specification.
www.nvidia.com/en-us/products/workstations/rtx-5000/

wolfI also disagree that 80 class cards don't have enough, it's a personal distinction that you think there isn't enough for the class or price.

Sure. That's exactly why I don't buy it. Nvidia is free to dictate to consumers what they think it's enough. You and I are free to accept or reject such proposition.

#133

JustBenching

Tek-CheckNot really. You said more than that, quote: "These [RDNA 1,2,3] severely lack in features that the most basic, cheapest entry level Turing card from 2018 has". If you want to look at it that way, you could also say that even 5080, and for that matter all previous class 80 cards, severely lack in features, basic hardware features such as enough VRAM for the class of cards and their price. It's the main reason I have never bought class 80 card from Nvidia. They destroyed this class of cards by restricting VRAM. Such as shame.

Nvidia offers the cards with the highest vram on the market for the last 6 years.

#134

Tek-Check

JustBenchingNvidia offers the cards with the highest vram on the market for the last 6 years.

Yes, the weather in towns across California is usually sunny.

#135

wolf

Better Than Native

Tek-CheckHere, it's more semantics. Nvidia explicitly lists the amount of VRAM as a feature. They also list it in the specification.

Yeah I agree too, it's 100% semantics. That's why I mean in relation to context of this topic and discussion, I don't consider VRAM amount of a given part within a generational product stack a feature in that respect, as it pertains to the amount of that specification that it possesses, rather than if it has a sort of yes/no level of capability to perform a function/utlise software (game or otherwise) features at all.

These charts are more what I mean, where the generation on the whole determines (first party) feature support, and within that generation you can pick the level of compute, VRAM etc you desire.

#136

Sol_Badguy

JustBenchingNvidia offers the cards with the highest vram on the market for the last 6 years.

Hmm...
The 7900 XTX which was/is a 4080/4080 Super competitor has 24GB vs 16GB.
The 7900 XT which was/is a 4070 Ti Super competitor has 20GB vs 16GB.
The 7900 GRE which was/is a 4070 Super competitor has 16GB vs 12GB.
The 7800 XT which was/is a 4070 competitor has 16GB vs 12GB.

#137

Tek-Check

Dr. DroCome on, dude. Don't be naïve. You know full well the reason this was delayed so much, and why RX 6000 series are being delayed even further, is not because of testing or optimization.

So far, we know that Sony wanted some primacy of the technology for PS5 Pro in the first place, and also there are no new GPU products until deep next year, so they need to do something.

#138

JustBenching

Sol_BadguyHmm...
The 7900 XTX which was/is a 4080/4080 Super competitor has 24GB vs 16GB.
The 7900 XT which was/is a 4070 Ti Super competitor has 20GB vs 16GB.
The 7900 GRE which was/is a 4070 Super competitor has 16GB vs 12GB.
The 7800 XT which was/is a 4070 competitor has 16GB vs 12GB.

The comparison is irrelevant. If you needed 24gb of vram in 2020, it doesn't matter whether or not the 6800xt had more or less than the 3080, because you needed 24, and therefore youd have to buy nvidia. If you needed 24gb in 2022, it doesn't matter whether or not the 7800xt or the 7900gre has more than 4070super, youd still need to buy a 4090 or a 7900xtx.

For example, im trying to play bannerlords right now and im consistently running out of vram and crashing on my 4090. It doesn't matter that the X card has more than Y card, I need more than 24 and only nvidia offers that. I'm not even sure if 32 will be enough though to be fair, might need more than that.

#139

Tek-Check

JustBenchingThe comparison is irrelevant. If you needed 24gb of vram in 2020, it doesn't matter whether or not the 6800xt had more or less than the 3080, because you needed 24, and therefore youd have to buy nvidia. If you needed 24gb in 2022, it doesn't matter whether or not the 7800xt or the 7900gre has more than 4070super, youd still need to buy a 4090 or a 7900xtx.

Your original claim was generic and irrelevant, and now you are desperately attempting to wriggle out of it. It's not going to work, ever.

If you needed 24GB back in 2020 for games, you were in less than 1% of gaming population who would buy a halo card. That's how 'widespread' that need was. You were right in this particular case, but it's so niche that makes your text petty, fringe and cringe.

JustBenchingFor example, im trying to play bannerlords right now and im consistently running out of vram and crashing on my 4090. It doesn't matter that the X card has more than Y card, I need more than 24 and only nvidia offers that. I'm not even sure if 32 will be enough though to be fair, might need more than that.

The "problem" of less than 1% of people in the first world countries. Your voice truly represents gamers' VRAM needs. What a nonsense and waste of debate time.

#140

JustBenching

Tek-CheckYour original claim was generic and irrelevant, and now you are desperately attempting to wriggle out of it. It's not going to work, ever.

If you needed 24GB back in 2020 for games, you were in less than 1% of gaming population who would buy a halo card. That's how 'widespread' that need was. You were right in this particular case, but it's so niche that makes your text petty, fringe and cringe.

The "problem" of less than 1% of people in the first world countries. Your voice truly represents gamers' VRAM needs. What a nonsense and waste of debate time.

Ah, so your personal preference of needing more than 16 are debate worthy but my needs of over 24 aren't, gotcha. Thanks.

#141

Tek-Check

wolfYeah I agree too, it's 100% semantics. That's why I mean in relation to context of this topic and discussion, I don't consider VRAM amount of a given part within a generational product stack a feature in that respect, as it pertains to the amount of that specification that it possesses, rather than if it has a sort of yes/no level of capability to perform a function/utlise software (game or otherwise) features at all.

Yes to this. It's even more complex. And yes to the amount of VRAM too, as it absolutely contributes to the level of capability in conjunction with those other features and can often be treated as yes/no feature, or part of specification, you name it.

We have seen time and again just how broken increasing amount of games are with insufficient VRAM, especially when users turn on those other features

Just recently Hardware Unboxed showed, yet again, how broken 8GB cards are, especially when upscaler, RT and other features are engaged. Have a watch.

In this instance, we could confidently say that:
8GB is NO level of capability in growing number of games, and
more VRAM is YES level of capability. It becomes a feature, not only specification, when people find it broken or a limiting factor. Planned obsolescence is a carefully crafted product strategy.

JustBenchingAh, so your personal preference of needing more than 16 are debate worthy but my needs of over 24 aren't, gotcha. Thanks.

Not really. Twisting words is not going to work either. You could do much better than this.

#142

wolf

Better Than Native

Tek-CheckYes to this. It's even more complex.

Definitely, and these types of features are the core of the discussion here.

I'm not too interested to go down a VRAM rabbit hole in this thread, though, despite varying amounts of agreement and disagreement I think it's increasingly off topic.

Like usual (for my commentary at least), product buying decisions are a multivariate assessment of all available information including but not limited to, my performance and productivity needs, specs (amounts of onboard hardware), local price, features, support and expected lifecycle, power consumption, size, efficiency and more.

I sincerely hope this situation factors into AMD's future feature support decisions, or at least how they handle potential leaks and PR blunders that lead them to situations like this. How they handled the Starfield DLSS fiasco was far from great at the time, but it appears since then they didn't land themselves in the very same hot water, and that's the sentiment I hope for out of this at least.

#143

zenlessyank

After gaming on a laptop with a gtx 1060 6GB for about 6 years, I wanted to build a new desktop so I scoured the benchmarks for new cards and the 7900XTX was trading blows with a RTX 4080 but came with more VRAM so I bought one even though I have long seated reservations about ATI/AMD.

I didn't really know or care about 'frame generation' at the time. Of course things change and DLSS, XeSS, and FSR is something I would learn more about. Throw in AI, which of course, I had head about and curiosity began to make me wonder what it took to get either working on this GPU. Eventually AMD released an AI 'package' that I installed and found interesting since I didn't have to connect to some AI server to utilize. Also there were ways to enable FSR on older cards but took some leg work. The forums were awash with ridicule and moaning about this 'missing' feature which was amusing and now this is coming our way.

I have noticed in a couple of games I have that they already have FSR 4 on my 7900xtx working in them that I assume is what the game developers did on their own dime.

This is just my experience with the situation. Cheers.

#144

Skimba

RecusRIP optimization.

Sad but true. But its not AMD's fault that developers don't optimize their games anymore. And AMD wasn't the first to start this AI everywhere race either. It was Nvidia.

Thou AMD introducing their solution is better than nothing. Unreal Engine example is so horribly experience without optimization. That upscaler type of solutions are just needed sadly. Rare are those developers who these days have actually time to optimize their games.

#145

Tek-Check

wolfDefinitely, and these types of features are the core of the discussion here.

We can overwhelmingly say it's a good thing that AMD is finally bringing the most recent upscaler to older cards. It took some time, there were complexities in a deal with Sony and no new dGPU hardware from either vendor entil deep next year. I don't think there is much else for gamers to celebrate in the age of slowly dying DIY market due to enormous and absurd prices of memory and storage.

#146

Dr. Dro

Tek-CheckSo far, we know that Sony wanted some primacy of the technology for PS5 Pro in the first place, and also there are no new GPU products until deep next year, so they need to do something.

This has nothing to do with Sony. This pathological defense of AMD baffles me. You can make an argument for so many things, but denying FSR 4 to their own customers just isn't one of them.

#147

AusWolf

JustBenchingAh, so your personal preference of needing more than 16 are debate worthy but my needs of over 24 aren't, gotcha. Thanks.

It's not a matter of personal preference, but the fact that most people buy midrange. That's the market that Nvidia (purposefully) starves of VRAM. If you can't see the world beyond x90 halo cards, it's not @Tek-Check's problem, and the discussion is really pointless.

#148

Tek-Check

Dr. DroThis has nothing to do with Sony. This pathological defense of AMD baffles me. You can make an argument for so many things, but denying FSR 4 to their own customers just isn't one of them.

Two things:
1. FSR4 is coming
2. You need to be willing to see the world more broadly and ask yourself why did Sony invest resources and manpower to develop PSSR for their console only, in collaboration with AMD folks. They did not do it to prop-up competitor's Xbox console, which also runs on AMD hardware. Just stop for a moment and think about it.

#149

JustBenching

AusWolfIt's not a matter of personal preference, but the fact that most people buy midrange. That's the market that Nvidia (purposefully) starves of VRAM. If you can't see the world beyond x90 halo cards, it's not @Tek-Check's problem, and the discussion is really pointless.

How are they starved though, really don't see it. They offer the same amount of vram in most of their lineup. Only ampere was "starved" and even that didn't apply universally to their midrange, since 3060 had more vram than whatever amd was offering at the time. Saying nvidia is vram starved in the midrange is akin to saying amd is vram starved in the lower end (because of the aforementioned 3060). People just like repeating stuff that aren't true.

#150

medi01

The accusation was that shit is restricted The Filthy Green style: to force to upgrade.

This looks like an actual porting effort is needed.

GodisanAtheistpeople have sworn off AMD,n possibly picked up a 5xxx series instead

Add your own comment

AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

252 Comments on AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts

AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

Related News

252 Comments on AMD FSR Upscaling 4.1 Officially Arrives to Radeon RX 7000/6000 GPUs

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts