So it seems that the 3090 will be priced at 1499$. This is kind of insane.
EDIT: For people comparing this to the Titan RTX, no. This GA102, not GA100. It's the cut-down version of Ampere. GA100 will come out, and it will be even more expensive.
EDIT: For people comparing this to the Titan RTX, no. This GA102, not GA100. It's the cut-down version of Ampere. GA100 will come out, and it will be even more expensive.
Back in Kepler, a 780Ti was 800$, and it had the GK110 chip, which was the full-fat chip. Now, the cut-down chip costs twice as much.
I'm tempted to get one just to avoid having to think about upgrading a graphics card for another 10 years. Plus I can do some ML messing about as well for resume-driven development.
That statement is absolutely true regarding me. Honestly, I don't need a gaming system at all, let alone one with such a powerful GPU. But even in terms of my leisure-time gaming, I could never justify the price difference over a much cheaper card.
Still, I could imagine it making a lot more sense for other people. E.g., pro gamers, people with big entertainment budgets, or people using CUDA for number-crunching.
Not to mention that it still has NVLink support! The 3-slot design is a bit challenging and for 4-card workstation, I need to rework my radiator mount to make space for another PSU.
If you're happy paying 40% more for a 50% faster card, that's okay. I just don't think it's very good for the industry.
That doesn't mean it's not the Titan equivalent for this generation. Titan X(pascal) and Titan XP were both GP102, and Titan RTX was TU102. AFAIK, only Titan V used the "100" chip, and that was sorta a fluke because there was no smaller volta chip. (and 3090 was explicitly introduced as the Titan RTX replacement)
It's not the Titan, because it's not the biggest chip, and also, it's not called "Titan". It fits the motif of the 2080Ti almost to a T.
Whether the chip number is right is pretty irrelevant.