Summary

  • RTX Pro 6000 Blackwell GPUs are more powerful than RTX 5090, targeted at workstation users.
  • They come with 24,064 CUDA cores, 600W power draw, 96GB of memory, and are offered in three variants.
  • Available from PNY and TD Synnex in April, other OEMs like Boxx, Dell, Lenovo, and HP in May.

Nvidia just wrapped its annual GPU Technology Conference (GTC), but one announcement slipped through the cracks of CEO Jensen Huang's two-hour keynote. Nvidia is introducing a new range of RTX Pro graphics cards for data centers, desktops, and laptops, and the flagship model goes beyond the power of even an RTX 5090.

RTX Pro 6000 Blackwell GPUs scale even higher

Just not for PC gaming

A render of Nvidia's Blackwell server GPU. Credit: Source: Nvidia

There are a ton of GPUs in Nvidia's RTX Pro Blackwell lineup, but the flagship model is the one that makes the RTX 5090 look like child's play. It's called the RTX Pro 6000 Blackwell Workstation Edition, and although the name implies an entirely new generation, it's built on the same architecture as the RTX 5090. This card is targeted at workstation users, leveraging powerful hardware for data science, simulation, and AI training, and we first learned about it last week.

It comes with the specs to match. The card comes with a staggering 24,064 CUDA cores, which is 10% higher than the RTX 5090. The RTX Pro 6000 Blackwell also comes with a rated power draw of 600W, along with 96GB of GDDR7 ECC memory.

Nvidia will offer the card in three variants, all of which come with the same number of CUDA cores and the same 96GB of memory. There are the Workstation and Server editions, which take slightly different form factors depending on where they're deployed. In addition, Nvidia is offering a Max-Q version. This is still a desktop card -- Nvidia uses Max-Q to refer to laptop power, mostly -- but it comes with a rated power draw of 300W.

The workstation model will be offered as a standalone GPU, fit with the dual flowthrough design that we've seen with Nvidia's Founder's Editions like the RTX 5070. Nvidia says the card will be available from PNY and TD Synnex directly starting in April, as well as from OEMs like Boxx, Dell, Lenovo, and HP starting in May. The server model will be exclusive to OEMs like Cisco, Dell, Supermicro, and Hewlett Packard Enterprise.

Nvidia hasn't shared many details for the Pro offerings lower down the stack, but it says the desktop cards will be available from OEMs starting this summer. For the laptop components, Nvidia says they'll arrive this year from brands like Razer, Lenovo, HP, and Dell. We don't know pricing details yet, but you can bet these cards will run a pretty penny.