A recent NBD expedition journal has revealed a new NVIDIA Blackwell GPU by the name RTX Pro 6000 X, with an incredibly high GDDR7 memory of 96 GB.
Nvidia is preparing a 96 GB RTX Pro 6000 x GPU in the Blackwell Pro range, intended for AI and intensive workloads
As Nvidia has done with its flagship product and other blackwell GeForce 50 Blackwell high -end standard cards, it seeks to extend the wallet by adding additional RTX GPUs. However, there will be no more powerful GPU than the RTX 5090, at least for the game segment.
The professional segment is a whole different story. Usually GPU manufacturers produce high -end graphics cards for professional use, and we could soon see one with Blackwell architecture. A few weeks ago, one of the 96 GB editions of a RTX 6000 Blackwell GPU was spotted in a NBD shipping manifesto (((And we also saw a RTX 4090 96 GB GPU earlier). Again, we saw the same GPU but with its full name.
The leakage @ Harukaze5719 spotted the list of “RTX Pro 6000 x Blackwell” in NBD shipping newspapers, which are sent to India for test purposes “. There are at least two different sender by the names N *** on and N ** D who sent the GPUs for the tests. As revealed in the shipping newspapers, the RTX Pro 6000 Blackwell has a GB202-870 GPU and of a true capacity of 96 GB.
This particular model will also include GDDR7 memory on a 512 -bit memory bus. The Memory Bus of the same width is also used on the GeForce RTX 5090, but it only brings 32 GB of memory. Since it is a gaming card, 32 GB of VRAM are more than sufficient, but the higher VRAM is very beneficial for intensive workloads such as AI applications, and the 96 GB edition will take care of it.
The TBP is about the same brand as the RTX 5090, but it is 25 W higher. No big change but important to take advantage of the full potential of the card. The RTX Pro Blackwell GPUs not only include the variant of 96 GB, but as spotted by the leak, there is another card with half of the VRAM capacity compared to the RTX Pro 6000 X.
This is with 48 GB of GDDR7 memory on a 384 bits. This looks new because we currently know no Blackwell GPU with a 384 -bit memory interface in the RTX 50 series. While the RTX Pro 6000 x uses a GB202 variant, which is also used for the RTX 5090, the following GPU is the GB203 in the RTX 5080, but it is able to take charge of a 256 -bit memory bus. We do not yet know what this GPU is exactly, but that is also part of the RTX Pro series dedicated to the workloads of the AI.
Nvidia Workstation Graphics Card Linep:
| Graphics card | RTX Pro 6000 x | RTX 6000 ADA | RTX A6000 | Quadro RTX 8000 | Quadro RTX 6000 | Quadro GV100 |
|---|---|---|---|---|---|---|
| Gpu | GPU Blackwell | GPU Ada Lovelace | Amper gpu | GPU Turing | GPU Turing | Volta GPU |
| GPU SKU | GB202 | AD102 | GA102 | Tu102 | Tu102 | GV100 |
| GPU process | 5nm | 5nm | 8nm | 12nm | 12nm | 12nm |
| Size | 750 mm2 | 608 mm2 | 628 mm² | 754 mm² | 754 mm² | 815 mm² |
| GPU hearts | TBD | 18176 hearts | 10752 cores | 4608 cores | 4608 cores | 5120 cores |
| Tensor nuclei | TBD | 568 hearts | 656 hearts | 576 hearts | 576 hearts | 640 hearts |
| Skip the clock | TBD | 2.50 GHz | 1.80 GHz | 1.77 GHz | 1.77 GHz | 1.62 GHz |
| Unique precision | TBD | 91.1 Tflops | 38.7 tflops | 16.31 Tflops | 16.31 Tflops | 16.66 Tflops |
| Specific of the rays layout | TBD | 210.6 TFOPS | 75.4 Tflops | 10 gigarays / sec | 10 gigarays / sec | N / A |
| Truum | 96 GB GDDR7? | 48 GB GDDR6 | 48 GB GDDR6 | 48 GB GDDR6 | 24 GB GDDR6 | 32 GB HBM2 |
| NVLink VRAM | N / A | N / A | 96 GB with NVLink | 96 GB with NVLink | 48 GB with NVLink | N / A |
| Memory bus | TBD | 384 bits | 384 bits | 384 bits | 384 bits | 4096 bits |
| Memory bandwidth | TBD | 960 GB / s | 768 GB / s | 672 GB / s | 672 GB / s | 870 GB / s |
| TDP | 600W? | 300W | 300W | ~ 225W | ~ 200W | 250W |
| Launch price | TBD | US $ 6,800 | $ 4650 US | US $ 10,000 | US $ 6,300 | US $ 9,000 |
| Launch date | 2025 | Q1 2023 | Q4 2020 | Q4 2018 | Q4 2018 | 2018 |