Thursday, October 30th 2025

No More Game Optimizations for AMD Radeon RX 6000 and RX 5000, Still Part of Main Driver Branch

In a rather shocking move, AMD has given the Radeon RX 6000 series and RX 5000 series GPUs a partial retirement by removing it from the regular monthly game optimization cycle. While the monthly driver updates will continue to support them, the game optimizations included will only be meant for RX 7000 series and RX 9000 series. The monthly driver releases will address critical security vulnerabilities and fix bugs. What makes the move surprising is the fact that the Radeon RX 6000 series in particular is less than 4 years old, and gamers who purchased those cards did so at extremely inflated prices because RX 6000 series, along with NVIDIA's RTX 30-series, were released amid the cryptocurrency mining rush that saw miners soak up GPU inventory.

In a statement to PCGH, machine translated from German, AMD says:
RDNA 1 and RDNA 2 graphics cards will continue to receive driver updates for critical security and bug fixes. To focus on optimizing and delivering new and improved technologies for the latest GPUs, AMD Software Adrenalin Edition 25.10.2 is placing Radeon RX 5000 and RX 6000 series graphics cards (RDNA 1 and RDNA 2) into maintenance mode. Future driver updates with targeted game optimizations will focus on RDNA 3 and RDNA 4 GPUs.
Sources: PCGH, VideoCardz
Add your own comment

286 Comments on No More Game Optimizations for AMD Radeon RX 6000 and RX 5000, Still Part of Main Driver Branch

#76
TheinsanegamerN
Vya DomusSo nobody can give me a concrete example of this super long amazing support translating into actual game optimisations all the way till the end. Got it.

Anyone who thinks that Pascal, for example, was receiving game optimisations till the end of driver support is uhm… being silly.
Can you give any examples of pascal falling off?

We've already seen what happened with rDNA2 when 3 came out. The 6800xt went from competing with the 3080 in new releases to competing with the 3070. That magically fixed itself when AMD started optimizing again.
Posted on Reply
#77
BoggledBeagle
Say I bought RX 6900 XT 5 years ago for $1000. It enabled me to play hunreads or thousands of games including all the new games introduced in this period.

Now the performance of this card sits between RX 9600XT and RX 9700. It lost more than half of its value now and is terribly energy inefficient compared to new AMD cards.

You could get 5060 Ti 16 GB for half the price and with half the power draw now with slightly better performance.

Where am I going with this. I need to leave the PC now.

WORK IN PROGRESS

I think I will continue with something like

"this card is no good today anyway and should be thrown away".

and

"You can still enjoy a lot of games even if you decided to keep the card for whatever strange reason."
Posted on Reply
#79
Unregistered
Vya DomusSo nobody can give me a concrete example of this super long amazing support translating into actual game optimisations all the way till the end. Got it.

Anyone who thinks that Pascal, for example, was receiving game optimisations till the end of driver support is uhm… being silly.
The Game-ready drivers still apply performance fixes to Pascal after all these years because most of them are just "game profiles" that tweak various backend settings.

edit:
nguyen1080Ti actually got a driver that boosted its performance 5.5 years after release LOL, meanwhile 6900XT didn't even get 5 years of full driver support
I was going to post that.
#80
enb141
I've been saying this, a few years ago I mentioned that AMD drivers were crap, I got a 6400 because back then getting an NVidia card was an impossible task.

Now the time proved me right.
Posted on Reply
#81
Ssenol
FR3D1This is not entirely correct;

Source ComputerBase (de)

With a friendly smile,
Fred.
You're absolutely right. AMD is losing our trust with its constant inconsistency. There's no reason to buy AMD anymore. Who knows if they won't do something similar to the 9070 XT in the future?
Posted on Reply
#82
damric
Well this sucks. Not for me since I don't really play the latest games, but my wife also has an RX 6900XT and she's gonna be pretty pissed. We were hoping to get another couple years out of that card since it has 16GB and still performs very well at raster.
Posted on Reply
#83
Ssenol
EatingDirtWould be kind of weird to upgrade from a 6900xt to a 5070 Ti/9070XT for ~20-40% more performance. Hell, I'm using a 6800XT and the only worthy performance upgrade in my opinion is the 5090, and I'm not spending that much money. I'm patiently waiting for the next gen GPU's.
This is a deliberate aging blow to us, we should not accept it.
Posted on Reply
#84
LastDudeALive
Dr. DroI'm going to be truthful with you, I am not upset at AMD's decision. But I am also privy to certain details that the general public is not, so that's a very important consideration as to why I am not raising my pitchforks when I would usually be the first to moan about this. I'm not sure I'm allowed to talk about it, so I won't, but just trust me when I say that there actually is a valid technical reason for this decision.
Nvidia could have dropped support for Maxwell and Pascal any time in the last 3 years with no DX12 Ultimate, RT cores, and Tensor cores being a perfectly legitimate "technical reasons." They could have kept DLSS 4 to Blackwell, with the new FP4 support being a perfectly legitimate "technical reason." AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, they can obviously work around technical limitations when they want to. That's a silly excuse that nobody believes.

The problem is that AMD does not understand the marketing power that gaming GPUs provide. Nvidia does. Let's be honest, at this point, the GeForce cards are basically a charity operation. There is so much demand for datacenter GPUs and AI accelerators, that Nvidia would be perfectly justified in ending their consumer operations all together and devoting 100% of their TSMC silicon to the AI business. Every GeForce card could have gone into an AI card and sold for 10x more money instead.

So why does Nvidia keep making consumer GPUs, with all the accompanying expenses (R&D, software support, having an entire datacenter run 24/7/365 improving DLSS, etc.)? Because it's invaluable as a marketing tool. GeForce might be "money lost" in opportunity costs. But DLSS and Frame Generation means Nvidia presents itself as the best AI company to the general public. The wide availability of GeForce cards with tensor cores and CUDA support meant that ordinary people could experiment with AI and Nvidia's software stack. Having "the world's most powerful gaming GPU" means your company is percieved as the best, even if it's $3000 and extremely rare.

AMD does not do this. Their RDNA product lines are extremely disjointed and artificially segmented. They shot themselves in the foot by not supporting DX 12 Ultimate on RDNA 1. ROCm support on consumer cards is an afterthought provided months or years after release, if their cards are supported at all. Their flagship cards haven't competed with Nvidia for close to a decade now. Nobody looks at their consumer GPUs and goes "wow, AMD is a great company."
Posted on Reply
#85
HairyLobsters
Vayra86AMD is not doing itself any favors here. Penny wise pound stupid
Drivers will still be supported, just not optimization for new games. All your old games will still work on your old video card.
Posted on Reply
#86
DaemonForce
enb141I've been saying this, a few years ago I mentioned that AMD drivers were crap, I got a 6400 because back then getting an NVidia card was an impossible task.
How does that in any way make sense? If I can't get the card I want, I get something equal or bide my time.
This year I bought a RX9070XT to sunset my worn and aging RX580 but I'm still holding onto it.
I also bought a P620 less than a week later because I'm not gonna pass up NVENC at $30.
You don't even have an encoder engine on the RX6400. Why the hell would anyone get it during the coof?
A display out? The RX6750XT was much more capable in price to feature+performance but it doesn't matter.
When you are glued to crap cards, expect crap drivers. The normal Wattman app is probably too much for you.
Posted on Reply
#87
Dr. Dro
Vya DomusAnyone who thinks that Pascal, for example, was receiving game optimisations till the end of driver support is uhm… being silly.
As shown above. Many of the driver optimizations apply to all hardware targets because of the shared user mode content and configuration profiling. Many of these benefits can be felt even on a GTX 900 series card.
persondbIf there are technical reasons why then they shouldn't have pushed RDNA 2 for so long and still are. The latest RDNA 2 dGPU was from Oct 2023, not to mention the APUs, with Rembrandt-R being from early 2023, Granite Ridge (desktop Zen 5) from mid 2024, the re-refreshed Rembrandt that they announced some days ago...

(Steam Deck doesn't matter since it's supported by linux drivers)

If there is some technical flaw that makes RDNA 2 unable to cope with some modern features then they shouldn't have kept releasing new products with it for so long. Those features are actually known years in advance to them so they can actually prepare new hardware...

And for some things, you can do software fallbacks like how Nvidia did a software path for Pascal in RT, which sucked, but at least worked or how you can run RT-only titles in linux with Vegas/RDNA 1 and it can work surprisingly well. But AMD just seems to be completely unwilling to spend on software.

At this point, I really guess that Nvidia is the actual finewine company. Anyone who brought a Turing card back in 2018 can still enjoy using the latest DLSS 4 and it seems that after months, maybe FSR 4 will come at most to RDNA 3.
Indeed, I agree. What will happen here going forward is that no features or major optimizations that require the kernel mode component to be changed in a substantial manner will be done going forward.

Optimizations and features that function at the user mode level can be backported to the 25.10 branch, and this is what AMD has committed to providing going forward with the "maintenance mode" schedule. I don't think FSR is a kernel mode function, so regarding that concern, I don't think this changes anything, at least regarding FSR 4. Redstone, perhaps, maybe, I don't know what changes would be necessary to make that tick.
LastDudeALiveNvidia could have dropped support for Maxwell and Pascal any time in the last 3 years with no DX12 Ultimate, RT cores, and Tensor cores being a perfectly legitimate "technical reasons." They could have kept DLSS 4 to Blackwell, with the new FP4 support being a perfectly legitimate "technical reason." AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, they can obviously work around technical limitations when they want to. That's a silly excuse that nobody believes.

The problem is that AMD does not understand the marketing power that gaming GPUs provide. Nvidia does. Let's be honest, at this point, the GeForce cards are basically a charity operation. There is so much demand for datacenter GPUs and AI accelerators, that Nvidia would be perfectly justified in ending their consumer operations all together and devoting 100% of their TSMC silicon to the AI business. Every GeForce card could have gone into an AI card and sold for 10x more money instead.

So why does Nvidia keep making consumer GPUs, with all the accompanying expenses (R&D, software support, having an entire datacenter run 24/7/365 improving DLSS, etc.)? Because it's invaluable as a marketing tool. GeForce might be "money lost" in opportunity costs. But DLSS and Frame Generation means Nvidia presents itself as the best AI company to the general public. The wide availability of GeForce cards with tensor cores and CUDA support meant that ordinary people could experiment with AI and Nvidia's software stack. Having "the world's most powerful gaming GPU" means your company is percieved as the best, even if it's $3000 and extremely rare.

AMD does not do this. Their RDNA product lines are extremely disjointed and artificially segmented. They shot themselves in the foot by not supporting DX 12 Ultimate on RDNA 1. ROCm support on consumer cards is an afterthought provided months or years after release, if their cards are supported at all. Their flagship cards haven't competed with Nvidia for close to a decade now. Nobody looks at their consumer GPUs and goes "wow, AMD is a great company."
You got it all right, down to the last detail. I'm in full agreement, I just have to point out the elephant in the room, though. The true problem here is technical debt, and fundamentally why I make the argument that this is a positive thing in the long run. The changes they made in the structure overall (and the reason RDNA 2 had to go) are directly pertaining to this.

Nvidia spent the R&D over the years, always developing, supporting and making every optional API extension under the sun at least functional, if not performant. AMD, on the other hand, has always taken the shallower road that required the least engineering and development time possible. This is what made it possible for Nvidia to support their cards for so long, mostly because the work was already done. You want to use some extension that will help you gain performance, optimize, or run something odd? The Nvidia card will do it. The AMD one will go belly up and croak.

The DirectX 11 driver command lists issue, the software RT driver that Nvidia provided but AMD refused to, RDNA/2 being unable to do HAGS, these are just smaller examples of the big picture I'm trying to present here. People glossed over the 5700 XT forgoing DirectX 12 Ultimate, calling RT needless, and all the other improvements and optimizations "superfluous"... this is your bill, and you should pay it in due. The new work graphs function they just added support for (exclusive to RX 9000 series, btw) in the new driver? Nvidia supports it in Ampere and has done so for over a year.
Posted on Reply
#88
enb141
DaemonForceHow does that in any way make sense? If I can't get the card I want, I get something equal or bide my time.
This year I bought a RX9070XT to sunset my worn and aging RX580 but I'm still holding onto it.
I also bought a P620 less than a week later because I'm not gonna pass up NVENC at $30.
You don't even have an encoder engine on the RX6400. Why the hell would anyone get it during the coof?
A display out? The RX6750XT was much more capable in price to feature+performance but it doesn't matter.
When you are glued to crap cards, expect crap drivers. The normal Wattman app is probably too much for you.
I got the 6400 because I needed a card for my HTPC, nothing fancy, just a low profile card that didn't consume a lot.

Now that HTPC has an RTX 4060. The 6400 is getting dust as their drivers.
Posted on Reply
#90
evernessince
Dr. DroYou got it all right, down to the last detail. I'm in full agreement, I just have to point out the elephant in the room, though. The true problem here is technical debt, and fundamentally why I make the argument that this is a positive thing in the long run. The changes they made in the structure overall (and the reason RDNA 2 had to go) are directly pertaining to this.
That's probably one of the few explanations that makes sense, because switching these cards over to maintenance mode is no small decision.

They are going to catch a ton of deserved flak for this, it would have been wise for them to have a sit down with the press to explain it or release a video regarding it. AMD's unilateral way of doing things is akin to Big Corpa but it would be greatly beneficial for them to collaborate and communicate more with the community. If they have a good reason, they should state it. People are still going to be pissed but at least they'd know why.
Posted on Reply
#91
persondb
LastDudeALiveRDNA product lines are extremely disjointed and artificially segmented. They shot themselves in the foot by not supporting DX 12 Ultimate on RDNA 1
They kinda backed themselves in a corner for that one as the hardware was probably already set in stone when that was decided. AMD those days really hates anything forward looking in terms of features set. GCN was great in that regard but RDNA is always looking like Nvidia Gen -2 in terms of features to implement.

They could have implemented the RT part of DX 12U in software but the rest of it like mesh shaders, VRS and so on aren't really possible. And since those things are decided years in advanced, I bet they probably were kinda aware but just ignored it, maybe thought it wasn't going to be important.

It's sad and funny at the same time as some things like saying they would BVH Traversal at the RT cores (i.e. implementing it in hardware) is something that Nvidia does since Turing...
Posted on Reply
#92
Unregistered
LastDudeALiveNvidia could have dropped support for Maxwell and Pascal any time in the last 3 years with no DX12 Ultimate, RT cores, and Tensor cores being a perfectly legitimate "technical reasons." They could have kept DLSS 4 to Blackwell, with the new FP4 support being a perfectly legitimate "technical reason." AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, they can obviously work around technical limitations when they want to. That's a silly excuse that nobody believes.

The problem is that AMD does not understand the marketing power that gaming GPUs provide. Nvidia does. Let's be honest, at this point, the GeForce cards are basically a charity operation. There is so much demand for datacenter GPUs and AI accelerators, that Nvidia would be perfectly justified in ending their consumer operations all together and devoting 100% of their TSMC silicon to the AI business. Every GeForce card could have gone into an AI card and sold for 10x more money instead.

So why does Nvidia keep making consumer GPUs, with all the accompanying expenses (R&D, software support, having an entire datacenter run 24/7/365 improving DLSS, etc.)? Because it's invaluable as a marketing tool. GeForce might be "money lost" in opportunity costs. But DLSS and Frame Generation means Nvidia presents itself as the best AI company to the general public. The wide availability of GeForce cards with tensor cores and CUDA support meant that ordinary people could experiment with AI and Nvidia's software stack. Having "the world's most powerful gaming GPU" means your company is percieved as the best, even if it's $3000 and extremely rare.

AMD does not do this. Their RDNA product lines are extremely disjointed and artificially segmented. They shot themselves in the foot by not supporting DX 12 Ultimate on RDNA 1. ROCm support on consumer cards is an afterthought provided months or years after release, if their cards are supported at all. Their flagship cards haven't competed with Nvidia for close to a decade now. Nobody looks at their consumer GPUs and goes "wow, AMD is a great company."
You nailed it, great post. The fact that even the lowest end Nvidia "gaming" cards come with the same CUDA API is very attractive, and goes a long way to increasing their market share. With AMD software and functionality has seemingly become an afterthought which is weird because when GCN was first introduced they pushed compute heavily with Khronos.
HairyLobstersDrivers will still be supported, just not optimization for new games. All your old games will still work on your old video card.
So when you're on RDNA2 and a new game comes out, and say it's broken or a stuttering mess, you're basically SOL because it will only get fixed in the new branch that your GPU doesn't support. This is one step away from those yearly legacy drivers they put out which is just flat out stupid. Whoever thought this was a good idea should promptly be fired.
Posted on Edit | Reply
#93
LastDudeALive
Dr. DroI'm not sure I'm allowed to talk about it, so I won't
12 hours later:
Dr. DroYou got it all right, down to the last detail. I'm in full agreement, I just have to point out the elephant in the room, though. The true problem here is technical debt, and fundamentally why I make the argument that this is a positive thing in the long run. The changes they made in the structure overall (and the reason RDNA 2 had to go) are directly pertaining to this.

Nvidia spent the R&D over the years, always developing, supporting and making every optional API extension under the sun at least functional, if not performant. AMD, on the other hand, has always taken the shallower road that required the least engineering and development time possible. This is what made it possible for Nvidia to support their cards for so long, mostly because the work was already done. You want to use some extension that will help you gain performance, optimize, or run something odd? The Nvidia card will do it. The AMD one will go belly up and croak.

The DirectX 11 driver command lists issue, the software RT driver that Nvidia provided but AMD refused to, RDNA/2 being unable to do HAGS, these are just smaller examples of the big picture I'm trying to present here. People glossed over the 5700 XT forgoing DirectX 12 Ultimate, calling RT needless, and all the other improvements and optimizations "superfluous"... this is your bill, and you should pay it in due. The new work graphs function they just added support for (exclusive to RX 9000 series, btw) in the new driver? Nvidia supports it in Ampere and has done so for over a year.
"Technical debt" is an extremely silly excuse. As I said, AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, obviously they can still make the architecture work if it's important to them. You think there's no technical debt in the Turing GTX cards? You think there's no technical debt in maintaining three different versions of DLSS 4 (FP4 for Blackwell, FP8 for Ada, FP16 for Turing and Ampere) and two different versions of Frame Generation (one for Ada with a hardware Optical Flow Accelerator, one for Blackwell where Optical Flow is done by the tensor cores)?

Again, AMD doesn't understand that "technical debt" is a small price to pay for the marketing that consumer GPUs provide. At this point in the AI bubble, all consumer cards are "technical debt." Enterprise AI gives you 10x-100x more money for the silicon. Maybe this decision frees 10 driver engineers up to work on UDNA. But if nobody buys UDNA GPUs because of this decision, AMD has shot themselves in the foot again.
Posted on Reply
#94
R-T-B
Vya DomusOfficially supported driver updates = game optimisations, correct ?
We all know that's a farce but you can probably still expect optimizations at 3 years. I'd say 5 is the cutoff.

Are we getting there yet?
Posted on Reply
#95
Dr. Dro
LastDudeALive12 hours later:

"Technical debt" is an extremely silly excuse. As I said, AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, obviously they can still make the architecture work if it's important to them. You think there's no technical debt in the Turing GTX cards? You think there's no technical debt in maintaining three different versions of DLSS 4 (FP4 for Blackwell, FP8 for Ada, FP16 for Turing and Ampere) and two different versions of Frame Generation (one for Ada with a hardware Optical Flow Accelerator, one for Blackwell where Optical Flow is done by the tensor cores)?

Again, AMD doesn't understand that "technical debt" is a small price to pay for the marketing that consumer GPUs provide. At this point in the AI bubble, all consumer cards are "technical debt." Enterprise AI gives you 10x-100x more money for the silicon. Maybe this decision frees 10 driver engineers up to work on UDNA. But if nobody buys UDNA GPUs because of this decision, AMD has shot themselves in the foot again.
I didn't say what exactly was done, so I held true to that. Anything else you can infer could be figured out by anyone who's not looking at this with blind rage, which is exactly: they dropped RDNA 2 to work on RDNA 3 and newer. I mean, doesn't take a genius to figure that out.

We both know the FSR issue specifically isn't a driver problem but rather, an artificial segmentation decision. The architecture might not be as fast, but it was never a showstopper. And I genuinely do think that there's more that RDNA 2 can offer, since it's DX12 Ultimate compliant, which makes that some well earned scorn. It's more of a PR disaster than anything, though, and I'd be lying if I said that I wouldn't be upset if I had been affected. It's more than a valid argument, just as @persondb mentioned there are products affected that are barely a year old.
Posted on Reply
#96
LastDudeALive
R-T-BWe all know that's a farce but you can probably still expect optimizations at 3 years. I'd say 5 is the cutoff.

Are we getting there yet?
Maxwell cards, first released in 2014, are supported by the latest driver and got optimizations for Battlefield 6 and Bloodlines 2. That's an 11 year lifecycle of optimizations.
Posted on Reply
#97
R-T-B
LastDudeALiveMaxwell cards, first released in 2014, are supported by the latest driver and got optimizations for Battlefield 6 and Bloodlines 2. That's an 11 year lifecycle of optimizations.
Yes, technically old cards are supported (wasn't support for maxwell dropped though?), but I doubt there is much more to wring out left post 5 years or so. But 3? 3 is just too young.
Posted on Reply
#98
Dr. Dro
R-T-BYes, technically old cards are supported (wasn't support for maxwell dropped though?), but I doubt there is much more to wring out left post 5 years or so. But 3? 3 is just too young.
The current Release 580 drivers still support Maxwell, Pascal and Volta. They will be dropped in the next release, which should come between December and January. Maybe a little earlier, but there's no need to release a new branch (and potential problems) during the holiday season I reckon
Posted on Reply
#99
Visible Noise
HairyLobstersDrivers will still be supported, just not optimization for new games. All your old games will still work on your old video card.
Yep, everything is fine as long as you don’t want to play new releases on your two year old AMD GPU.

That’s the equivalent of Nvidia dropping support for the RTX 4090.
Posted on Reply
#100
Hecate91
Dr. DroI only really have one problem with this comparison, AMD must be held to a higher standard than Intel, because it needs to push towards excellence, not mediocrity. Alder Lake's graphics were, are, and will always be... mediocre and passable. RDNA 2, on the other hand, isn't.
It makes no sense why anyone would hold a company to any higher standard than another, besides the obvious double standards.
LastDudeALive"Technical debt" is an extremely silly excuse. As I said, AMD is getting FSR 4 to work on the RDNA 2 GPU of the PS5 Pro, obviously they can still make the architecture work if it's important to them. You think there's no technical debt in the Turing GTX cards? You think there's no technical debt in maintaining three different versions of DLSS 4 (FP4 for Blackwell, FP8 for Ada, FP16 for Turing and Ampere) and two different versions of Frame Generation (one for Ada with a hardware Optical Flow Accelerator, one for Blackwell where Optical Flow is done by the tensor cores)?

Again, AMD doesn't understand that "technical debt" is a small price to pay for the marketing that consumer GPUs provide. At this point in the AI bubble, all consumer cards are "technical debt." Enterprise AI gives you 10x-100x more money for the silicon. Maybe this decision frees 10 driver engineers up to work on UDNA. But if nobody buys UDNA GPUs because of this decision, AMD has shot themselves in the foot again.
It isn't silly when it's not just technical debt, but a lack of financial resources as well on the Radeon side. I doubt AMD would have FSR4 working as well as it does without Sony co-developing the tech as PSSR.
Nvidia keeping driver support for as long as they do is just change found under the sofa cushions for them, a rounding error for a $5 trillion dollar company. And the "optical flow accelerator" turned out to be a bunch of BS as to why previous generations didn't get every feature for quite a while. Even so there never was any outrage for previous gens not getting frame gen.
Consumer cards are especially technical debt for AMD when Nvidia gets closer towards a full monopoly, or who knows how many driver engineers AMD has, freeing up 10 gaming graphics driver engineers for UDNA may help if they only have a team of 30 people.
Posted on Reply
Add your own comment
Nov 14th, 2026 19:05 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts