I called this a year ago - GPU sales are crashing.

[Tagrineth]

Even before the Crypto crash started, as soon as the Ethereum Proof of Stake merge was announced, I called this.
I knew ADA and RDNA3 were being engineered for a crypto mining-dominated market. I remember what happened to Turing when Nvidia priced themselves out of the market.
I knew if crypto mining was going to fizzle out, that the market would no longer support $1000+ GPUs.
I've been around long enough to see this sort of thing happen over and over again. People will pay any price for the halo product. The 4090 and RX 7900 XTX selling through was a given, a foregone conclusion - but 4080s and 7900 XTs sitting on store shelves gathering dust so soon after release? Even I didn't expect that, I figured the first wave at least would sell through on day 1 like is typically the case.
People don't want these ridiculously overpriced GPUs in the middle of the product stack.
We're seeing another Turing in progress right now. People didn't want a $1000 2080 and they don't want a $1200 4080.
I doubt we'll see a market correction during ADA and RDNA3, sadly. The chips are over-engineered as hell and simply aren't viable for the normal market. But market pressure is happening. Nvidia and AMD cannot ignore this downward trend. Something will have to change.
Personally I'm expecting Blackwell and RDNA4 might come out sooner than expected, with significantly more aggressive pricing. We might see this generation of GPUs end up being very limited in the market space and have a much shortened lifespan. Maybe if we're lucky we might see ADA Supers like we did with Turing, with much better specs for roughly the same prices, in order to try to counteract their own hubris with these stupid products.
We can only be so lucky.

 

[homerdog]

View attachment 7998Even before the Crypto crash started, as soon as the Ethereum Proof of Stake merge was announced, I called this.
I knew ADA and RDNA3 were being engineered for a crypto mining-dominated market. I remember what happened to Turing when Nvidia priced themselves out of the market.
I knew if crypto mining was going to fizzle out, that the market would no longer support $1000+ GPUs.
I've been around long enough to see this sort of thing happen over and over again. People will pay any price for the halo product. The 4090 and RX 7900 XTX selling through was a given, a foregone conclusion - but 4080s and 7900 XTs sitting on store shelves gathering dust so soon after release? Even I didn't expect that, I figured the first wave at least would sell through on day 1 like is typically the case.
People don't want these ridiculously overpriced GPUs in the middle of the product stack.
We're seeing another Turing in progress right now. People didn't want a $1000 2080 and they don't want a $1200 4080.
I doubt we'll see a market correction during ADA and RDNA3, sadly. The chips are over-engineered as hell and simply aren't viable for the normal market. But market pressure is happening. Nvidia and AMD cannot ignore this downward trend. Something will have to change.
Personally I'm expecting Blackwell and RDNA4 might come out sooner than expected, with significantly more aggressive pricing. We might see this generation of GPUs end up being very limited in the market space and have a much shortened lifespan. Maybe if we're lucky we might see ADA Supers like we did with Turing, with much better specs for roughly the same prices, in order to try to counteract their own hubris with these stupid products.
We can only be so lucky.

At least people aren’t buying this crap. Nvidia needs a successor to TU116. Ada supers at the same prices wouldn’t solve much IMO. Like you said, nobody wants to spend $1200 for a distant second best.

 

[Tagrineth]

At least people aren’t buying this crap. Nvidia needs a successor to TU116. Ada supers at the same prices wouldn’t solve much IMO. Like you said, nobody wants to spend $1200 for a distant second best.

It wouldn't solve the problem but it would help. Part of the issue is that the price/perf is entirely stagnant, the new GPUs are right in line with Ampere and RDNA2's price/perf which is a complete fucking joke.

 

[McHuj]

For me personally this is the first gen of GPUs I’m skipping since whatever released pre-GTX4xx series.

The perf/cost at the x80 level isn’t there but more importantly I don’t think software exists to really stress these new cards.

 

[DegustatoR]

Personally I'm expecting Blackwell and RDNA4 might come out sooner than expected, with significantly more aggressive pricing.

One is in a direct opposition the the other. They may either come sooner than expected or with a more aggressive pricing (in which case they are highly likely to come later than expected, not sooner).

We might see this generation of GPUs end up being very limited in the market space and have a much shortened lifespan.

Turing did fine and was your typical 2 years generation. Why would this generation be any different?

Maybe if we're lucky we might see ADA Supers like we did with Turing, with much better specs for roughly the same prices, in order to try to counteract their own hubris with these stupid products.
We can only be so lucky.

This is highly likely to happen but why is this us being "lucky"? A mid-gen refresh is as expected as anything.

 

[Henry swagger]

Crypto miners had the money not gamers.. majority of pc gamers are on a 1060 most can't afford even a 3080

 

[arandomguy]

It's counterintuitive to most people but in a macro situation with contracting demand from a business stand point it can makes sense to sell for as much per unit as possible instead of going for volume. The reason being that the volume simply isn't there unlike in an macro environment with growth. It's not that the margins aren't their to sell at lower prices, it's that the increased sales would not make up the lower per unit margin.

GPUs are durable goods that also not perishable, these are important characteristics that greatly change the dynamics on how to approach pricing with respect to demand. This is even of greater important for the actual majority of market as they look to upgrade based on usability as opposed to product improvements. A "cheap" GPU sold now that is "good enough" essentially is also a lost future sale. You see this sentiment being brought up often along the lines of say "I got my RTX 3080 for $700 and it's good enough for what I play, I see no reason to upgrade at these prices" without the people posting that realizing why being aggressive on pricing, especially at the onset, does not make sense from a business stand point and is actually not sustainable.

There is a reason why market segmentation is rife in the tech space. If say 2/10 customers are willing to pay $400 while the rest are only willing to pay $200, pricing will not drop until those 2/10 customers have reached some level of saturation. Dropping to $200 right at the onset would represent a significant opportunity loss.

 

[Cyan]

Crypto miners had the money not gamers.. majority of pc gamers are on a 1060 most can't afford even a 3080

they forgot to make attractive GPUs for the average gamer. Now we have GPUs with a power consumption like an electric thermal radiator that cost like a second hand car, for which you need a new PSU or a new PC case. Then you read that the RTX 4090 is the biggest leap in years. That's how this generation is, accompanied with some miserable sales and I wish they don't sell much tbh. I am not contributing to this decadence so I am not buying this, they don't even hype me.

https://twitter.com/i/web/status/1608199772506128388

 

[DavidGraham]

Turing did fine

Don't know where all the fuss about Turing is coming from! Turing did fine sales and market share wise.

 

[yuri]

When Turing released Reddit/HW forums were full of users who were not satisfied with the perf/$ vs the older gen. Many of them even claimed going Radeon instead of Turing.

As always, those revolting users were just a vocal minority and many of them still bought nVidia.

 

[DegustatoR]

When Turing released Reddit/HW forums were full of users who were not satisfied with the perf/$ vs the older gen. Many of them even claimed going Radeon instead of Turing.

As always, those revolting users were just a vocal minority and many of them still bought nVidia.

I somewhat struggle to think of a new lineup where every position was clearly better than what was available at (more or less) the same price previously.
The top end is basically the only segment where each new generation has been bringing enough change in perf/features per price for owners of the a previous top end product to upgrade.
Below that? It has always been a disappointment for someone.

 

[Seanspeed]

I knew ADA and RDNA3 were being engineered for a crypto mining-dominated market.

But this isn't correct. The pricing might be engineered for crypto-influenced pricing expectations, but the hardware is clearly not. I mean, we've seen fairly limited increases in actual memory bandwidth here, which would be one of the biggest priorities if you were gonna do a chip for cryptomining specifically.

The high power ceilings are also not cryptomining-friendly choices, either.

There's no reason for these Nvidia GPU's to be so expensive other than greed. Nvidia doesn't want to have to 'come back to reality' following the crypto crash and want to see if they can get away with gouging gamers to prop up their profits.

The idea that they're 'overengineered' and that's the problem is just bizarre. :/

 

[Seanspeed]

Don't know where all the fuss about Turing is coming from! Turing did fine sales and market share wise.

It very obviously didn't. Steam hardware survey was enough to make it clear how much less popular it was for gamers compared to Maxwell and Pascal before it.

 

[Malo]

Nvidia won't want to drop prices because they and everyone else knew there would be a massive drop in sales. So for the small amount of sales they were going to get anyway at the high end, maximize profits.

 

[hurleybird]

Nvidia won't want to drop prices because they and everyone else knew there would be a massive drop in sales. So for the small amount of sales they were going to get anyway at the high end, maximize profits.

I think that's the sort of short sighted, one dimensional MBA-type logic that sinks companies, and almost killed Intel. If we were just coming off a generation like Terrascale with cheap GPUs about, then sure, that might track. But the reality is that for many years a large swath of gamers have been priced out of the market and would jump at the chance to buy an aggressively priced GPU.

 

[Remij]

Well deserved!

 

[Potato Head]

You know those numbers are from the prior generation GPU’s right?

 

[PSman1700]

Nvidia won't want to drop prices because they and everyone else knew there would be a massive drop in sales. So for the small amount of sales they were going to get anyway at the high end, maximize profits.

Explains not just the gpu market then.

 

[Cyan]

But this isn't correct. The pricing might be engineered for crypto-influenced pricing expectations, but the hardware is clearly not. I mean, we've seen fairly limited increases in actual memory bandwidth here, which would be one of the biggest priorities if you were gonna do a chip for cryptomining specifically.

The high power ceilings are also not cryptomining-friendly choices, either.

There's no reason for these Nvidia GPU's to be so expensive other than greed. Nvidia doesn't want to have to 'come back to reality' following the crypto crash and want to see if they can get away with gouging gamers to prop up their profits.

The idea that they're 'overengineered' and that's the problem is just bizarre. :/

not to mention the issue here is that contrary to other generations where there were the typical complains -yet they were selling-, the GPUs manufacturers aren't selling. It's not that the 4090 is bad, on the contrary, it is the most capable GPU build to date, nor that the 4080 is also bad per se. It's just that they are trying to sell the premium-tier to get as much money as possible while they release the mid and low tier cards in dribs and drabs.

They sold 22 millions of GPUs during Q4 in 2009 between AMD and nVidia combined. Now compare with the 6 million sold between AMD, Intel and nVidia combined this year during Q3.

 

[DavidGraham]

Steam hardware survey was enough to make it clear how much less popular it was for gamers compared to Maxwell and Pascal before it.

Maybe less polular than Ampere and Maxwell, but still, Steam survery and market share numbers showed Turing dominating over RDNA1/Radeon VII. and when Pascal went end of life, Turing went ahead strong.