GPU prices — is the worst behind us ?

The difference between a 5090 and everything else is pretty substantial. I'll take the performance it offers over anything else. (Granted, I paid MSRP for my card.)
Right love my 5090 on water massive upgrade over my old 3090.
 
No question the performance is phenomenal but it's also a room heater. If you're making money from using it, it's still worthwhile but I run most of my AI workflows on our enterprise accounts and Bedrock. Too bad I only captured the upside on one card, sold a tuf 5090 and a FE here basically at cost last year and actually returned a Zotac to MC (MSRP one)...I never learn...Halo GPUs are unobtanium forever.
If you are running a 5090 without a mini split or a window unit in the room you have no business owning a 5090.
 
Is that still true in 2026 ?

I think vray has an HIP version now:
https://documentation.chaos.com/space/VMAX/113587140/V-Ray+GPU+Setup
https://www.pugetsystems.com/labs/a...g-performance-is-radeon-a-viable-alternative/

Redshift has well, they are still so far behind that the support must be limited work too.
This is the field i work in as a hobby other then gaming and yes it is 100% true. I should say that it will run but it is 50% slower in speed so in effect yes you still can't use them. The speed is is so massive even if you can run them no one will. I mean when you are doing Fluids or really anything and a frame takes 5-7min on a 5090 who in the fuck is going to run anything amd when the hit is 50% or MORE slower. It all comes down to render times and like another has said amd is not a software company and that sucks for the rest of us. If they could get Render times with in say a 10% difference people would start to use them but man 50% slower per frame. You can't use that.
 
This is a ridiculous statement. I'm not doing either of these things and its fine.

Really depends where you live. You live in Texas where the A/C is almost always running. Some parts of the country the A/C is not really needed or does not need to run all the time in the summer. So you would be far more likely to notice a 5090 heat output in a room that does not have A/C blowing in it or a running unit in the room.
 
This is a ridiculous statement. I'm not doing either of these things and its fine.
I live in Southeast Louisiana even with the AC running and door open the 1000w heat dump form the PC+ tv + all the networking shit and online ups will easily get my room to 75-80. that's to hot for my liking adding a mini split and keeping the room at 60-65 is optimal to me.

Also I was replying to the guy that sold his because it put to much heat out. If you can afford a 5090 you can afford to keep it cool.
 
I live in Southeast Louisiana even with the AC running and door open the 1000w heat dump form the PC+ tv + all the networking shit and online ups will easily get my room to 75-80. that's to hot for my liking adding a mini split and keeping the room at 60-65 is optimal to me.

Also I was replying to the guy that sold his because it put to much heat out. If you can afford a 5090 you can afford to keep it cool.
Lol it's about priorities, when I scored a below MSRP 5080 from woot, I knew what I had to do....I'm fine turning down a setting or two from ludcirous to ultra or high if needed. Not the end of the world.
 
What's also gone largely ignored due to the AI boom is that we are effectively on the same process node (5nm) and arch since 2022 (the 4090 to 5090 was mostly a die size increase with minor arch revisions), in a more sane/competitive world we'd have had 5090 levels of performance in a 300w envelope with 32 GB VRAM for around $1500. If nothing else, NV should do a 2 or 3nm node shrink for their current GPUs.
 
3nm node shrink for their current GPUs.
The GB10/GB20 on spark upcoming N1X could show a little bit what it could look like on a more recent 3N FinFET node.
https://en.eeworld.com.cn/news/manufacture/eic718138.html

Maybe it is a low effort versus an actual new generation and knew it was for expensive product in small chassis and aimed at optimising power usage/cooling vs performance per mm. So maybe not the perfect comp, but once the N1X with good gaiming drivers is out, it could give some idea, it seem to run 5070 type of core counts at 2.5 ghz at only 80-100watt, fit them in 173mm versus the 263mm 5070 (but the soc could have some memory stuff or others on the cpu tile instead of the gpu tile to help this happen....)
 
Yeah, when less than one out of ten buy AMD cards it's close enough to say nobody wants them. They are fucked.
No, you are wrong. First, as 1_rick pointed out that some people do want AMD cards so your original assertion is wrong. Now, you can say that practically nobody wants Intel discreet GPUs as they are hovering around 1% sales versus a wobbly 7-10% sales for AMD depending on the quarter. With Intel, your statement is valid, but towards AMD, you are overstating.
 
The consumer desktop graphics market means so little at this point in time unfortunately. Lisa Su is over in korea and taiwan making memory deals for AI Accelerator chipsets, not desktop. Nvidia, AMD and even Intel are all profiting. There are no losers in that list. Although intel is at the bottom.
 
You missed or i guess it would fall under cuda that AMD has ZERO way of doing any 3d rendering with Octane/Redshift/Arnold/Karma/Renderman or Vray. Like NONE of the major players support AMD. NONE OF THEM. The only thing a amd card is good for is games. It literally can't do anything else. The world runs on Cuda.
That would be the OptiX > OpenCL/HIP and of lot of software skip the OpenCL/HIP implementation due to low marketshare by AMD....cost/benefit rules in business
 
No, you are wrong. First, as 1_rick pointed out that some people do want AMD cards so your original assertion is wrong. Now, you can say that practically nobody wants Intel discreet GPUs as they are hovering around 1% sales versus a wobbly 7-10% sales for AMD depending on the quarter. With Intel, your statement is valid, but towards AMD, you are overstating.
The 8% are overstating, as they are basicly the left over die hard fans.
Have 8% of sales for 1½ year is adisaster.
If you do not care about performance/features/image quality, you should buy a console.
If you do care about performance/features/image quality you buy NVIDIA.

8% sales is fucked.
Intel is trying to get a foot hold, but lie AMD they have left the halo segment to NVIDIA and they are also noton par with NVIDIA's features.

Again, the reading forums you would think AMD is doing great and on par with NVIDIA, but th reality is not this.
 
I'll solve the thread for y'all for atleast the next year. The answer is unequivocally and resoundingly NO.


As for the typical arguments that always break out regarding brands: I have an 9070 XT, it's a great card for what it is. Would I rather game on a 5080 or 5090? Yeah obviously. Did I want to pay more than 700 after tax? No. It's just a consumer choice at the end of the day. Both companies are leaning into AI and don't care about consumer graphics that much anymore anyway. With how quickly emulation and native ports are coming for PS 5 games, the inability to sell consumer hardware at a price the majority of consumer will accept, and the continued increases in cost for memory. I don't see a future right now where game streaming doesn't become the status quo in a hurry unfortunately. I think we're essentially a dying market that won't be catered to in a little over a decade. Big changes are coming, and they're coming quickly.
 
No, you are wrong. First, as 1_rick pointed out that some people do want AMD cards so your original assertion is wrong. Now, you can say that practically nobody wants Intel discreet GPUs as they are hovering around 1% sales versus a wobbly 7-10% sales for AMD depending on the quarter. With Intel, your statement is valid, but towards AMD, you are overstating.
Yeah, some want AMD cards, less than 10% of the market wants them. From a business standpoint, that means no one wants your shit. No business can survive that kind of market share unless it is ultra-high margin/volume.

With that said, I take no joy in this reality. I buy halo and NV provides that product. Unfortunately, AMD barely sells non-halo cards and doesn't offer halo cards.

I want cheaper cards. NV dominating the market isn't going to give me that option.
 
As a gamer who doesn't do anything with CUDA, I have to say, "yeah? So?"
It is nor just that.
You get less performance topped off with lesser visual quality.
So lower FPS with lower visual quality, not something a serious gamer would call "desireable":

View: https://www.youtube.com/watch?v=SlRpJ553RzE
Again, if you do not care about visual quality a console would be a better financial choice.

Just because you are one of the remaning 8% die hard fans, doesn't mean AMD is in a great position where at increase to 10% of sales would be considered a success.

You are the minority BIGTIME and biased on top of that making you blind.
 
The 8% are overstating, as they are basicly the left over die hard fans.
Have 8% of sales for 1½ year is adisaster.
If you do not care about performance/features/image quality, you should buy a console.
If you do care about performance/features/image quality you buy NVIDIA.

8% sales is fucked.
Intel is trying to get a foot hold, but lie AMD they have left the halo segment to NVIDIA and they are also noton par with NVIDIA's features.

Again, the reading forums you would think AMD is doing great and on par with NVIDIA, but th reality is not this.
You like making stuff up to fit your perceived narrative? The 8-10% aren't merely die hard fans. I've owned both Nvidia and AMD throughout the years. I'm also a bang for my buck PC DIYer. Given the current situation, AMD made more financial sense. I'm quite happy with my 9070 and 9070 XT.
 
You like making stuff up to fit your perceived narrative? The 8-10% aren't merely die hard fans. I've owned both Nvidia and AMD throughout the years. I'm also a bang for my buck PC DIYer. Given the current situation, AMD made more financial sense. I'm quite happy with my 9070 and 9070 XT.
There is still a place for choosing AMD as a rasterization-per-dollar high detail(ray tracing at some titles) within older titles. AMD is superior in the midrange market for that purpose alone as far as I understand looking at the data.
I just wish that they were not a niche offering and made a successor to the 7900XTX. A dual 9070 XT is possible. We need AMD to succeed for the market to move forward positively for everyone value wise. I just don't see good things ahead for midrange budget gamers. Prices may only go up from here. :(
 
You like making stuff up to fit your perceived narrative? The 8-10% aren't merely die hard fans. I've owned both Nvidia and AMD throughout the years. I'm also a bang for my buck PC DIYer. Given the current situation, AMD made more financial sense. I'm quite happy with my 9070 and 9070 XT.
They wish it was 10%
It is not.
And havn't been for 1½ years.

Take a step back and listen to your self, you are trying to inflate sales by 2% of the market to have a "win".

It is that bad.

EDIT: Those 2% from 8% to 10% is actually inflating sales by 25% FYI)
 
Last edited:
There is still a place for choosing AMD as a rasterization-per-dollar high detail(ray tracing at some titles) within older titles. AMD is superior in the midrange market for that purpose alone as far as I understand looking at the data.
I just wish that they were not a niche offering and made a successor to the 7900XTX. A dual 9070 XT is possible. We need AMD to succeed for the market to move forward positively for everyone value wise. I just don't see good things ahead for midrange budget gamers. Prices may only go up from here. :(
Getting harder and harder to recoup R&D when your sales only are 8%, AMD is not a charity, they have stockholders too.
And it is even worse in the datacenter/A.I. market where NVIDIA is +90%.
 
Before the most recent price increase they really were more attractive value wise. That recent price increase was nails in the coffin. It is really sad to see.
 
Just stop. I don't know what Lisa Su did to hurt you so much, but this is ridiculous.
Was anything I posted factual wrong?
No.
Lower FPS with lower visual quality.
Like it or not.
(And that is whitout all the added bonus features.)

There is a reason they have droped to 8% of sales, ignoring facts doesn't alter reality, no matter how angry you get (your bias/ego showing there)).
 
Yeah, some want AMD cards, less than 10% of the market wants them. From a business standpoint, that means no one wants your shit. No business can survive that kind of market share unless it is ultra-high margin/volume.

With that said, I take no joy in this reality. I buy halo and NV provides that product. Unfortunately, AMD barely sells non-halo cards and doesn't offer halo cards.

I want cheaper cards. NV dominating the market isn't going to give me that option.

littlegreenmenfromhome
The 8% are overstating, as they are basicly the left over die hard fans.
Have 8% of sales for 1½ year is adisaster.​


Again, both of you are wrong. If AMD was only Radeon GPUs and no other products, then I would agree with you, but they have a robust CPU side that is kicking Intel's butt and eating away at their dominance every quarter in Enterprise spaces and they are competing with Nvidia in the AI space, probably better than they are in the consumer discrete GPU space if comparing. They are also selling tens of millions of custom chips for Playstation, Xbox, handhelds, Steam hardware, etc. 10% in discrete GPU market share is not that bad when they are selling very well in other areas because they are a diversified company, not a one trick pony. Both Nvidia and ATI/AMD have expanded beyond being one trick pony discrete GPU designers because there isn't that much money in discrete GPUs, not compared to the deep pockets of big tech companies all fighting each other for chips.
 
A RTX 5080 Super 24GB for $2k launch would sell fine given current market conditions. I think Nvidia is too preoccupied with making ludicrous money in the datacenter market to do bother with it.
The space between RTX 5080 and RTX 5090 is widening in street price.
 
Again, both of you are wrong. If AMD was only Radeon GPUs and no other products, then I would agree with you, but they have a robust CPU side that is kicking Intel's butt and eating away at their dominance every quarter in Enterprise spaces and they are competing with Nvidia in the AI space, probably better than they are in the consumer discrete GPU space if comparing. They are also selling tens of millions of custom chips for Playstation, Xbox, handhelds, Steam hardware, etc. 10% in discrete GPU market share is not that bad when they are selling very well in other areas because they are a diversified company, not a one trick pony. Both Nvidia and ATI/AMD have expanded beyond being one trick pony discrete GPU designers because there isn't that much money in discrete GPUs, not compared to the deep pockets of big tech companies all fighting each other for chips.
A good CPU does not negate failing the GPU market for 1½ years.

Consoles are low revenue.
NVIDIA dominates everything else (the high margin markets).
 
A RTX 5080 Super 24GB for $2k launch would sell fine given current market conditions. I think Nvidia is too preoccupied with making ludicrous money in the datacenter market to do bother with it.
The space between RTX 5080 and RTX 5090 is widening in street price.
It takes 6 months to produce a RTX 5080 die from start to finish.
The order has to be put in at least 1 year before producion start.

That is a minimum lead time of 1½ years.
You cannot just flick a switch and poor out dies the next day,

You view is very simplistic.
 
A good CPU does not negate failing the GPU market for 1½ years.

Consoles are low revenue.
NVIDIA dominates everything else (the high margin markets).
Dr Lisa Su has been chasing the higher margin markets, discrete GPUs are not high margin, all the bitching you guys do about GPU prices is proof enough. I never said that a strong CPU side negates the GPU side. The GPU side has not had its Ryzen moment yet and some at AMD have talked about it, just has not materialized yet. Maybe it never will, who knows. I think that having AMD GPUs in the major consoles will help drive game devs to better support AMD uArch and may result in better desktop GPU performance if the uArch is not vastly different. Now that Sony and MS are driving harder towards ray tracing capability, AMD has their work cut out for them for next gen consoles to be what Sony and MS want.

Consoles may be lower revenue, but it is volume revenue, something investors like to see.

Nvidia does not dominate Enterprise Server markets, AMD is getting closer to being the dominate one in that area in a few years when they exceed Intel's market share. Nvidia is not really competing in that space. Now with Agentic AI, Nvidia does see that as a revenue stream and is making hardware for that aspect of AI which AMD has already been working on, so they are behind the curve in that regard for now.
 
It takes 6 months to produce a RTX 5080 die from start to finish.
The order has to be put in at least 1 year before producion start.

That is a minimum lead time of 1½ years.
You cannot just flick a switch and poor out dies the next day,

You view is very simplistic.
Nvidia already has Super dies per reputable leakers, it is the cost of 3 GB memory chips that makes discrete GPU prices closer to Enterprise GPU prices.
 
It takes 6 months to produce a RTX 5080 die from start to finish.
The order has to be put in at least 1 year before producion start.

That is a minimum lead time of 1½ years.
You cannot just flick a switch and poor out dies the next day,

You view is very simplistic.
Now you are just being condescending. It also takes time for board manufacturers to decide what board designs, cooling solutions, marketing box designs, access to memory since Nvidia has been offering GPU dies without memory bundled, and distribution channels for international sales. I could keep going on with additional details just so I could condescend with a comment like "You(r) view is very simplistic."

Don't be that guy. If you are that guy, choose to be better.
 
Dr Lisa Su has been chasing the higher margin markets, discrete GPUs are not high margin, all the bitching you guys do about GPU prices is proof enough. I never said that a strong CPU side negates the GPU side. The GPU side has not had its Ryzen moment yet and some at AMD have talked about it, just has not materialized yet. Maybe it never will, who knows. I think that having AMD GPUs in the major consoles will help drive game devs to better support AMD uArch and may result in better desktop GPU performance if the uArch is not vastly different. Now that Sony and MS are driving harder towards ray tracing capability, AMD has their work cut out for them for next gen consoles to be what Sony and MS want.

Consoles may be lower revenue, but it is volume revenue, something investors like to see.

Nvidia does not dominate Enterprise Server markets, AMD is getting closer to being the dominate one in that area in a few years when they exceed Intel's market share. Nvidia is not really competing in that space. Now with Agentic AI, Nvidia does see that as a revenue stream and is making hardware for that aspect of AI which AMD has already been working on, so they are behind the curve in that regard for now.
NIVIDA has a even bigger total marketshare on the AI market:
https://carboncredits.com/nvidia-co...n-2025-and-reveals-next-gen-ai-supercomputer/

AMD is a 4%:
1791602560695.png


Wishful thinking doesn't change reality.

We have only NVDIA GPU's in our datacenters and 0 AMD GPU's and that is not chagning as long I am have the power to decide.
Partly because the software is supporting CUDA, not AMD and partly because AMD is a bad performance/$ and partly because AMD lack features in the same way they do on the desktop.

Again the data doesn't agree with the far to vocal minority on forums.
 
Last edited:
Now you are just being condescending. It also takes time for board manufacturers to decide what board designs, cooling solutions, marketing box designs, access to memory since Nvidia has been offering GPU dies without memory bundled, and distribution channels for international sales. I could keep going on with additional details just so I could condescend with a comment like "You(r) view is very simplistic."

Don't be that guy. If you are that guy, choose to be better.
If you want NVIDIA to launch a lot of 5080's at YOUR price, you are naive, I will repreat it again:

Your views are for to simplistic.

And if you want them to launch a lot more 5080's, you have to tell the shareholders you are diverting dies away from Nvidia RTX PRO 4500 Blackwell's in order to sell a product with less profit magin.

Bad business deal to sell a $2000 5080 compared to a $5500 Nvidia RTX PRO 4500 Blackwell.
 
Now with Agentic AI, Nvidia does see that as a revenue stream and is making hardware for that aspect of AI which AMD has already been working on, so they are behind the curve in that regard for now.
Agentic AI has been most of openAI/Anthropic revenues, Nvidia can power GPT 6 astra fast-mode agentic ai, it is not really "new" in AI term and Nvidia is the clear leader in this has well. Agentic loops have been the big token consumer for a while.

Nvidia networking and ability to make coherent memory-work over hundreds of gpu make their dominance in agentic work even bigger than the previous AI models, giant multi agentic work and reasoning loop having made inference a bit more like training in that regard..
 
I really wish Intel didn't seemingly throw in the towel on their discrete GPUs, would have been nice to have some more competition.
 
If you want NVIDIA to launch a lot of 5080's at YOUR price, you are naive, I will repreat it again:

Your views are for to simplistic.

And if you want them to launch a lot more 5080's, you have to tell the shareholders you are diverting dies away from Nvidia RTX PRO 4500 Blackwell's in order to sell a product with less profit magin.

Bad business deal to sell a $2000 5080 compared to a $5500 Nvidia RTX PRO 4500 Blackwell.
RTX Super Dies were already made and sent to board partners. We just don't know if the board partners have access to the density of VRAM they need yet. You keep saying this condescending "Your views are for(far) to(too) simplistic."

Your perspective was missing critical data and reasoning as to why. It is VRAM density availability and the RTX 5080 Super is not using 2GB modules like those used on the Nvidia RTX PRO 4500 Blackwell. The RTX 5080 Super was set to use 3GB modules which is being used on other products. If it was 2GB modules in a clamshell configuration it would have been released already. Not all cards use the same modules. "Your views are for(far) to(too) simplistic." Apparently it applies to you by your own logic.
 
RTX Super Dies were already made and sent to board partners. We just don't know if the board partners have access to the density of VRAM they need yet. You keep saying this condescending "Your views are for(far) to(too) simplistic."

Your perspective was missing critical data and reasoning as to why. It is VRAM density availability and the RTX 5080 Super is not using 2GB modules like those used on the Nvidia RTX PRO 4500 Blackwell. The RTX 5080 Super was set to use 3GB modules which is being used on other products. If it was 2GB modules in a clamshell configuration it would have been released already. Not all cards use the same modules. "Your views are for(far) to(too) simplistic." Apparently it applies to you by your own logic.
So are NVIDIA releasing a lot of consumer GB203-400-A1/GB203-450-A! dies nor not?

And just because a FEW engineering samples was shipped is not the same a luanch level dies.

So who are "simplistic"?
Me, saying this will not happen anytime soon and a handfull of sample dies does not constitute a luanch preperation.
Or you sayingall the dies are ready, but something is blocking the launch.

If I am so simplistic, tell me how many dies have been shpped to partner, beacuse once again (a common pattern in the this) a lot of personal views are being posted, despite the facts being something else.
 
Back
Top