Some of NVIDIA’s GPUs are confusingly named. In particular, there are at least 4 recent (ish) models with the number
6000 in their name, all with significantly different performance and pricing.
What is the NVIDIA (Quadro) RTX (Pro) (A)6000 (Ada/Blackwell) GPU?
These are the four models and their specs:
| Model name | Year introduced | Architecture | VRAM | CUDA cores | Tensor cores | Memory bandwidth |
|---|---|---|---|---|---|---|
| RTX Pro 6000 | 2025 | Blackwell | 96GB | 24,064 | 752 | 1,597 GB/s |
| RTX 6000 | 2023 | Ada | 48GB | 18,176 | 568 | 960 GB/s |
| RTX A6000 | 2020 | Ampere | 48GB | 10,752 | 336 | 768 GB/s |
| Quadro RTX 6000 | 2018 | Turing | 24GB | 4,608 | 576 | 672 GB/s |
As is clear from the table, these GPUs are very different, and the newer models will be better than the older models. As well as the listed specs, the more modern architectures have other benefits (like better low-precision support).
Note also that some of these have multiple sub-variants — e.g., the RTX Pro 6000 has datacenter and workstation versions. See the linked page on NVIDIA’s site for more details.
To find cloud providers who offer these GPUs for rental, see our cloud GPU comparison page.