NVIDIA now has a solution for professionals ready to get a grip on its new Ampere architecture: the RTX A6000 and the A40. They are two powerful cards that comfortably outperform the drool-worthy $1,499 RTX 3090, with 10,752 CUDA cores and 48 GB of VRAM. And indeed, in case you were curious, they were the predecessors of the previous RTX 8000 and 6000 Quadro GPUs.It's not clear why NVIDIA removes the Quadro branding this time around, but the name on the card doesn't matter to anyone seeking professional results.
NVIDIA has not yet announced clock speeds and precise technical information about these cards, but you can think of them as super-powered versions of the RTX 3080 and 3090. They run the same entirely GA102 GPU as the two cards and are otherwise similar to each other, except that the A40 is entirely passively cooled and has a slower 14.5 Gbps memory clock. They also run slower GDDR6 RAM compared to faster GDDR6X on user cards, so there could be some notable performance differences between 3080 and 3090, as Anandtech points out. The company claims that the A6000 is twice as fast as the RTX 8000 in some cases.
As usual, these are not cards that many customers would need to consider, but they are compelling choices for anyone who needs GPU power for professional work. The A40 will be a reliable option for dense computer-focused servers. NVIDIA also features three DisplayPort links, while previous passively cooled cards were designed for headless computing tasks. The company says the RTX A6000 will be available in mid-December, while the A40 will be available early next year.