NVIDIA Grabs Market Share, AMD Loses Ground, and Intel Disappears in Latest dGPU Update

[Tomorrow]

Missed that, and without surprise... it's got basically no fixes over the previous release and has received no new features whatsoever. Boring.

It's not supposed to. It's essentially a security update.

It's still the same old 23.19 driver, codebase from mid 2023 or so. It is not the same 25.5.1 for RDNA cards.

Like you say yourself. It was frozen in 2023. Now it only has security fixes.

You're mistaken about Turing, though. Nvidia still supports hardware much older than that in their mainline driver release. The GTX 900 series (Maxwell) from 2014 are still supported, and only recently they started warning customers that they intend to stop supporting these in a future release.

Supported and receiving new game optimizations are two very different things. You dont really think Nvidia is putting in an effort to optimize new games for Turing do you? They likely just make sure those games run and provides security fixes. That's about it.

The RTX 20 series GPUs (Turing) remain under active development and support every new RTX feature released thus far.

That's clearly false because you yourself acknowledge that FG is not supported. Therefore it cannot support every new RTX feature.
I ran 2080 Ti from 2021 to 2025. If there's one series i can comment from experience it's Turing.

During those years did i receive any meaningful updates via drivers? Only DLSS upscaling. Oh and i think there was the video upscaling in control panel, but that just made the card consume 250W and was useless. In 2024 they finally started to make a new unified control panel at last so i guess that could be considered a new feature.

It is also explicitly compatible with DLSS 4 super resolution and ray reconstruction features,

But not frame generation or multi-frame generation. AMD proved that FG works and works well on 20 series.

the only things missing from 20 and 30 series are frame generation support.

20 series is also missing ReBAR support that could have been an easy add-on. Im still salty over this one.

Transformer model upscaling (preset K on the latest DLL), which is the star feature regardless, is fully supported even on the RTX 2060.

Transformer model runs worse on 20 series than it does on newer cards.

 

[Dr. Dro]

That's clearly false because you yourself acknowledge that FG is not supported. Therefore it cannot support every new RTX feature.
I ran 2080 Ti from 2021 to 2025. If there's one series i can comment from experience it's Turing

FG was not supported on 30 series either, and we know why. It's the only thing that doesn't work, Nvidia claims it's because of the optical flow performance of 20 and 30 series, I personally don't buy it, and I agree that it could be backported if they really wanted to, and even Nvidia engineers have said so in the past. But everything else? Fully supported. If you want to use FSR 4, at least for now, hope you have a RX 9000 GPU

During those years did i receive any meaningful updates via drivers? Only DLSS upscaling. Oh and i think there was the video upscaling in control panel, but that just made the card consume 250W and was useless. In 2024 they finally started to make a new unified control panel at last so i guess that could be considered a new feature.

Yes, yes you have. Very many, and you are well aware of that...

But not frame generation or multi-frame generation. AMD proved that FG works and works well on 20 series.

FSR 3's frame generation component is platform agnostic, so yes, it works.

20 series is also missing ReBAR support that could have been an easy add-on. Im still salty over this one.

Resizable BAR support has absolutely nothing to do with feature support, and to this day the performance advantage of having it enabled is questionable, the only GPUs that need it are Intel's, because the drivers were written from the beginning with that in mind. Nvidia uses app profiles to disable it in many applications, and I'm sure the same happens with AMD

Transformer model runs worse on 20 series than it does on newer cards.

It requires more processing power, so naturally. It also takes longer to infer than the CNN model on newer hardware, including the 50 series. Even then, it's only 5% slower than the CNN model, and it's doubtful that this will cause any perceptible increase in latency unless you're already running at sub 20 fps IMHO. The quality of the image produced is the same across all supported GPUs.

 

[Rover4444]

FSR 3's frame generation component is platform agnostic, so yes, it works.

I wonder if somebody could combine DLSS4 upscaling with FSR3 frame generation for the older NVIDIA cards.

 

[Tomorrow]

Resizable BAR support has absolutely nothing to do with feature support, and to this day the performance advantage of having it enabled is questionable, the only GPUs that need it are Intel's, because the drivers were written from the beginning with that in mind. Nvidia uses app profiles to disable it in many applications, and I'm sure the same happens with AMD

ReBAR adds roughly 5%. In some games less, in some games more. This is free performance Nvidia did not want to give their buyers.
With AMD there is no whitelist for ReBAR. It's is enabled for all games by default, but can be disabled on a game-per-game basis if issues occur.
So far i have not run into such issues yet and i have it enabled for everything.

It requires more processing power, so naturally. It also takes longer to infer than the CNN model on newer hardware, including the 50 series. Even then, it's only 5% slower than the CNN model, and it's doubtful that this will cause any perceptible increase in latency unless you're already running at sub 20 fps IMHO. The quality of the image produced is the same across all supported GPUs.

That's extra performance the older series loses despite supporting the feature.
if AMD ever ports FSR4 in some form to RDNA3 i fully expect that one to perform worse too.
I think i saw a proof-of-concept test where someone got it to run on RDNA3, but the results were not great.

I wonder if somebody could combine DLSS4 upscaling with FSR3 frame generation for the older NVIDIA cards.

This is game dependent. If a game allows combining different upscaler and FG then it's perfectly doable. Like in Ghost of Tsushima where older Nvidia cards can run DLSS4 upscaling along with Reflex, combined with FSR FG.

 

[Redwoodz]

It goes in circles, we had the exact same discussion like 5 days ago, told him vega not on the maintenance line (basically it's fixing bugs that exist on the last driver) but nope, some people just don't give a damn about reality if it align with their brand of preference.

Mind you, Vega is still being sold new as a new product even to this day, something that isn't the case with Turing even though turing is still under full support. Go figure.

"Vega is still being sold new as a new product even to this day"
What? Proof?

 

[JustBenching]

"Vega is still being sold new as a new product even to this day"
What? Proof?

Again? Come on now....Barcelo R has vega graphics. So 7730u, 7530u, 7430u, 7330u etc.

There are many more families of products that are new (2023 or later) but im not going to go through all of them, just google.

 

[Tomorrow]

"Vega is still being sold new as a new product even to this day"
What? Proof?

I think he means iGPU's on some 3+ year old laptops:

Notebooks mit iGPU-Modell (AMD): Radeon RX Vega 10 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics Preisv

Preisvergleich und Bewertungen für Notebooks mit iGPU-Modell (AMD): Radeon RX Vega 10 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics

geizhals.eu

Only one model from HP shows as listed since 2024 and this on HP for making a new model in 2024 with Vega when there are much better RDNA options available. All other models listed are from 2022 or earlier when AMD froze Vega feature development. So these laptops are just stock clearance and anyone buying these cant complain that Vega only receives security updates now. Only 29 are actually in stock and only two have 20+ offers while rest are 1-2 offers.
There's also zero reason to buy any of these are for the same price you can get RDNA.

That's a pretty weak argument that Vega is still being "sold" but that user is grasping at anything just to prove is point that "Vega is abandoned" or something so im not surprised. Just put him in the ignore list like i have done and move on. Not worth arguing with that person.

 

[RandallFlagg]

I think he means iGPU's on some 3+ year old laptops:

Notebooks mit iGPU-Modell (AMD): Radeon RX Vega 10 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics Preisv

Preisvergleich und Bewertungen für Notebooks mit iGPU-Modell (AMD): Radeon RX Vega 10 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics/Radeon Vega 3 Graphics/Radeon Vega 6 Graphics/Radeon Vega 8 Graphics/Radeon Vega 9 Graphics

geizhals.eu

Only one model from HP shows as listed since 2024 and this on HP for making a new model in 2024 with Vega when there are much better RDNA options available. All other models listed are from 2022 or earlier when AMD froze Vega feature development. So these laptops are just stock clearance and anyone buying these cant complain that Vega only receives security updates now. Only 29 are actually in stock and only two have 20+ offers while rest are 1-2 offers.
There's also zero reason to buy any of these are for the same price you can get RDNA.

That's a pretty weak argument that Vega is still being "sold" but that user is grasping at anything just to prove is point that "Vega is abandoned" or something so im not surprised. Just put him in the ignore list like i have done and move on. Not worth arguing with that person.

There are a ton of laptops being sold brand new with those 4 year old rebranded chips.

Now these below are direct from Mfr. sites, the retail channel is far worse than this.

It's obvious that this is the entire reason that AMD went to that cryptic naming convention.

To illustrate how bad that is, notice the Dell Inspiron 7520U. The 2 means Zen 2. That has a 6 year old architecture chip in it.

 

[JustBenching]

There are a ton of laptops being sold brand new with those 4 year old rebranded chips.

Now these below are direct from Mfr. sites, the retail channel is far worse than this.

It's obvious that this is the entire reason that AMD went to that cryptic naming convention.

To illustrate how bad that is, notice the Dell Inspiron 7520U. The 2 means Zen 2. That has a 6 year old architecture chip in it.

View attachment 403606

View attachment 403607

View attachment 403611

Yeah well, but now you are using facts, and this thread is a clear demonstration that some people don't care about them. As if any of this would make a difference, support for these gpus has stopped 3 years ago, even if these products aren't sold NOW - they were still sold as brand new during the time that support has stopped. Grasping at straws to deny the obvious....

 

[Tomorrow]

There are a ton of laptops being sold brand new with those 4 year old rebranded chips.

Define "tons"? I linked one of the biggest Europe's price aggregator sites where only 29 models were in stock. Only two of them had any meaningful number of offers and only one was introduced after 2022.

So the important takeaway here is not how many are available, but when did they first go on sale and are these new laptops or just clearance sales on old ones?
Because based on what i see these are almost exclusively 3+ year old laptops on clearance.

Now these below are direct from Mfr. sites, the retail channel is far worse than this.

Not necessarily.

It's obvious that this is the entire reason that AMD went to that cryptic naming convention.

Probably, but at the time it was more about selling old Zen CPU models as new.

To illustrate how bad that is, notice the Dell Inspiron 7520U. The 2 means Zen 2. That has a 6 year old architecture chip in it.

And what affects the overall performance more? A Zen 2 chip instead of Zen or a Vega iGPU instead of RDNA?
I think we all know the answer to that question. Zen 2 is far more noticeable than Vega.

 

[RandallFlagg]

Probably, but at the time it was more about selling old Zen CPU models as new.

That's exactly what it was, and is, about. 90%+ of people buying those laptops has no idea how deceptive that naming convention is. They're going to see the 7 at the front and think it is just one gen behind the ones with an 8 at the front.

And what affects the overall performance more? A Zen 2 chip instead of Zen or a Vega iGPU instead of RDNA?
I think we all know the answer to that question. Zen 2 is far more noticeable than Vega.

It's all a reprehensible and overtly deceptive practice on AMDs part. They make it look like they have kept a consistent naming convention, when they have not.

 

[Tomorrow]

That's exactly what it was, and is, about. 90%+ of people buying those laptops has no idea how deceptive that naming convention is. They're going to see the 7 at the front and think it is just one gen behind the ones with an 8 at the front.

It's all a reprehensible and overtly deceptive practice on AMDs part. They make it look like they have kept a consistent naming convention, when they have not.

Yeah unfortunately all manufacturers are, or have been guilty of this at some point by deceptively naming their old stuff.
The only thing we as end users can do is not to buy these and educate our circle of people to at least check with someone before buying.

 

[RandallFlagg]

Yeah unfortunately all manufacturers are, or have been guilty of this at some point by deceptively naming their old stuff.
The only thing we as end users can do is not to buy these and educate our circle of people to at least check with someone before buying.

They have, but nowhere near this level and this widespread. AMD literally did this across their entire mobile stack. If those laptops had their old designation of 3XXXU or 4/5XXXU, they'd be $300 laptops not $800 laptops. And that is why they did it.

 

[blinnbanir]

They have, but nowhere near this level and this widespread. AMD literally did this across their entire mobile stack. If those laptops had their old designation of 3XXXU or 4/5XXXU, they'd be $300 laptops not $800 laptops. And that is why they did it.

It is funny that you are blaming AMD for laptop prices from vendors that only buy the chip(s) from AMD.

 

[Dr. Dro]

It is funny that you are blaming AMD for laptop prices from vendors that only buy the chip(s) from AMD.

Laptop vendors buy these processors in bulk at discount prices. There is no defending the indefensible, the product designation is very much intended to mislead buyers who aren't keenly aware of what those numbers mean. Which are next to nobody, even in tech circles. It is utterly irrational to expect the target audience of low cost Dell Inspirons and HP Pavilions to know what any of this means:

 

[RandallFlagg]

It is funny that you are blaming AMD for laptop prices from vendors that only buy the chip(s) from AMD.

That's because, before AMD renamed its chips, a laptop with a three gen old CPU was selling cheap.

If they put the old naming nomenclature chips out there, they'd sell cheap again - because nobody would want them.

One of those 'new' chips is a lot like a 4600U.

As stated, this is completely and utterly indefensible except to the most rabid brand 'loyalists'. You know what you are.

 

[blinnbanir]

That's because, before AMD renamed its chips, a laptop with a three gen old CPU was selling cheap.

If they put the old naming nomenclature chips out there, they'd sell cheap again - because nobody would want them.

One of those 'new' chips is a lot like a 4600U.

As stated, this is completely and utterly indefensible except to the most rabid brand 'loyalists'. You know what you are.

They renamed their chips because a $800 APU is still cheaper than a $6000+ DGPU based laptop. It is all relative

 

[RandallFlagg]

They renamed their chips because a $800 APU is still cheaper than a $6000+ DGPU based laptop. It is all relative

 

[blinnbanir]

View attachment 403656

Indeed look at what Nvidia did when they replaced the 3060 laptops with the 4050 with 4GB for $300 more. Stop supporting the narrative that AMD is somehow more evil than Nvidia

Laptop vendors buy these processors in bulk at discount prices. There is no defending the indefensible, the product designation is very much intended to mislead buyers who aren't keenly aware of what those numbers mean. Which are next to nobody, even in tech circles. It is utterly irrational to expect the target audience of low cost Dell Inspirons and HP Pavilions to know what any of this means:

View attachment 403655

You are an unabashed Nvidia fan boy so nothing you say has any merit to me

 

[Dr. Dro]

Indeed look at what Nvidia did when they replaced the 3060 laptops with the 4050 with 4GB for $300 more. Stop supporting the narrative that AMD is somehow more evil than Nvidia

If you're going to do this, at least get the facts right... the RTX 4050 has always had 6 GB by default, and the 4050 was never intended to replace the 3060, but rather the 3050, which was also refreshed from the original 4 GB design to a newer 6 GB one with more cores but narrower bus.

You are an unabashed Nvidia fan boy so nothing you say has any merit to me

Ad hominem. Furthermore, in my defense, it is a statement which has no basis in reality. The people who question my stances, however, are admitted AMD diehards. All 5 of them, shout out to my fan club. If I got that reputation because I don't seem to necessarily think everything Nvidia does is unquestionably evil, then it's a reputation I pretty much don't mind having.

 

[Rover4444]

The people who question my stances, however, are admitted AMD diehards.

I'm an admitted Intel diehard, actually. I will not tolerate the slander!

shout out to my fan club

 

[JustBenching]

If you're going to do this, at least get the facts right... the RTX 4050 has always had 6 GB by default, and the 4050 was never intended to replace the 3060, but rather the 3050, which was also refreshed from the original 4 GB design to a newer 6 GB one with more cores but narrower bus.



Ad hominem. Furthermore, in my defense, it is a statement which has no basis in reality. The people who question my stances, however, are admitted AMD diehards. All 5 of them, shout out to my fan club. If I got that reputation because I don't seem to necessarily think everything Nvidia does is unquestionably evil, then it's a reputation I pretty much don't mind having.

Both intel and AMD have gone rogue when it comes to laptop naming schemes, but amd has taken it a step further. Or - a lot of steps I guess. Nvidia is well behaved in comparison to these 2, with amd being by far the worst offender.

 

[Dr. Dro]

I'm an admitted Intel diehard, actually. I will not tolerate the slander!

You're a reasonable fellow, lol

Both intel and AMD have gone rogue when it comes to laptop naming schemes, but amd has taken it a step further. Or - a lot of steps I guess. Nvidia is well behaved in comparison to these 2, with amd being by far the worst offender.

At least Intel clearly distinguished their rereleased and new products with the Ultra moniker and seem to be avoiding using the exact same numbers, very weird marketing, but relatively valid IMO. Better than the "13th and 14th Gen" Alder Lake parts, clearly rebadged 12th Gen CPUs. Nvidia's puzzles me a bit too, for example, there's the RTX 2050 which is clearly an Ampere GA107 model, basically the same RTX 3050M but with a very narrow 64 bit bus... and they also sell that exact thing as a GeForce MX 250, so the buyer can at least expect to have generally the same ballpark of performance by buying any of these models.

The last big, big oof NV did with branding imo was the GT 1030 with DDR4 and GDDR5, DDR4 just wasn't good enough for that little thing and the performance difference between both was massive. The 3050 6 GB and 3050 8 GB can also be a tad misleading, but in general the 6 GB models don't have a power connector and are clearly labeled to have 6 GB on the box, so that can often be forgiven depending on the context as they make no effort to conceal that it is an entry level offering to the RTX 3050 series. They are all parts for mass market with insane volume, anyway.

 

[blinnbanir]

It does not matter what the Nvidia fan boys say. They are the only ones that wax negatively on AMD's offerings without any real experience with AMD's current products. Someone who buys a 5090 and commenting negative should be taken with a huge grain of salt. All consoles with the exception of the Switch 2 are AMD. The same thing can be said for handhelds as even MSI will be offering AMD based offerings. While they have been complaining about AMD I have seen a user say that he runs his 5090 at 66% and is patiently waiting for fixes for his drivers. This is the same person that will tell you that my screenshot of AMD software showing 160 FPS is a lie because AMD cannot be that good,. I am not saying that CUDA and AI have not pushed Nvidia into the pole position but that does not mean that AMD's offerings are objectively worse when the are the basis of PC Gaming and Steam OS currently only supports AMD and Intel.

 

[Dr. Dro]

It does not matter what the Nvidia fan boys say. They are the only ones that wax negatively on AMD's offerings without any real experience with AMD's current products. Someone who buys a 5090 and commenting negative should be taken with a huge grain of salt. All consoles with the exception of the Switch 2 are AMD. The same thing can be said for handhelds as even MSI will be offering AMD based offerings. While they have been complaining about AMD I have seen a user say that he runs his 5090 at 66% and is patiently waiting for fixes for his drivers. This is the same person that will tell you that my screenshot of AMD software showing 160 FPS is a lie because AMD cannot be that good,. I am not saying that CUDA and AI have not pushed Nvidia into the pole position but that does not mean that AMD's offerings are objectively worse when the are the basis of PC Gaming and Steam OS currently only supports AMD and Intel.

Not that I have to justify myself, but the only reason I have an RTX 5090 is because AMD doesn't have anything on the level and won't for at least a couple of years. The closest thing to it they offer is but half as fast, and not suitable for my needs. I built a 4K ultra gaming machine. Running at 66%? Do you mean that thread that @JustBenching asked me to try Wukong on it at a similar wattage to his 4090? Bro, lol.

This is clearly the hill you want to die on, spreading misinformation with lies and fallacies... nobody's buying it.