[Updated] Your Thoughts on Die Sizes and Pricing in the Last 6 Years

[Brahmzy]

Keep in mind we pay for the next gen, next process change R&D with this gen. Not that nV doesn’t have cash coffers busting at the seams for their R&D budget, but they’re going to pay for their future product’s R&D with today’s profit. So 7nm is being being paid for by this and last gen’s profits.
Also, call it IPC or whatever, but gains have been getting less and less as we go through these process shrinks. Certainly with intel, and I think that is across the board. Things have been optimized and efficiencies have been honed and we’re approaching diminishing returns, assuming they’re doing things the same way (same general architecture(think Core from intel)).
SO, if the expectation is we need to continue to see massive gains each gen (call it 50%), I think it’s unrealistic to assume linear price increases.
Every kid is an armchair CFO on the forums, bitching about the pricing and blah blah. They’re also the ones playing at 1440 and below.
Those of us playing at 4K have already been eagerly waiting on the 2080Ti for a LONG time as it looks to be a real single card solution to play 4K proper with AA.
It’s expensive, but playing at the top resolution has always been expensive. It usually requires SLI. I’m excited to be moving back to a single card after playing 4K since early 2015. This card is welcome with open arms here. I knew it wasn’t going to be cheap.

 

[PeterScott]

You tell me how 3 independent companies which do not correlate their capacity increase in any way manage to stay well under both projected and actual demand for DRAM chips, increasing their operating profit by 3637% and operating profit margin by 1315%. We have an expression in my country which roughly translates as "kid's bedtime story" - some made up narrative to help reassure the little ones and help them go to sleep easily.

Demand can outstrip production resulting in inflated prices. Having you already forgotten the impact of mining on GPUs?

Were AMD and NVidia colluding?

 

[PrincessFrosty]

Losing sales in light of good supply/inventory buildup, is really the only thing that will drive down pricing.

Competitive products might drive down sales, and lead to lower pricing, or they might not.

I think it's more subtle than this. When you're top dog and you have your faster GPUs out and some features advantage like RT, it means people have to come to you if they absolutely must have those features, and so you can set prices higher because some subset of people will be willing to pay more money for more features. However if the price becomes too high that people cannot individually afford or justify the product (i.e they perceive the value of the product less than what it's priced at), then you'll make less sales in total.

What they're interested in is the maximum profit margins possible, so they'll jack up the price and make more per card up until a point where total number of cards sold start to drop, there's some happy medium in the middle which is where the market as an aggregated lump of people essentially arrive at a maximum profit price point. Usually market research is done before hand to establish where this point in precisely.

Competition adds another factor into play, it means that you'll lose sales if your price is higher than your competitors, and so you only need 2 competitors competing for them to price war down to a point where the profit margins are fairly slim. If AMD enter the market with a RT capable card which is also fast at rasterization and they see Nvidia taking huge margins, they'll undercut them to take their customers away. And Nvidia will respond with their own price cuts, that competition drives down prices. The only time that's untrue is when businesses collude to price fix. At the moment there's no evidence of that.

The moment AMD has a better GPU than NVidia, is the moment they raise their prices higher than NVidia. It really won't change the pricing landscape.

That gives them leverage to price higher than Nvidia because they have a better product, BUT in order to compete with that product Nvidia have to lower their prices, that's the only way they can win, by saying "hey our product isn't quite as good, but look how much cheaper it is" and some people will respond to that and say OK i'll take the hit of a few FPS by not having the best, but save all that extra cash. AMD in response have to lower their prices to make their product more attractive again. They'll still have a price point higher than Nvidia in that scenario, but less than what they'd ideally sell for.

AMD has underdog pricing because they are the underdog, not because they have some kind of moral stance against high margins. Everyone wants high margins. The moment they have that top dog product, they will try to get top dog pricing, and the fat margins to go with it.

More specifically everyone wants the biggest possible margins. Just because you're top dog doesn't give you unlimited scope for as much margin as you like, because almost everyone in the market is willing to trade quality for price to some point. Only and extremely tiny number of people must have the best thing and can also afford any arbitrarily high price it comes with. So you can't just skyrocket your pricing, you have to consider at some point, customers will flee to an marginally inferior product if that product is priced substantially lower. And the market share bares that out, most of the discreet GPU sales are done in the low/mid range, the high end is an exception.

 

[PeterScott]

I think it's more subtle than this. When you're top dog and you have your faster GPUs out and some features advantage like RT, it means people have to come to you if they absolutely must have those features, and so you can set prices higher because some subset of people will be willing to pay more money for more features. However if the price becomes too high that people cannot individually afford or justify the product (i.e they perceive the value of the product less than what it's priced at), then you'll make less sales in total.

What they're interested in is the maximum profit margins possible, so they'll jack up the price and make more per card up until a point where total number of cards sold start to drop, there's some happy medium in the middle which is where the market as an aggregated lump of people essentially arrive at a maximum profit price point. Usually market research is done before hand to establish where this point in precisely.

Competition adds another factor into play, it means that you'll lose sales if your price is higher than your competitors, and so you only need 2 competitors competing for them to price war down to a point where the profit margins are fairly slim. If AMD enter the market with a RT capable card which is also fast at rasterization and they see Nvidia taking huge margins, they'll undercut them to take their customers away. And Nvidia will respond with their own price cuts, that competition drives down prices. The only time that's untrue is when businesses collude to price fix. At the moment there's no evidence of that.



That gives them leverage to price higher than Nvidia because they have a better product, BUT in order to compete with that product Nvidia have to lower their prices, that's the only way they can win, by saying "hey our product isn't quite as good, but look how much cheaper it is" and some people will respond to that and say OK i'll take the hit of a few FPS by not having the best, but save all that extra cash. AMD in response have to lower their prices to make their product more attractive again. They'll still have a price point higher than Nvidia in that scenario, but less than what they'd ideally sell for.



More specifically everyone wants the biggest possible margins. Just because you're top dog doesn't give you unlimited scope for as much margin as you like, because almost everyone in the market is willing to trade quality for price to some point. Only and extremely tiny number of people must have the best thing and can also afford any arbitrarily high price it comes with. So you can't just skyrocket your pricing, you have to consider at some point, customers will flee to an marginally inferior product if that product is priced substantially lower. And the market share bares that out, most of the discreet GPU sales are done in the low/mid range, the high end is an exception.

Of course somewhere marketing has sales vs price projection spread sheets. Prices are usually set to maximize overall profit depending on an array of factors. Generally, the higher price something, the less of them you sell, so there is a peak of overall profit projected at some price. The overall interactions including competition are VERY complex.

But the point I want to highlight is the limited impact of AMD, even with competing products. A few people have made the point that people just want AMD more competitive, so they can get cheaper NVidia cards. This points out how the competitive affect of AMD on NVidia pricing is blunted. In the eyes of the many in the market, these are not two equal products. So even when AMD offers competing performance for a lower price. For Example RX 480 vs GTX 1060. RX 480 was essentially equal performance for less money(before mining), and most people still bought GTX 1060.

That is central to my point. In this world, a merely competitive AMD isn't enough to really hold down NVidia pricing. At this point you need AMD to dominate to really have an impact.

Another point, is you can't point to one specific price increase and say: "That increase right there is lack of competition". Especially when that particular price increase followed a big increase in input costs.

 

[ozzy702]

Of course somewhere marketing has sales vs price projection spread sheets. Prices are usually set to maximize overall profit depending on an array of factors. Generally, the higher price something, the less of them you sell, so there is a peak of overall profit projected at some price. The overall interactions including competition are VERY complex.

But the point I want to highlight is the limited impact of AMD, even with competing products. A few people have made the point that people just want AMD more competitive, so they can get cheaper NVidia cards. This points out how the competitive affect of AMD on NVidia pricing is blunted. In the eyes of the many in the market, these are not two equal products. So even when AMD offers competing performance for a lower price. For Example RX 480 vs GTX 1060. RX 480 was essentially equal performance for less money(before mining), and most people still bought GTX 1060.

That is central to my point. In this world, a merely competitive AMD isn't enough to really hold down NVidia pricing. At this point you need AMD to dominate to really have an impact.

Another point, is you can't point to one specific price increase and say: "That increase right there is lack of competition". Especially when that particular price increase followed a big increase in input costs.

In general I agree. AMD would need to be at levels of competition that we haven't seen in 4+ years for it to have any impact on NVIDIA's pricing.

That said, how do we know that there was a large increase in costs to NVIDIA? Margins keep climbing year after year and I suspect the margins on the 20X0 cards to be extremely high.

How expensive is GDDR6 vs GDDR5X? We're still looking at the same bus widths. Yes it's on a "new" fab but even that shouldn't be much difference in cost, especially compared to when Pascal was introduced. Lack of competition really does mean that even if there were a tiny increase in materials cost, we're looking at many many times that in increased MSRP. Do we actually know that these GPUs cost that much more, if any, to produce?

 

[PeterScott]

In general I agree. AMD would need to be at levels of competition that we haven't seen in 4+ years for it to have any impact on NVIDIA's pricing.

That said, how do we know that there was a large increase in costs to NVIDIA? Margins keep climbing year after year and I suspect the margins on the 20X0 cards to be extremely high.

Do we having gaming card only margins? Because NVidia now sells the $10000 Pro GPUs, and I suspect that they might be helping margins, and mining will have been boosting margins in recent quarters as well.

How expensive is GDDR6 vs GDDR5X? We're still looking at the same bus widths. Yes it's on a "new" fab but even that shouldn't be much difference in cost, especially compared to when Pascal was introduced. Lack of competition really does mean that even if there were a tiny increase in materials cost, we're looking at many many times that in increased MSRP. Do we actually know that these GPUs cost that much more, if any, to produce?

The much bigger dies are a virtual certainty pushing increased input costs. We aren't talking small increase in size either. While I haven't seen official numbers, the RTX 2080 die size looks at least as big as the GTX 1080 Ti die size, if not larger. That is increasing to the next size class. The RTX 2080 Ti die is off the charts large. Unprecedented size in a consumer GPU. Dies are almost certainly the most expensive components in this class of GPU. So most expensive components growing a huge amount in size, gives hefty input cost increases.

 

[ozzy702]

Do we having gaming card only margins? Because NVidia now sells the $10000 Pro GPUs, and I suspect that they might be helping margins, and mining will have been boosting margins in recent quarters as well.
The much bigger dies are a virtual certainty pushing increased input costs. We aren't talking small increase in size either. While I haven't seen official numbers, the RTX 2080 die size looks at least as big as the GTX 1080 Ti die size, if not larger. That is increasing to the next size class. The RTX 2080 Ti die is off the charts large. Unprecedented size in a consumer GPU. Dies are almost certainly the most expensive components in this class of GPU. So most expensive components growing a huge amount in size, gives hefty input cost increases.

Good info. I don't know and haven't seen anything substantiated which is why I asked. I'll be skipping this generation since I'm on a 1080 at the moment and don't game enough to justify a new GPU but 3000 series on 7nm is a lock for me.

 

[jpiniero]

TU104 is roughly 520 mm2 or so.

 

[PrincessFrosty]

Of course somewhere marketing has sales vs price projection spread sheets. Prices are usually set to maximize overall profit depending on an array of factors. Generally, the higher price something, the less of them you sell, so there is a peak of overall profit projected at some price. The overall interactions including competition are VERY complex.

But the point I want to highlight is the limited impact of AMD, even with competing products. A few people have made the point that people just want AMD more competitive, so they can get cheaper NVidia cards. This points out how the competitive affect of AMD on NVidia pricing is blunted. In the eyes of the many in the market, these are not two equal products. So even when AMD offers competing performance for a lower price. For Example RX 480 vs GTX 1060. RX 480 was essentially equal performance for less money(before mining), and most people still bought GTX 1060.

I agree with what you're saying, to some degree Nvidia are the premium product, if you look at their drivers, their features, functionality, buy in with game devs, their reach and overall user experience, they probably beat AMD by having a single large cohesive platform. So AMD could tie with Nvidia in terms of GPU performance and many would still go Nvidia and just enjoy the lower prices overall. This is kinda saying that fanboyism is a real thing and that it's based of a whole bunch of different factors. But they're no factors that Nvidia has a patent on, any other business can do this, So in some sense we just have to admit that part of the overall value judgement is not just power of the GPU but the overall experience.

People can only use the competing AMD products to get cheaper Nvidia GPUs if there's actually other people who buy the AMD GPUs, which there is. So it's not like one dominates over the other by an extreme amount. Someone posted a relative share of the market graph in another thread showing AMD behind but still taking a decent share. if AMD was making almost no sales they'd go bust and not be able to drag down prices from Nvidia. The whole thing is made more complex by the fact that most people aren't interested in raw speed but rather $ to performance ratio and mid range cards, because they have a budget. Which is where AMD cleans up a lot more.

 

[PeterScott]

The whole thing is made more complex by the fact that most people aren't interested in raw speed but rather $ to performance ratio and mid range cards, because they have a budget. Which is where AMD cleans up a lot more.

Not really. Most of the people on a budget still buy NVidia, the GTX 1060 being a prime example. It's the most popular GPU on Steam, and the AMD RX cards barely register.

It's kind of like the Halo effect, where people look at reviews saying NVidia has the best top end GPUs, so when it comes time to buy, even though they are buying far from the top end they still choose NVidia. It takes a lot of work to overcome that.

 

[PeterScott]

Good info. I don't know and haven't seen anything substantiated which is why I asked. I'll be skipping this generation since I'm on a 1080 at the moment and don't game enough to justify a new GPU but 3000 series on 7nm is a lock for me.

I expect you are going to wait 18-24 months for 3000 series, but 7nm will likely be coming to the 2000 series when it is economically viable for consumer volume.

NVidia has done this a couple of times before, where they release a big die at the end of one process, and die shrink in the same family (Like GTX 280 (65nm)-> GTX 285(55nm) 6 months later).

 

[crisium]

I didn't say no matter what. I said if AMD dominated, you would have cheaper NVidia products. But you would be complaining about the expensive AMD products.

It's possible of course (what isn't?) but doubtful. What you said before "The moment AMD has a better GPU than NVidia, is the moment they raise their prices higher than NVidia. It really won't change the pricing landscape."

Do you forget all history? Competition constantly chances the pricing landscape.

$400 GTX 285 was dethroned by the faster $380 Radeon 5870. That was a 6 month AMD lead on the next generation.

Literally days after $550/$400 Radeon 290X/290 launch:

$650 GTX 780 -> $500
$400 GTX 770 -> $330

You don't even need to have the performance crown to have healthy competition that drives down prices.

Mere weeks after launch due to $300 Radeon 4870:

$650 GTX 280 ->$500
$400 GTX 260 ->$300

BTW, the GTX 280 was at that point the largest die size card Nvidia had ever made. They still had to cut prices to lower than smaller die previous generation flagships due to competition.

I will not get onboard your train that we'd have the same thing if AMD could actually compete. Even the 7970 launch saw pp$ improvements compared to Nvidia's 500 series, and consumer friendly price wars ensued throughout 2012-2013 once the 600 series launched. Nvidia is literally not offering a same-price upgrade path for $700 1080 Ti and $350-$450 GTX 1070 customers. If AMD could compete they'd have an opening for cards at those price points to take advantage of $1200 2080 Ti and $600 2070. We have a thread discussing this possible opening.

 

[PeterScott]

Mere weeks after launch due to $300 Radeon 4870:

$650 GTX 280 ->$500
$400 GTX 260 ->$300

BTW, the GTX 280 was at that point the largest die size card Nvidia had ever made. They still had to cut prices to lower than smaller die previous generation flagships due to competition.

I will not get onboard your train that we'd have the same thing if AMD could actually compete. Even the 7970 launch saw pp$ improvements compared to Nvidia's 500 series, and consumer friendly price wars ensued throughout 2012-2013 once the 600 series launched. Nvidia is literally not offering a same-price upgrade path for $700 1080 Ti and $350-$450 GTX 1070 customers. If AMD could compete they'd have an opening for cards at those price points to take advantage of $1200 2080 Ti and $600 2070. We have a thread discussing this possible opening.

Again, I never said NVidia was immune when being drastic upsets.

I totally concede that if AMD releases a card close in performance and prices it at less than half the price,then NVidia would have to adjust pricing or remove the more expensive card from the market. The $300 4870 was quite close to the $650 GTX 280, even matching/beating it some games. The pricing was a massive upset that NVidia had to react to. But even after the NVidia price cuts, AMD's 4870 was still a much better price/$ option, and yet NVidia still outsold them. This is the crux of the issue, and why AMD needs utterly dominant products to shift the market.

So I don't think AMD is prone to making such drastic pricing moves anymore, because in the last several attempts it never really won them market share, and AMD is more prone to seeking better margins these days, rather than trying to pry market from NVidia with pricing.

 

[crisium]

Thinking about it, I actually do think if AMD gets the performance crown that they should price their flagship above 2080 Ti and perhaps save the PP$ winners for mid tier cards. The halo effect of Nvidia's flagships hurt them a lot during the 4000-6000 series.

 

[PrincessFrosty]

It's kind of like the Halo effect, where people look at reviews saying NVidia has the best top end GPUs, so when it comes time to buy, even though they are buying far from the top end they still choose NVidia. It takes a lot of work to overcome that.

There's definitely a level of fanboyism on both sides for sure, one thing that hurts AMD is that they've kinda had years of being somewhat of the underdog, they never really stomped Nvidia except back when the FX5xxx range was dominated by the 9700 I think it was, they took a huge market share back then. But for a very long time they've just not really gone down the huge expensive die route that much, they went through a whole series of cards around the time I had 2x 4870s in CF where they were touting he high end for gamers was essentially 2 inexpensive but reasonably fast cards in CF.

It's decades of reputation that influence purchasers decisions. There's a whole load of other factors as well, when you go to market and the timing, how soon do you release after your competitors products, where are the Nvidia users in their upgrade cycle, and what was the last jump in performance like in the Nvidia range. What time of year is it and how much spare cash do people have at this time. What game titles are out there that can justify an upgrade. What other tech (4k monitors for example) is maturing and becoming useable. What other features do you offer (PhysX, Gsync, 3D stereoscopic, Ray Tracing) and all that stuff.

One thing I stand by that I said before is that Nvidia do seem to offer a much more well packaged bundle of features, I can plug in my 3D glasses and I know it'll just all work together. AMDs more noble approach to go open source on many of these things just doesn't have has as much of a premium feel to it.

I'm not really a fanboy for either although I used to be, I have owned cards from both camps over the decades but it's been mostly Nvidia. If they release a 7nm card and ignore raytracing for now and dedicate most of the transistors to classical renders and smash Nvidias numbers, I'll probably go that route as I don't think I'm ready to step away from 4k, the detail increase is just too good.