Why wouldn't it work? Do you feel "gamers" aren't going to pay for it? How much do you feel is the breakover point for too much? And it's a dumb, leading question, but what happens if you're wrong and gamers do pay for it?
Speculation and Rumors: Nvidia Blackwell ...
- Thread starter pharma
- Start date
DegustatoR
Legend
I'd say that +30% ($2000) is "much higher" and yet it should work fine - especially as it launches alongside a 5080 which will supposedly be considerably cheaper while still running all this stuff just fine.
Completely agree.
All depends on price. If it’s $3k I think far few gamers bother with it. The launch 4080 proves that even for NVIDIA the GPU market isn’t entirely inelastic.
The original post I replied to suggested 2500-3000. I actually think 2500 is fine (I would pay it). However there is a limit and it’s anybody’s guess what that is. $3000 is getting there. $4-5000 would be a problem.
If I’m wrong and a $4000 5090 sells enough to show up on the steam hardware survey then clearly there’s no problem and Nvidia can just keep raking it in.
Last edited:
I think interest will begin to fall off dramatically the further over $2k it goes. I expect it to be roughly 2x the price of the 5080, so $2-2.4k.
Put another way, I expect demand at $2k to be nearly as high as the 4090 (tempered a bit just because there is a segment of 4090 owners who are content to skip a generation). Beyond that price is where demand will start to slacken.
DegustatoR
Legend
A $4000 card which will be some 50% faster than a (supposedly) $1200 one makes zero sense though.
The main problem with the 4090 is that it was too cheap. Or at least the MSRP was too low. Even now it's impossible to find one at or close to MSRP. Cheapest ones on Newegg are $1900-$2000. NVIDIA would be crazy to price the 5090 below ~$2200. I'd do $2500-$3000 if I were wearing the leather jacket.
A lot of models sold/available will be marked up AIB skus - the average MSRP is probably at least $200 above the FE base.
I’d honestly be shocked at a base MSRP over $2.5k - and I’ll come here to eat my words if I’m wrong!
Yeah the average price of a 4090 in the US is closer to $1800-$2000 after tax. There was decent availability for anyone willing to pay that much for most of 4090's lifetime. A base MSRP of $2500 is easily $3000+ after AIB markup and taxes.
When I look at Newegg there is 1 model at $1899.99 before tax. Only goes up from there. So they are all easily over $2000 after tax.
The 5090 with its 512-bit bus and 32GB of RAM is potentially not gonna be some pure 'gaming' GPU, though. I think it's quite likely it'll be snatched up massively by smaller scale AI users and businesses, much like the 4090 was. Plus Nvidia is continually testing the extremes of what gamers are willing to pay. And more often than not, are able to normalize quite a bit higher prices in the process. Like, even if they raise prices by like 'x' percent and it doesn't work out, they can just do 'x-10%' or something and people will cheer for it like it's a bargain, even though it's still a big price hike from before. Consumers are stupid and Nvidia knows it.
I really dont think a $3000 pricetag is off the table here. And I'd be downright surprised if it's less than $2500, unless it's quite cut down from what's being reported now.
Microcenter has 5 models less than $2000 of which 3 are less than $1864. I think pricing on the 5090 won't be too much different then what we see for the 4090. As usually the Founders Edition will set the floor.
Frenetic Pony
Veteran
Guess:
5080, 4080 +27%, $1,200 (still < 4090 @4k, for reference)
5090, 24gb, 520w, 4090 + 33%, $1,800
5090 AI, 32gb+, 600w, 4090 +48%, $2,500+
5090 AI will just be for existing lower power datacenters to run AI inference workloads, but it'll "technically" be available for consumers to purchase. If you can even fit a ginormous quad slot into your case without blowing up your power supply off a 600w GPU assuming you can even cool it, after you already spent an extra $2.5k on a 9% overclock.
Maybe the 24gb 5080ti will come out for$1200 24gb after the 5080 drops to $1k sometime in early 2026?
Anyway they'll claim their new stochastic anisotropic filtering units are the shit, despite this costing like < 1ms to replicate on compute shaders. I shouldn't have been that mean about the in ram neural net compression; but trying to claim you came up with something cool when what you're really doing is trying to force gamedevs to spend $$ and time on your platform exclusive thing so your corporate overlords $10 in ram per unit instead of just upping minimum to 12gb+ of ram is the time to ask "are we the baddies?" Nvidia is trying to push towards killing PC gaming inadvertently because this ups profit margin, and that's just depressing (not that AMD's "just below Nvidia" pricing strategy is helping).
5080, 4080 +27%, $1,200 (still < 4090 @4k, for reference)
5090, 24gb, 520w, 4090 + 33%, $1,800
5090 AI, 32gb+, 600w, 4090 +48%, $2,500+
5090 AI will just be for existing lower power datacenters to run AI inference workloads, but it'll "technically" be available for consumers to purchase. If you can even fit a ginormous quad slot into your case without blowing up your power supply off a 600w GPU assuming you can even cool it, after you already spent an extra $2.5k on a 9% overclock.
Maybe the 24gb 5080ti will come out for$1200 24gb after the 5080 drops to $1k sometime in early 2026?
Anyway they'll claim their new stochastic anisotropic filtering units are the shit, despite this costing like < 1ms to replicate on compute shaders. I shouldn't have been that mean about the in ram neural net compression; but trying to claim you came up with something cool when what you're really doing is trying to force gamedevs to spend $$ and time on your platform exclusive thing so your corporate overlords $10 in ram per unit instead of just upping minimum to 12gb+ of ram is the time to ask "are we the baddies?" Nvidia is trying to push towards killing PC gaming inadvertently because this ups profit margin, and that's just depressing (not that AMD's "just below Nvidia" pricing strategy is helping).
Last edited:
No doubt the 4090 had big appeal to those users, but I would wager a few Nvidia shares worth that the majority of 4090s were bought by people for gaming. If Nvidia prices most of them out (which they would at $3k) they aren’t selling nearly as many. That might be fine with them - I don’t know what their production plans for GB202 are. But the halo cards typically have the best margins, so they have an interest in moving a good number of them. As a side note, I completely reject the oft-repeated sentiment (not here, elsewhere) that anyone buying a $1600 GPU has enough money to not care if it’s $3000 or higher. If you imagine someone saying that about similarly priced vacations, the absurdity becomes apparent.
In short, I don’t believe AI/business will pick up the volume lost from gamers at $3k. But if NVIDIA only have the headroom to produce a third as many, maybe that won’t matter. But I suspect they don’t want 5090 volume to be dramatically lower than 4090.
Last edited:
arandomguy
Veteran
I don't think that is the right way to look at it as it depends on why that person is a buying a $1600 GPU or really the current flagship.
If they're buying it purely because they want to buy the flagship and best as a luxury product without respect to cost then it wouldn't matter. Those buyers will buy anyways if the GPU is now $3000. The thing it keep in mind here is that $3000 for something like (which would cost less in real terms if you also factor in resale value) is actually a very achievable spending amount for quite a large amount of the population in the context of something that you'll likely use for years.
If they're buying for more of a functional productive usage stand point then it's more of a objective rate of return calculation, in which $3000 could make sense or not over a $1600 product.
If they're buying it because there spending threshold is somewhat around that then yes they wouldn't be able to buy a $3000 or higher GPU, but doesn't mean there wouldn't exist options for them in the $1600 price bracket.
And the above seems to be a repeated argument from some people which I disagree with. There seems be some people who are stuck on the idea of things like if you buy a 4090 you're forever locked into looking at xx90 GPUs (what did they buy before the 3090 then?) and this is somehow representative. Even the idea that if you spent say $1000 on your last GPU you're somehow now locked into spending around $1000 is not going to be the norm.
I know personally speaking I've bought the entire range of the spectrum, the lowest end to the highest end and everything in between. I don't think I've ever bought in the exact same arbitrary product or price segment from one upgrade to the next much less pidgeonholed myself into thinking I have to do so. I simply bought based on what I felt was the best deal relative to my needs and spending desire at the time. As opposed to because I bought the original Radeon and now I can't buy GPUs anymore since nothing uses that naming and tier scheme. Okay thats a bit ancient of an example but what if someone came into the hobby with the 7870 or 9800GT, according some viewpoints those people had no forward upgrade paths. Are GTX 1660ti buyers supposed to be waiting for GTX 1960ti or something?
Last edited:
arandomguy
Veteran
I won't repeat some of other responses so I'll just try to add elsewhere.
Do they only showcase on the 4090 though? Their DLSS3 and frame generation marketing page for example uses the 4060ti as the show case - https://www.nvidia.com/en-us/geforce/technologies/dlss/
Their Reflex page compares both a RTX 3060 and RTX 4080 - https://www.nvidia.com/en-us/geforce/technologies/reflex/
It's been mentioned but their technology featureset is broadly comptabilite across the product range typically. There are some things, I'm guessing you're referring to RT (and specifically PT), that due to performance limitations may not scale across the entire product range. However in terms of RT I think we've had this debate every since Turing debated in terms of the relevance for the entire product stack, without relitigating it in detail I think we need to agree it's far from settled that you have to have the highest end GPU to leverage it. Cyberpunk PT I believe is generally considered usable with a RTX 4070 depending on subjective preferences, so it's not something that is definitively RTX 4090 or not usable.
Yes I’m referring to graphical enhancements that require more performance not less. If Nvidia launches Blackwell with the 5080 as the premier gaming SKU while putting the 4090 to shame I will be very impressed. Doubt it though.
Okay, in what way? How would a $2000+ 5090 be detrimental to nVidia? I can see a spell of internet ranting about nVidia's ridiculous prices, but will they actually lose market share, or be in any way worse off pushing higher up the price/demand curve than before?
Depends on the price. There is a price where people will still buy it in droves and there’s a price where they won’t. My first and only point was that Nvidia will price the 5090 to sell to gamers. If they don’t then they can’t credibly use the 5090 to market gaming performance of Blackwell. It seems like a very obvious thing to me.
But why? That's the bit I'm not getting. They can showcase games on the 5090, that no-one can afford. What's the downside to that for nVidia? Will they lose sales? It's a little analogous to car adverts saying, "available from only £15,000," and then in the small print saying the model on show isn't the cheapest model but the model with all the expensive extras. If the card is available to buy from somewhere, even an exclusive nVidia dealership, it's not false advertising so not illegal. The negative would be how the market would react, and I can't see any bad reaction there.
In the event of a $2000 5090, I can see a lot of Reddit and X posts saying, "this is disgusting," and then record profits for nVidia for selling their silicon at the highest margins they can command.