Nvidia has finally debuted its RTX 4000 series of graphics cards, starting with the RTX 4090 and RTX 4080, and as expected, they're hot, heavy, and very expensive. Even more expensive than the RTX 3000 series if you can believe it, and though the performance gains are impressive, they're very focused on ray tracing and even then, only when leveraging the new DLSS 3 algorithm.
The flagship is the RTX 4090 and it's built on a 4nm process using the new Ada Lovelace architecture. It will have over 16,000 CUDA cores, a more than 50% uptick over the RTX 3090 Ti, and clock speeds that can reach over 2.5GHz -- an 800MHz increase in turn. That should give it unprecedented performance, with Nvidia claiming that it'll be capable of hitting two times the performance of the RTX 3090 Ti, and up to four times when it comes to ray tracing. That is, however, only when it's using DLSS 3, which uses a new scheduling algorithm to improve performance by dramatic amounts.
The only problem is, the card will have a TDP of 450W, which is a lot of power and a lot of heat. It will also cost $1,600 at launch. That's $100 more than the RTX 3090 did, and with the RTX 4080 prices rising in turn, and the lower specced version of that card being effectively an RTX 4070 in all but name, it's clear what's going on here.
Nvidia is sitting on a lot of RTX 3000 series stock due to the cryptocurrency crash and Ethereum mergers, so it's going to keep selling those cards for the near future at least. That means we won't see more modest RTX 4000 cards for some time and prices aren't going to come down either.
That is unless AMD's RDNA3 can compete favorably at a more affordable price.