NVIDIA RTX PRO 6000 Blackwell rental: early access

NVIDIA RTX PRO 6000 Blackwell specifications

Architecture

Blackwell

Announced

2025

Memory

96 GB

Memory type

GDDR7

Memory bandwidth

1.6 TB/s

FP32

120 TFLOPS

FP16/BF16, dense

467.8 TFLOPS

FP8, dense

935.6 TFLOPS

FP4, dense

1,871.2 TFLOPSDerived: NVIDIA’s whitepaper per-SM Tensor Core rate × the 2,430 MHz boost clock in the product brief. For the Server Edition, NVIDIA gives FP16/FP8/FP4 only as rounded 1/2/4 PFLOPS peaks, the level of sparse figures; it publishes no dense ones.

FP64

not in NVIDIA’s card specsNVIDIA’s whitepaper gives FP64 on the GB202 chip as 1/64 of FP32.

RT cores

yes

NVENC/NVDEC

NVENC 4, NVDEC 4

MIG

up to 4 instances

GPU interconnect

PCIe 5, no NVLink

Power (TDP/TGP)

up to 600 W

Form factor

Server Edition (being confirmed)

Memory is specified per GPU. Multiple GPUs have separate memory: using a larger model across them requires compatible software and model partitioning. NVLink does not make them one GPU.

Other variants: RTX PRO 6000 Blackwell Workstation Edition (opens in a new tab), RTX PRO 6000 Blackwell Max-Q Workstation Edition (opens in a new tab)

The RTX PRO 6000 form factor is still being confirmed; memory bandwidth and other specifications depend on it. Specs from NVIDIA’s official materials, checked 26 September 2026.

The RTX PRO 6000 by workload

• Inference: LLMs up to 70B in FP8 on one GPU
• Fine-tuning: QLoRA up to 123B, LoRA up to 32B on one GPU
• Training from scratch: models up to 3B on one GPU
• Rendering: scenes up to 64 GB
• FP64 computing: not suited to this GPU

An estimate from open formulas, not a benchmark or an invoice.

Servers with the NVIDIA RTX PRO 6000 Blackwell

Server configuration (CPU, RAM, disk, network) is being confirmed.

Sizing assumption: up to 8 GPUs per server. We’ll confirm the actual configuration in our reply.

Sources

• NVIDIA RTX PRO 6000 Blackwell Server Edition Datasheet (opens in a new tab)
• NVIDIA RTX PRO 6000 Blackwell Server Edition (opens in a new tab)
• NVIDIA RTX PRO 6000 Blackwell Server Edition Product Brief (opens in a new tab)
• NVIDIA RTX PRO Blackwell GPU Architecture (v1.1) (opens in a new tab)
• NVIDIA Video Encode and Decode Support Matrix (opens in a new tab)
• NVIDIA Newsroom, March 18, 2025 (opens in a new tab)

Similar GPUs

• NVIDIA H200

  141 GB HBM3e · 4.8 TB/s · NVLink

  H200 leads on 4 of 5 metrics, RTX PRO 6000 on 1. H200 has more memory bandwidth (4.8 vs 1.6 TB/⁠s) and FP8 (1,978.9 vs 935.6 TFLOPS). But RTX PRO 6000 has more FP32 (120 vs 67 TFLOPS). FP64: 34 TFLOPS on H200; not in NVIDIA’s card specs for RTX PRO 6000 (NVIDIA’s whitepaper gives 1/64 of FP32 for the GB202 chip).

• NVIDIA B200

  180 GB HBM3e · 7.7 TB/s · NVLink

  B200 leads on 5 of 6 metrics, RTX PRO 6000 on 1. B200 has more memory bandwidth (7.7 vs 1.6 TB/⁠s) and FP16/BF16 (2,250 vs 467.8 TFLOPS). But RTX PRO 6000 has more FP32 (120 vs 75 TFLOPS). FP64: 37 TFLOPS on B200; not in NVIDIA’s card specs for RTX PRO 6000 (NVIDIA’s whitepaper gives 1/64 of FP32 for the GB202 chip).

• NVIDIA B300

  270 GB HBM3e · 7.7 TB/s · NVLink

  B300 leads on 5 of 6 metrics, RTX PRO 6000 on 1. B300 has more FP4 (14,000 vs 1,871.2 TFLOPS) and memory bandwidth (7.7 vs 1.6 TB/⁠s). But RTX PRO 6000 has more FP32 (120 vs 75 TFLOPS).

Dense FP16/BF16 and FP8 from the H100 SXM5 column of NVIDIA’s whitepaper: for the H200, NVIDIA lists only with-sparsity figures, the same as the H100 SXM’s.

The RTX PRO 6000’s FP16, FP8 and FP4 are derived figures. Derived: NVIDIA’s whitepaper per-SM Tensor Core rate × the 2,430 MHz boost clock in the product brief. For the Server Edition, NVIDIA gives FP16/FP8/FP4 only as rounded 1/2/4 PFLOPS peaks, the level of sparse figures; it publishes no dense ones.