Wow, Nvidia picked a great time to launch their extortionately priced 4000 series.
Shipments of integrated and discrete graphics processing units dropped to a 10-year low in the third quarter as PC OEMs reduced procurements of CPUs, and gamers lowered their purchases of existing graphics cards while waiting for next-generation products. In contrast, miners ceased to buy graphics boards due to changes that happened to Ethereum. In general, sales of standalone graphics cards for desktops hit a multi-year low.
Regarding the thread title, it sure doesn't look that way at the high end. The 4090 sells out in a few seconds when it comes in stock, and goes for huge markups on ebay. It's just like the 3080/3090 at the height of the mining boom two years ago.
The days of gaining supremacy through a superior node are going to come to an end. Manufacture is too expensive for even Intel to keep it to themselves, especially when they keep losing market share. They will sell their nodes, so if they have a truly superior node, then AMD and Nvidia will let Intel make the GPUs.
But the price/mm^2 is not the same. The Geforce 280 is from 2008 using 65nm where a wafer was ~$2000, the 5nm in 2020 was ~$17000 so the, if napkin maths can be applied then $/mm^2 has increased by a factor 8.5
It is not about sympathy, and I'm personally I think it is crazy that someone is willing to pay that amount of cash for a video card, but as long as there is, Nvidia would be stupid not to sell it at that price.Yea but you can't look at it solely that way. Raw die prices are a fraction of the entire card. If it was $30 with GTX 280 but sold for $599, and die cost becomes $300(assuming 60 per wafer), if you sell it at $999 you more than make up for the cost. So it should be something like $1200-1300, but it's $1599.
And if you take into account optimizations done at the board level and volume deals across the stack, you can see why there's no sympathy for Nvidia.
It is not about sympathy, and I'm personally I think it is crazy that someone is willing to pay that amount of cash for a video card, but as long as there is, Nvidia would be stupid not to sell it at that price.
I think that the $1k crowd at least acknowledge that the 7900XTX is about to be launched and wait until a proper comparison before buying.I'm honestly surprised more gamers didn't buy a 4080. I full expected insane mobs of impulsive gamers to be pepper spraying and clubbing each other to be first in line at Micro Center on release day, but it was mostly scalpers and they didn't even sell out. The way gamers typically buy GPUs makes me see them as creatures instead of people. So easy to ripoff just by putting something they want in front of them, like dropping food in a cage full of starving rats. They just go nuts every time and will pay any price every time no mater what. If Nvidia drops the price even to $1000, it will flip a primal switch in the gamer brain and they will instantly start clawing each other's eyes out to get their hands on the card.
Also sinc
I think that the $1k crowd at least acknowledge that the 7900XTX is about to be launched and wait until a proper comparison before buying.
If you only look on physical die size, you are correct.
My argument is that there is a reason why nvidia can and will sell cards in this price range. The market is there. There is also a reason why a GPU the same size keeps getting more expensive every generation, as the $/mm^2 increases. But most of all there is a lack of competition in high end GPU and leading edge chip production.
It was clear that Nvidia knew that they messed up the pricing once they canceled the 4080 12GB, so some consumers are thinking. However, the 7900xtx is not out yet so why lower the prices of the 16GB version? Once reviews are out they might.I'm honestly surprised more gamers didn't buy a 4080. I full expected insane mobs of impulsive gamers to be pepper spraying and clubbing each other to be first in line at Micro Center on release day, but it was mostly scalpers and they didn't even sell out. The way gamers typically buy GPUs makes me see them as creatures instead of people. So easy to ripoff just by putting something they want in front of them, like dropping food in a cage full of starving rats. They just go nuts every time and will pay any price every time no mater what. If Nvidia drops the price even to $1000, it will flip a primal switch in the gamer brain and they will instantly start clawing each other's eyes out to get their hands on the card.
It was clear that Nvidia knew that they messed up the pricing once they canceled the 4080 12GB, so some consumers are thinking. However, the 7900xtx is not out yet so why lower the prices of the 16GB version? Once reviews are out they might.
Haven't they already bought the 4090?Yeah, but that requires thinking, which is giving panic stricken, impulsive buyers way more credit than they deserve based on their usual insane buying behavior.
Haven't they already bought the 4090?
With fixed wafer cost and defect rate you could get something like 530 100mm square dies but only 50 good 600mm square dies. If the A16 really did cost $110 to make then AD102 would be something like $1100. Clearly, that is not feasible. I think that estimate of A16 price is *very* wrong. More like $30-40 without all the other parts that are packaged with it.Big monolithic dies are too expensive to make. Nvidia will need to move to a save money route like AMD.
If Nvidia's 4N has similar cost to Apples N4(both these are custom 5nm processes), then the chip pricing of the RTX 4xxx series make a bit more sense.
Apple's a16 chip made on N4 cost them $110 dollars per chip for a die that is just a little over 100mm2.
A16 Bionic reportedly costs more than twice as much as A15 | AppleInsider
The A16 Bionic chip found in the iPhone 14 Pro models costs $110 to manufacture, which is 2.4 times more than the A15 found in the iPhone 13 series and in the iPhone 14.appleinsider.com
if it cost $1 per mm2 of die, this would make sense why Nvidia is shifting some production away from it's consumer chips to hopper because even at $1599, nvidia's margin will be worse than last gen.
With fixed wafer cost and defect rate you could get something like 530 100mm square dies but only 50 good 600mm square dies. If the A16 really did cost $110 to make then AD102 would be something like $1100. Clearly, that is not feasible. I think that estimate of A16 price is *very* wrong. More like $30-40 without all the other parts that are packaged with it.
None of that really applies. Even if you assume, at worst case, Apple is somehow paying N3 wafer prices for N4 it ends up in the $30-40 price range. And they're not paying N3 prices.-Apple is also the earliest adopter of virtually all of these nodes and practically bankrolls the development of new nodes for TSMC.
Entirely possible Apple's price is the cost of the bleeding edge tech, not what clients like NV pay after the tech has effectively been paid for by Apple (so cost per mm2 might be much much lower for NV, although still high in absolute terms, just because TSMC has recouped dev cost from apple).
GPU market may be imploding but that isn't stopping NVIDIA from gaining market share:
Total PC GPU shipments:
View attachment 71718
Desktop GPU market share:
View attachment 71719
Q3 2022 Discrete GPU Market Share Report: NVIDIA Gains 88% Market Share Hold, AMD Now at 8% Followed By Intel at 4%
Jon Peddie Research (JPR) has published the full GPU market share report for Q3 2022 that cover AMD, Intel & NVIDIA GPUs.wccftech.com
Fire sale? Man I missed that one!
I mean, its going on now. nVidia cards are still over priced (though much cheaper than they were), but AMD cards are discounted quite a bit.
Yeah, AMD stuff is fire-salin' away. Not sure how much further that they are going to drop.