NVIDIA reportedly preparing RTX 5000 ADA GPU

NVIDIA has just announced its RTX 4000 GPU for workstations. This is an entirely different class of product than RTX 6000 Ada, the flagship professional GPU featuring the most powerful AD102 GPU with 18176 cores. The RTX 4000, has nearly 3 times fewer cores, but at the same time it has a much smaller physical footprint and no requirement for auxiliary power.

Presently, NVIDIA has a large gap in desktop workstation RTX series based on Ada Lovelace architecture. This void is likely to be filled in new SKUs like RTX 5000 and 4500 later, but thus far NVIDIA made no official comments on such plans.

Turns out, one of the most known NVIDIA hardware leakers, kopite7kimi, expects RTX 5000 GPU already. In fact, there is some confidence that it may feature 15360 CUDA Cores and up to 32GB of GDDR6 memory. That’s still a lot more than any consumer RTX 40 GPU (up to 24GB), but at the same time, these professional RTX GPUs are not using the more capable GDDR6X standard.

Of course, should those specs be accurate, this means that the RTX 5000 must be using AD102 GPU as well, because AD103 can only offer 10240 cores. Furthermore, NVIDIA would have to cut the AD103 memory bus to 256-bit, to match the 32GB config.

There is no information when would such a card be released, but given that the AD102 has been on the market for months and this SKU clearly uses a cut-down variant, the launch could happen at any moment now.

NVIDIA RTX ADA Workstation GPUs
VideoCardz.com	RTX 6000 ADA	RTX 5000 ADA	RTX 4000 SFF ADA
Picture		[ to be leaked ]
GPU	AD102	AD102	AD104
CUDA Cores
Tensor Cores
Boost Clock		TBC
Memory
Memory Bus
Memory Clock		TBC
Max Bandwidth		TBC
Single Precision		TBC
TDP		TBC
Launch Price		TBC
Launch Date	January 2023	TBC	March 2023

Source: Kopite7kimi