Nvidia has just introduced 12 new Blackwell GPUs for its RTX Pro series

Nvidia gave us a sneak peek before the launch of their RTX 5090, and it has finally arrived. The Blackwell series has made its debut with the monstrous RTX PRO 6000 graphics cards.

The new lineup is comprehensive, featuring:

• Datacenter GPU: Nvidia RTX Pro 6000 Blackwell Server Edition
• Desktop GPUs: Nvidia RTX Pro 6000 Blackwell Workstation Edition (Founder's Edition cooler design), Nvidia RTX Pro 6000 Blackwell Max-Q Workstation Edition (blower-style cooler), Nvidia RTX Pro 5000 Blackwell, Nvidia RTX Pro 4500 Blackwell, Nvidia RTX Pro 4000 Blackwell.
• Laptop GPUs: Nvidia RTX Pro 5000 Blackwell, Nvidia RTX Pro 4000 Blackwell, Nvidia RTX Pro 3000 Blackwell, Nvidia RTX Pro 2000 Blackwell, Nvidia RTX Pro 1000 Blackwell, Nvidia RTX Pro 500 Blackwell.

Nvidia’s previous flagship, the RTX 5090, is no longer the top RTX-series graphics card. With over 10% more CUDA cores, the title now goes to Nvidia’s new PRO series GPUs. Featuring three times the memory of Nvidia’s 32GB RTX 5090, these new PRO GPUs excel in memory-intensive tasks thanks to an impressive 96GB of GDDR7 EEC memory on a single card.

Boasting TDPs of 600W, the new 6000-series PRO Workstation and server GPUs are at the pinnacle of Nvidia’s PRO range. The workstation variant showcases a cooler reminiscent of the RTX 5090, while the server version is equipped with a heatsink optimized for rack-mounted server settings.

Essentially, these two GPUs are quite similar, with the primary distinction being their intended usage. One has a heatsink not suited for PC/workstation setups, whereas the other is not ideal for server environments. Both devices exceed the CUDA cores and memory of Nvidia’s RTX 5090. Moreover, they are compatible with Multi-Instance GPU (MIG) technology, enabling this single GPU to be split into up to four instances, reducing workload interference and enhancing flexibility as needed.

Nvidia has also announced that the RTX 6000 Server edition will deliver a remarkable 125 TFLOPS of single-precision compute performance. Furthermore, it will operate completely passively, without any fans, and is intended to be packaged with third-party server configurations.

Up to this point, Nvidia's focus has been on the business and AI sectors, with no announcements aimed at thrilling everyday consumers. This may change at Computex.

As always, for the latest news on hardware launches and industry developments, be sure to follow our dedicated hardware coverage.