AMD's Radeon RX 9070 Launch Faces Pricing Hurdles

[InVasMani]

I could only see the 32GB being useful if you run LLM's on the GPU.

But AMD would rather you buy a 7900XTX (24GB) on the consumer side or a W7800 (32GB) which means more profit for them.

Flight Simulator apparently can use up to 24GB VRAM at least co-pilot said that's the case and unsurprising given the scope of it. There is also outside of AI 3D modeling and CUDA along with like GPU Ram Drive possibilities. It's very fast high bandwidth storage. There is always a use case that can be made for more of it. You don't have to go that far back before someone said you wouldn't use X capacity VRAM or system memory in general. At the same time developers target up to what available on the market at the time of development or some update to on going development. I mean look at Skyrim and we're not even on version 9000 yet.

 

[Dr. Dro]

I've had a couple of NVIDIA products, the 9500 GT DDR2 (awful, but cheap) and the GTX 970 (3.5GB really, cheap as well).

Other than that, I had to buy Radeon simply because of the pricing over here.

People are generalizing by suggesting that their market is the only one that matters.

Are you implying that there's a huge pricing disparity between AMD and NV in our country? Because this is absolutely not the case.

https://www.pichau.com.br/placa-de-video-xfx-radeon-rx-7900-xtx-mercury-24gb-gddr6-384-bit-rx-79xmairb9

Exact same price I paid on my ROG Strix OC White RTX 4080 18 months ago. I just sold it for 8k. Can you find 7900 XTXs for less? Yes. Not the fancy magnetic fan one from XFX, no, but you can also find cheaper GeForce RTX cards. Or at least could, if there was any stock left...

 

[Makaveli]

Flight Simulator apparently can use up to 24GB VRAM at least co-pilot said that's the case and unsurprising given the scope of it. There is also outside of AI 3D modeling and CUDA along with like GPU Ram Drive possibilities. It's very fast high bandwidth storage. There is always a use case that can be made for more of it. You don't have to go that far back before someone said you wouldn't use X capacity VRAM or system memory in general. At the same time developers target up to what available on the market at the time of development or some update to on going development. I mean look at Skyrim and we're not even on version 9000 yet.

I don't see those as strong uses cases for AMD to release high memory capacity gpu's in the consumer space especially at the mid tier level that will just hurt sales of the workstation gpu's.

Nvidia also does the same thing there is a reason the 5090 only has 32GB of VRAM and not 48GB's like a A6000 model.

For the 3d modeling and CUDA you mention above both AMD and NV want you buying their workstation videocards for that use case.

 

[eidairaman1]

After all this calamity, the 9070 XT must be either the best or the worst GPU in the last decade. In either case, I want one for my collection.

Same here

 

[ModEl4]

I really would have zero problem with RX 9070 series launching March If this was AMD’s original plan.The problem is that it wasn't the original plan since some retailers and reviewers had already received stock and review samples the last 2-3 weeks, so the original plan was to be sold nearly 2 months earlier. Surely the delay must be correlated with what the competition will offer since according to AMD’s statement : “Radeon 9000 series hardware and software are looking great” so no last minute hardware bugs or software optimizations or whatever.
AMD (just like many from the industry) must have been surprised with RTX 5000 pricing, probably they thought $50 more or whatever for RTX 5070/5070Ti MSRP based on the info that they had from the industry so the RX 9070 series original pricing plan was accordingly higher.
And instead of AMD dropping immediately the RX 9070 series pricing accordingly based on the competition’s announced MSRPs and cover their partner’s channel they just delay the launch 2 months in order to see the exact performance of RTX 5070Ti/5070 and if Nvidia MSRPs are going to stick (probably hoping for lower performance for RTX 5000 series or for inflation or both lol) just to save a few bucks per card.
Just looking at the below slide (if we take it at face value), if the correlation is based on price, AMD's original pricing plan for RX 9070 was lower than RX 7900GRE and higher than RX 7800XT so let's say between $529-$479 (lower than RX 7900GRE MSRP and higher than RX 7800XT street price).
If the correlation in the slide is based on performance the conclusion is the same (then RX 9070 would had lower performance than RX 7900GRE but higher than RX 7800XT, so the MSRP should have been max $529 just to match the previous gen performance/price ratio…)
The difference between RX 9070 and RX 9070XT is small for reference type models (even if RX 9070 is 10% lower clocked than RX 9070XT and also has 10% lower memory clock (18Gbps instead of 20Gbps for example) based on the leaked specs (4096 vs 3584 SPs) the difference wouldn't be higher than 17% at 4K and even lower in QHD (RX 9070XT=100 / RX9070=83) so the logical conclusion for the original RX 9070XT plan was to be max +$100 vs RX 9070 so let's say between $629-$579 (or even less ($50-$70) if RX 9070 isn't so lower clocked and especially if we take on account the differences for older generations - RX5700XT/RX5700 $50 difference, RX6800XT/RX6800 $70 difference, RX7800XT/RX7700XT $50 difference etc)

 

[john_]

Huge fail the March release. Intel did it right, rushing to release their cards way before Nvidia's cards.
Something is really wrong with Radeon group and Nvidia gets again a free pass to price and sell their cards as they like.

 

[Zach_01]

Huge fail the March release. Intel did it right, rushing to release their cards way before Nvidia's cards.
Something is really wrong with Radeon group and Nvidia gets again a free pass to price and sell their cards as they like.

Oh yeah... nVidia cards will be inflated by who knows how much because 9070s are launching on March... not because nVidia will bottleneck the supply "to handle scalpers"

NVIDIA GeForce RTX 5090 And RTX 5080 Likely to Be Extremely Hard to Get at Launch

Recently, we covered a rumor arising from PCGH which stated that the RTX 5090, and perhaps even the RTX 5080, are gearing up to face an extremely stock-limited launch in Germany. Now, it appears that customers in North America will likely face very similar circumstances, if the claims made by...

www.techpowerup.com

Intel did it very right... releasing a single low tier GPU thats nowhere to be found, that requires a high-end CPU to play nice. Cudos!

 

[JustBenching]

Latest rumors are that the 479$ for reference cards was right, but it's in reference to the 9070. The XT is @ 599$ - leaks from chiphell.

 

[AusWolf]

Latest rumors are that the 479$ for reference cards was right, but it's in reference to the 9070. The XT is @ 599$ - leaks from chiphell.

Then the XT needs to be as fast as the 7900 XT, if not the XTX, otherwise there's little incentive to buy the new stuff. A GRE level product at GRE price would be meh.

 

[Krit]

Something is really wrong with Radeon group

FSR4

 

[JustBenching]

Then the XT needs to be as fast as the 7900 XT, if not the XTX, otherwise there's little incentive to buy the new stuff. A GRE level product at GRE price would be meh.

For 600$? I see no reason why it shouldn't smack the 5080 in raster. 2 anemic nvidia generations back to back, normally the 5080 is a 5060 / 5060ti tier. A 600$ GPU should be a lot faster than that, at least in pure raster. It probably won't be, cause greed is good, but one can hope.

 

[AusWolf]

For 600$? I see no reason why it shouldn't smack the 5080 in raster. 2 anemic nvidia generations back to back, normally the 5080 is a 5060 / 5060ti tier. A 600$ GPU should be a lot faster than that, at least in pure raster. It probably won't be, cause greed is good, but one can hope.

Unless the guys at AMD put all their cards on RT and FSR 4 this time, but let's hope you're right.

 

[Assimilator]

View attachment 380769
AMD confirms Radeon RX 9070 series launching in March

This actually may not be the worst decision in the world.

Looking at it positively, given how AMD has literally renamed their GPUs to crib their competitor's naming standard, it makes sense to go all-in on this copycat/confusion strategy and launch their similarly-named GPU at the same time NVIDIA launches the 5070, for maximum effect at spoiling the latter's launch; NVIDIA won't be able to prevent this by launching early. Having two months' of stock ready and waiting to ship also means that AMD will be able to overwhelm 5070 stock levels and essentially nullify that launch with hordes of (hopefully cheaper) 9070s. Intel's new GPUs aren't competing in this segment so can't spoil anything (B770 is nowhere to be seen and scuttlebutt suggests it may have already been cancelled).

On the negative side, the hype for 9070 is right now. In two months' time, that hype bubble will have been significantly deflated by the raw performance of 5090 and 5080 (or rather, these cards will have shifted the market's expectation of performance) and the NVIDIA 5000-series hype train will be going full steam ahead. There is always the possibility that NVIDIA decides to lop another $50 off the 5070's price to rain on AMD's parade, given that the former's traditional avenue of combatting competition by launching early isn't an option here. And of course the big one... for the first time in a long time AMD has the opportunity to launch a mid-range product before their competitor, and "first to market wins" is as true as it's always been; a two-month head start is nothing to sneeze at.

I honestly don't know. I can understand the reasoning behind this move, but I can't see a marketing strategy that is entirely predicated on what your competitor decides to do, being sustainable in the long term - because all it takes is one change in behaviour from said competitor, to throw your strategy for an unrecoverable loop. NVIDIA only really got to their current position of market dominance when they stopped reacting to what AMD was doing, and started doing what they believed would make them successful... I honestly can't remember the last time that AMD's GPU division was setting the pace in the graphics industry, as opposed to following the trail that NVIDIA has previously blazed.

Let's just hold thumbs that the stars align for AMD this time around, because if they don't...

 

[DaemonForce]

And instead of AMD dropping immediately the RX 9070 series pricing accordingly based on the competition’s announced MSRPs and cover their partner’s channel they just delay the launch 2 months in order to see the exact performance of RTX 5070Ti/5070 and if Nvidia MSRPs are going to stick (probably hoping for lower performance for RTX 5000 series or for inflation or both lol) just to save a few bucks per card.

The one and only problem I see with this is there's already a very large customer base on very old cards like the 480/580 looking to upgrade to new Radeon. I'm in it. We don't need the performance metrics when the upgrade already looks like ~400% ballpark estimates. If the flagship supposedly replaces the 7900GRE, all the other details are just odds and ends.
Oh 3GHz boost? Cool.
16GB memory? Infinitely wise choice.
AI performer? Great.
Using cheaper GDDR6 memory that keeps the price in check? Awesome.

We really aren't asking for much here. Minimal marketing materials and a solid launch date. Go!

If this card drops anywhere near $600 USD it won't compete anywhere. Everyone (other than AMD) knows this.
I've waited since last January to get things rolling with a working 7900XT and failed. Waited for 9K series and now AMD is using shops for cheap storage rental. Wat?
This nonstop nutfuckery will have me considering team green and I know better. I already know the encoder solves one problem and every other detail creates 5 new ones.
Their cards won't play ball with the rest of my gear and I need the bandwidth. I need Radeon's raster and I'm not going to get it from Jensen's el cheapo AI units at any price.
AMD needs to give us the launch or cancel for UDNA since the bridge from here to there is going to sour a lot of people buying this generation of card.
After this I won't be making the jump to it anyway. This breaks any confidence I had in their ability to run a statement of work.
Sell off the Radeon branch to literally anyone that cares about making GPUs because whatever you want to call this strategy, ain't it.

"first to market wins" is as true as it's always been; a two-month head start is nothing to sneeze at.

Except we already know how this will go. AMD will sit on the advantage until it spoils.
A two month lead becomes a zero month lead becomes a four month lag behind nVidia.

 

[lilhasselhoffer]

Why? Semiconductor fabrication costs have far outpaced the rate of inflation.



Holy crap. Disastrous comes to mind. Another Months Delay.

You are incorrect. The cost of subsequent nodes has increased at a pace much faster than inflation. The cost of the same nodes actually decreases, as the yield per node increases. IE, the same input wafer can now produce 95% good product instead of 85% of good product means input costs are stagnant but output is greater, so cost overall is less. Your statement about silicon costs is about the unusually low yields of the latest processes...which is a failure of new processes and not an increase in all production costs. For a rather simplistic example, note that the PCH in most computers used to be a generation or two behind the CPU to cut platform costs. Now the CPU itself might use multiple generations of lithography.

Why then model inflation at all? I'm putting that question in your mouth, because it's a reasonable ask. Development costs, low initial yields, and all of that fun is amortized. It's built into the cost of the silicon overall, because any good foundry is constantly improving. The old silicon is getting cheaper, but price stays stagnant because the new silicon's development is built into the cost....because foundries do like to stay in business. As such, you model your costs on a regular cadence of cheap oldest processes, improving older processes, and high cost new processes. This is the same thing all manufacturing does...because if they didn't they wouldn't make a profit.


Regarding the announced delay...this sounds like a bad time for AMD. Even the most optimistic interpretation of this is that a two month delay with cards in the wild points to shenanigans behind the scenes. There's no greater failure than having inventory sit unable to be sold at retailers...because that crap is going to get out. This points to deeper issues than a pricing concern...

 

[AusWolf]

This actually may not be the worst decision in the world.

Looking at it positively, given how AMD has literally renamed their GPUs to crib their competitor's naming standard, it makes sense to go all-in on this copycat/confusion strategy and launch their similarly-named GPU at the same time NVIDIA launches the 5070, for maximum effect at spoiling the latter's launch; NVIDIA won't be able to prevent this by launching early. Having two months' of stock ready and waiting to ship also means that AMD will be able to overwhelm 5070 stock levels and essentially nullify that launch with hordes of (hopefully cheaper) 9070s. Intel's new GPUs aren't competing in this segment so can't spoil anything (B770 is nowhere to be seen and scuttlebutt suggests it may have already been cancelled).

On the negative side, the hype for 9070 is right now. In two months' time, that hype bubble will have been significantly deflated by the raw performance of 5090 and 5080 (or rather, these cards will have shifted the market's expectation of performance) and the NVIDIA 5000-series hype train will be going full steam ahead. There is always the possibility that NVIDIA decides to lop another $50 off the 5070's price to rain on AMD's parade, given that the former's traditional avenue of combatting competition by launching early isn't an option here. And of course the big one... for the first time in a long time AMD has the opportunity to launch a mid-range product before their competitor, and "first to market wins" is as true as it's always been; a two-month head start is nothing to sneeze at.

I honestly don't know. I can understand the reasoning behind this move, but I can't see a marketing strategy that is entirely predicated on what your competitor decides to do, being sustainable in the long term - because all it takes is one change in behaviour from said competitor, to throw your strategy for an unrecoverable loop. NVIDIA only really got to their current position of market dominance when they stopped reacting to what AMD was doing, and started doing what they believed would make them successful... I honestly can't remember the last time that AMD's GPU division was setting the pace in the graphics industry, as opposed to following the trail that NVIDIA has previously blazed.

Let's just hold thumbs that the stars align for AMD this time around, because if they don't...

This is the most sensible comment in the thread by far.

The one and only problem I see with this is there's already a very large customer base on very old cards like the 480/580 looking to upgrade to new Radeon. I'm in it. We don't need the performance metrics when the upgrade already looks like ~400% ballpark estimates. If the flagship supposedly replaces the 7900GRE, all the other details are just odds and ends.
Oh 3GHz boost? Cool.
16GB memory? Infinitely wise choice.
AI performer? Great.
Using cheaper GDDR6 memory that keeps the price in check? Awesome.

We really aren't asking for much here. Minimal marketing materials and a solid launch date. Go!

I'm in the same shoes as you, just with a 6750 XT. I used to have a 7800 XT which was more than enough for my needs, only that I was in need for some cash, so I had to sell it. Otherwise, I wouldn't be looking at RDNA 4 right now. I just thought, if I was gonna re-buy the 7800 XT anyway, then why not wait a bit longer to get a 9070 XT instead, since the price was predicted to be similar, and the launch was near. Now, the launch is whenever the launch is, and the price is uncertain. Don't get me wrong, I can wait, it just doesn't make sense, and the perfectionist in me hates it when something doesn't make sense.

 

[Zach_01]

This is the most sensible comment in the thread by far.


I'm in the same shoes as you, just with a 6750 XT. I used to have a 7800 XT which was more than enough for my needs, only that I was in need for some cash, so I had to sell it. Otherwise, I wouldn't be looking at RDNA 4 right now. I just thought, if I was gonna re-buy the 7800 XT anyway, then why not wait a bit longer to get a 9070 XT instead, since the price is predicted to be similar, and the launch is near. Now, the launch is whenever the launch is, and the price is uncertain. Don't get me wrong, I can wait, it just doesn't make sense, and the perfectionist in me hates it when something doesn't make sense.

If you think about it the 5070s/5070Ti are also launching by the end of Feb if not in March.

 

[AusWolf]

If you think about it the 5070s/5070Ti are also launching by the end of Feb if not in March.

Sure, but hype is at its peak now, right after CES, and the stock is there, waiting in stores. This is a risky game for AMD either way.

 

[Assimilator]

This is the most sensible comment in the thread by far.

I'm sure team red lovers will find some aspect of it to use for accusing me of being an NVIDIA shill. Again. I'm honestly considering quitting this forum for good because it's become so difficult to have a rational discussion here - the team red camp are so biased and so reactive to anything critical of their dear company (which BTW doesn't give a flying f**k about them) that having an opinion that differs from theirs, even if it's backed up with a facts-based explanation of how that opinion was reached, results in the internet equivalent of being burned at the stake. And I'd have no problem with that if it was backed up with facts to counter the ones I'm presenting, but it almost never is. When you're interested in having honest discourse in order to potentially have your mind changed by information and viewpoints you've not considered, and all you're getting is the equivalent of chimps throwing their feces, it's incredibly exhausting and disheartening. The few rational people like you are all that are keeping me here.

 

[john_]

Oh yeah... nVidia cards will be inflated by who knows how much because 9070s are launching on March... not because nVidia will bottleneck the supply "to handle scalpers"

When a company doesn't have competition, can price it's products as high as it wants, no matter if scalpers manage to grab most of the supply or supply is artificially restrained to push prices higher than MSRP. AMD's choice here is questionable at best. Maybe they feel their product isn't good enough, or on the contrary it is good but maybe they don't want to sell much because they don't have wafers to cover the supply. So they wait Nvidia to release 5070, so most consumers will be driven to that card. Don't know. Just throwing random thoughts here. Probably illogical thoughts, but I can't connect AMD's choice to postpone the card's release with logic. Maybe FSR4 is not ready, as implied to another reply to me. Maybe they found other problems that they need some time to solve, like the power consumption problem with RDNA3 in video playback, but worst.
When you want to get market share you rush to be first in the market. Intel did that and it's new cards where tested against 2 years old RX 7600 and RTX 4060, that's why they got so favorable reviews. AMD waiting for 5070 and for the new techs Nvidia is promoting with 5000 series to get traction in the market, to start getting implemented in games, this is at least stupid if not on purpose. In NO way it will help AMD to get market share.
Just remembered reading a few days ago some analysts who downgraded AMD to expect AMD to do even worst in gaming this year and now I think I understand what that analyst knew then, that we learned just now.

 

[AusWolf]

I'm sure team red lovers will find some aspect of it to use for accusing me of being an NVIDIA shill. Again. I'm honestly considering quitting this forum for good because it's become so difficult to have a rational discussion here - the team red camp are so biased and so reactive to anything critical of their dear company (which BTW doesn't give a flying f**k about them) that having an opinion that differs from theirs, even if it's backed up with a facts-based explanation of how that opinion was reached, results in the internet equivalent of being burned at the stake. And I'd have no problem with that if it was backed up with facts to counter the ones I'm presenting, but it almost never is. When you're interested in having honest discourse in order to potentially have your mind changed by information and viewpoints you've not considered, and all you're getting is the equivalent of chimps throwing their feces, it's incredibly exhausting and disheartening. The few rational people like you are all that are keeping me here.

Likewise.

Although, you started your comment with some positives, for which, you could equally be branded an AMD shill. Maybe you're a shill of both companies, who knows.

All joking aside, I feel the same as you (exhausted), albeit, on the other side of the spectrum. It definitely has become excruciatingly difficult lately to have a decent conversation about processors and graphics cards, their makers and their features without being labelled a fan of the other side whose opinion should under no circumstances be taken seriously by anyone. If you mention your dislike towards a single thing that Nvidia does or says (regardless of your reasons), boom, you're an AMD fan and your opinion is invalid. I'm not saying that it's any better from the other side. I'm just sad that we as a society have sunk so deep where our love for tech is divided into "camps", where everybody sees the world through red or green glasses. It makes me wonder what I'm really doing here wasting my time replying to individuals who will never care for my reasoning, just use the opportunity to throw some more excrement my way because I've hurt their precious feelings towards their precious brand. Herd mentality never ceases to amaze me.

I've also come to the point when I've deleted my system specs from my profile, and filled it with some random crap, because it immediately made me look like a shill, and every single one of my points in every thread invalid, even though I own lots of stuff from all three major chip makers, so I know what I'm talking about. Most people here give advice based on reviews they've read - I give advice based on stuff that I actually own or have owned at some point. I used to be proud of my new purchases, but now I feel more like hiding, because whichever brand you buy, it basically makes you a shill.

Anyway, hang in there and keep up the good stuff. 'Twas a nice chat.

 

[gffermari]

AMD may be waiting the 5070 reviews for confirming that is not as fast as the 4090 at humane and usable settings.
Right now, after CES, what's left in people's mind is that a 549$ card matched the 1600$ flagship....

 

[ModEl4]

This actually may not be the worst decision in the world.

Looking at it positively, given how AMD has literally renamed their GPUs to crib their competitor's naming standard, it makes sense to go all-in on this copycat/confusion strategy and launch their similarly-named GPU at the same time NVIDIA launches the 5070, for maximum effect at spoiling the latter's launch; NVIDIA won't be able to prevent this by launching early. Having two months' of stock ready and waiting to ship also means that AMD will be able to overwhelm 5070 stock levels and essentially nullify that launch with hordes of (hopefully cheaper) 9070s. Intel's new GPUs aren't competing in this segment so can't spoil anything (B770 is nowhere to be seen and scuttlebutt suggests it may have already been cancelled).

On the negative side, the hype for 9070 is right now. In two months' time, that hype bubble will have been significantly deflated by the raw performance of 5090 and 5080 (or rather, these cards will have shifted the market's expectation of performance) and the NVIDIA 5000-series hype train will be going full steam ahead. There is always the possibility that NVIDIA decides to lop another $50 off the 5070's price to rain on AMD's parade, given that the former's traditional avenue of combatting competition by launching early isn't an option here. And of course the big one... for the first time in a long time AMD has the opportunity to launch a mid-range product before their competitor, and "first to market wins" is as true as it's always been; a two-month head start is nothing to sneeze at.

I honestly don't know. I can understand the reasoning behind this move, but I can't see a marketing strategy that is entirely predicated on what your competitor decides to do, being sustainable in the long term - because all it takes is one change in behaviour from said competitor, to throw your strategy for an unrecoverable loop. NVIDIA only really got to their current position of market dominance when they stopped reacting to what AMD was doing, and started doing what they believed would make them successful... I honestly can't remember the last time that AMD's GPU division was setting the pace in the graphics industry, as opposed to following the trail that NVIDIA has previously blazed.

Let's just hold thumbs that the stars align for AMD this time around, because if they don't...

I agree that's why i said that if this was their plan from the start i wouldn't have a problem, but they keep miscalculating and going back and forth and in every navi generation we had at least one street price drop shortly after announcing MSRPs except maybe in covid era.The only really good launch MSRPs (without counting price drops after announcing original pricing like in RX5700/5700XT case for example) that they had after 2017 was 6800XT/6800/7800XT models, i don't remember anything else that made an impression.
And the marketing really sucks, for example one of the strong features of AMD is that as you start lowering quality settings, AMD is gaining performance vs the competition (usually all the sites are testing at max or mix/ulta-high, but you can see it for example in some tests like igp vs entry discreet, just check for example w1zzard's 8500G review , GTX 1060 is around +30% normally in ultra 1080p vs RX6400, but at 1080p lowest settings in this games selection RX6400 is +5% vs GTX 1060.
They really don't know what are their strong points and how to market them (in the above example, surely for many folks that don't buy $600 plus and keep their VGAs 4-5 years this is a strong incentive for buying AMD and i doubt many are aware of this)

Waited for 9K series and now AMD is using shops for cheap storage rental. Wat?

I suspect that this is causing a big problem for some retailers/etailers that they don't have good cashflow situation.
AMD should do something to help them, it's not cheap storage it's free storage lol

 

[Random_User]

Using cheaper GDDR6 memory that keeps the price in check?

I'm sorry, but from all rumors it seems that AMD tried to price their GDDR6 cards as nVidia's GDDR7 ones. Yes, the amount of VRAM on 5070 is unacceptable. But it's still the new VRAM standard, that comes at steeper price. I can't see how the four GB more of an older, cheaper VRAM can cost as much.