As an Amazon Associate, we earn from qualifying purchases. TweakTown may also earn commissions from other affiliate partners at no extra cost to you.
Priced at over $10,000 in some markets, the new RTX PRO 6000 GPU is a new workstation GPU that can be viewed (in some ways) as a GeForce RTX 5090 with 96GB of VRAM. This thing is a monster, and we got to go hands-on with it at PNY's booth at Computex 2025. Well, hold it and take a look at its physical design.

The NVIDIA RTX PRO 6000 GPU looks like the GeForce RTX 5090 Founders Edition.
Yes, it looks almost identical to the GeForce RTX 5090 Founders Edition (check out our review here), but it has an all-black finish and an 'RTX PRO 6000' label. The cooler design and form factor are identical, with a two-slot thickness, custom PCB, and NVIDIA's new Double Flow Through cooling.
The big thing here, of course, is that it packs a whopping 96GB of fast GDDR7 memory and the full GB202 GPU chip with 24064 CUDA Cores, compared to the RTX 5090's 21760 CUDA Cores. To fit all of that memory inside the card, NVIDIA has placed 48GB on either side of the PCB.
Even though this is a workstation GPU targeting the AI market, it still features the latest generation of Tensor and RT Cores, so it can be used for gaming. Not that you would, as the performance difference probably wouldn't be that noticeable in most games.

This beast is designed for deep learning and scientific computing. Even though it's the single most powerful GPU we've ever held, PNY also gave us a look at a System 76 Thelio Mega system that can house up to 3 x RTX PRO 6000 GPUs paired with a 96-Core Threadripper Pro 7995WX CPU, 512GB of memory, and 64TB of storage. Impressive stuff, but a system that would cost at least $60,000 all up.
For something a little more down to earth, here's PNY's custom chrome-plated GeForce RTX 5080, which was brought to the show and signed by NVIDIA CEO Jensen Huang.
