Well there’s the looming threat of the fact that Nvidia is making margins that are way bigger so they can easily match any AMD undercutting and then undercut AMD if they want to.

The products that are red hot right now also use HBM not VRAM.

Right now, there's a big gap in nvidia's product offering: nvidia is happy to let their gaming/affordable cards languish, to stop them competing with the server products at 4x the price.

And the ML community isn't very motivated to port their code from nvidia to amd because the nvidia tax isn't too bad.

If AMD could take one of their existing 24GB consumer cards and quadruple the RAM, though? Open source ML efforts would have support within days.

And nvidia couldn't respond to undercut them without also undercutting their cash cow server GPUs.

AMD is kind of ironically hamstring by x86. They make a lot more profit off their CPUs than they do their GPUs and they only have so many TSMC wafers for both products. If they try to buy their way into more GPU market share they're losing margin on the GPUs they do sell and denying themselves more profitable CPU sales. And Nvidia has historically been pretty good about aggressively back filling any holes in their product stack AMD figures out how to exploit with a late product introduction. AMD seems like they've mostly given up on gaining in the discrete market.

Intel is making some better moves lately but the impression I get is they're losing money on every sale and need to buy their wafers from TSMC like AMD. I'm not sure how long they'll stomach that situation or if they'll eventually get their GPUs onto their own process.

> If AMD could take one of their existing 24GB consumer cards and quadruple the RAM, though? Open source ML efforts would have support within days.

They ironically could do exactly this because the MCD disambiguates the memory config from the gpu config. They would need a new MCD die but they could put four or 8 memory controllers on a MCD instead of 2.

Operating margins are actually down compared to previous gens, roughly similar to 2012 levels, until AI comes into the picture. And gaming cards haven’t increased their margins equivalently.

Nvidia isn’t going to sell gaming cards at an operating loss out of AI money. If the consumer market is soft then it’s soft, and with the general slowdown in Moores law cost progression it doesn’t make sense to have “firesale gens” that set an impossible standard for future gens to beat.

~20% perf/$ progression per generation is going to be about what’s sustainable, because that’s about what tsmc is giving the chip companies. You can “cheat” by using older worse nodes like ampere and turing did, and like AMD did with the RX 7600, to keep costs down, but that’s a one-time boost and consumers grow to expect that pricing going forward. It’s been a market-strategy mistake and has resulted in mis-calibrated consumer expectations around how much costs are actually increasing. Just make the products you can make and sell them at the same margins you’ve always operated at.

Super refresh should recalibrate a couple of the worst misses, 4080 MSRP is indefensible and 4070 ti is just way too expensive for a 12gb card, but, 4060 and 4070 and 4090 are all pretty on-target and are not particularly high-margin products especially compared to dirt cheap Turing and ampere chips.