NVIDIA RTX Pro 5000 Blackwell with 72GB is Out

0
NVIDIA RTX Pro 5000 Blackwell With 72GB
NVIDIA RTX Pro 5000 Blackwell With 72GB

NVIDIA has a new GPU, or perhaps it is a GPU we have already seen, just with a lot more memory. The RTX Pro 5000 Blackwell has been available for some time with 48GB. Now, we have the NVIDIA RTX Pro 5000 Blackwell with 72GB, increasing the memory capacity of the card.

NVIDIA RTX Pro 5000 Blackwell with 72GB Launched

On the specs side, this appears to be a straightforward upgrade. Instead of 48GB, the new card has 72GB of GDDR7 and ECC support.

NVIDIA RTX Pro 5000 Blackwell With 72GB Specs
NVIDIA RTX Pro 5000 Blackwell With 72GB Specs

There was one item that stuck out from the specs. NVIDIA’s MIG feature, or Multi-Instance GPU, changes because of the shift in memory capacity. Of course, the card can use its maximum memory capacity, either 48GB or 72GB, in a single instance. The NVIDIA RTX Pro 5000 Blackwell supports splitting the card into two instances. As a result, there are two 24GB MIG instances on the existing 48GB version and two 36GB instances on the new 72GB version.

Final Words

You may have seen in our recentĀ Dell Pro Max 18 Plus Review that the system came with the NVIDIA RTX Pro 5000 Blackwell laptop GPU. That is a 24GB GPU that is designed for “mobile” platforms. We are putting that in quotes since that was a huge laptop.

Dell Pro Max 18 Plus NVIDIA RTX PRO 5000 Blackwell Flux2 FP8 Task Manager
Dell Pro Max 18 Plus NVIDIA RTX PRO 5000 Blackwell Flux2 FP8 Task Manager

This new 72GB version is a solid option for anyone who wants to stay within what a 300W blower-style cooler can handle in terms of TDP, while also needing a lot of memory. Having 72GB will hold larger models and brings it much closer to the NVIDIA RTX 6000 Pro Blackwell cards. What will be interesting to see is whether there is a market for 8 GPU systems with these cards, or if the expense of going to that large of a server will push people to just get the RTX Pro 6000 anyway. Our sense is that this is likely for smaller systems where 1-2 GPUs is the targeted deployment.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.