The biggest driver of chip cost is the size of the chip itself. .
Why: The bigger the chip, the less you can bake on a wafer.
De 2080 was: 545 mm²
The 3080 was massive: 628.4mm²
The 4080 is only: 379mm²
The price increase from 2080 to 3080 was therefore not surprising at all. And it was also much less of an issue because you still made a huge step forward in performance per euro compared to 2080.
The 4080 is a much smaller chip than the 3080, up to 40% smaller, but retails for 50% more in price. And so you really don’t make any progress when you look at performance per euro.
Sure, that size isn’t the only thing that drives the cost, it’s also using a more expensive process, it’s a new architecture, but those are all things that happened with previous generations. But a 40% smaller chip that must therefore be sold at 50% more? I don’t believe it.
The much more likely story to me is that Nvidia sensed during the mining boom that there is a group of people who are willing to pay ridiculous amounts like this for GPUs and want to keep the new price at that level instead of going back to the old level.
Look at AMD, for example, who also managed to create a card with the 7900XTX that shows a significant step up in performance per dollar for the same money as the 6900XT.
Mind you, I buy Nvidia cards way more often than AMD cards, but they can really keep up with this new generation. And for people like you to agree with the story that it’s all because it’s so expensive to produce is exactly what Jensen wants. What do you think about why he yells things like “Moore’s Law is dead”? Meanwhile, AMD knows just how to innovate without massively raising prices.
