Back to all GPUs

NVIDIA RTX PRO 6000 Cloud GPU Specs

HexGrid Cloud rents the NVIDIA RTX PRO 6000 Blackwell Server Edition at $1.99 per GPU hour, billed per minute — about 5% below the $2.09/hr average across 1 other clouds. The RTX PRO 6000 carries 96GB of GDDR7 memory on the Blackwell architecture.

96GB VRAMBlackwellDeploys in ~1 minSecure Cloud

HexGrid Cloud price

$1.99/hr

Billed per minute

VRAM

96GB

GDDR7

FP32

120 TFLOPS

Compute performance

Market average

$2.09/hr

You save 5%

Running the RTX PRO 6000 on HexGrid Cloud

What you get when you deploy with us, beyond the hourly rate.

Live in under a minute

Pick a region, hit deploy, and SSH in. No quota requests, no sales call.

Per-minute billing

You pay for the minutes you use. Stop the instance and billing stops with it.

Ready for training day one

CUDA, PyTorch and the usual drivers are preinstalled, or bring your own image.

No egress charges

Move checkpoints and datasets out without a surprise bandwidth bill.

RTX PRO 6000 pricing across the market

Published on-demand rates from other clouds, so you can see where our RTX PRO 6000 price sits.

Market average $2.09/hr across 1 cloud

5% below market average
  • HexGrid CloudYou're here$1.99/hr
  • RunPod$2.09/hr

Other providers' rates are their published on-demand list prices. They exclude committed-use discounts and can change at any time. Shown for reference only.

About the NVIDIA RTX PRO 6000 Blackwell Server Edition

What the RTX PRO 6000 is built for, and where it falls short.

The NVIDIA RTX PRO 6000 Blackwell Server Edition is a professional datacenter GPU based on NVIDIA Blackwell architecture. It combines 96GB of ECC GDDR7 memory, 1.6 TB/s of memory bandwidth, fifth-generation Tensor Cores, fourth-generation RT Cores, and PCIe 5.0 connectivity. Its large memory capacity and support for FP4, FP8, BF16, FP16, and TF32 make it particularly capable for large language model inference, model fine-tuning, multimodal generative AI, scientific computing, rendering, video processing, and virtual workstation workloads.

Best suited for

The RTX PRO 6000 Blackwell Server Edition combines 96GB of ECC GDDR7 memory with fifth-generation Tensor Cores, FP4 and FP8 acceleration, and professional graphics capabilities, making it well suited to enterprise AI inference, fine-tuning, multimodal AI, rendering, and virtual workstation workloads.

  • LLM inference
  • LLM fine-tuning
  • Agentic AI
  • Generative AI
  • Image generation
  • Video generation
  • 3D rendering
  • Scientific computing
  • Virtual workstations

Strengths

  • 96GB GDDR7 ECC memory
  • 1,597 GB/s memory bandwidth
  • Fifth-generation Tensor Cores
  • FP4 and FP8 AI acceleration
  • Up to four MIG instances
  • Four NVENC and four NVDEC engines
  • Strong AI and professional rendering performance
  • PCIe 5.0 x16 connectivity

Limitations

  • No NVLink support
  • Lower memory bandwidth than HBM-based H100, H200, and B200 accelerators
  • FP64 performance is not optimized for traditional HPC workloads
  • Up to 600W power consumption

NVIDIA RTX PRO 6000 Blackwell Server Edition specifications

Full technical specifications for the NVIDIA RTX PRO 6000 Blackwell Server Edition.

VendorNVIDIAArchitectureBlackwellCategoryDatacenter GPUReleased2025Form factorPCIe dual-slot, FHFLInterfacePCIe 5.0 x16

Memory

Memory
96 GB
Type
GDDR7
Bandwidth
1,597 GB/s
Bus width
512-bit
ECC
Yes

Compute

CUDA cores
24,064
Tensor cores
752
Tensor generation
Gen 5
RT cores
188

AI & compute performance

FP32
120 TFLOPS
TF32 Tensor
117 TFLOPS / 234 TFLOPS sparse
BF16 Tensor
500 TFLOPS / 1,000 TFLOPS sparse
FP16 Tensor
500 TFLOPS / 1,000 TFLOPS sparse
FP8 Tensor
1,000 TFLOPS / 2,000 TFLOPS sparse
RT
355 TFLOPS

Power & physical

TDP
600 W
Form factor
PCIe dual-slot, FHFL
Cooling
Passive
Length
267 mm

Interconnect

Interface
PCIe 5.0 x16
NVLink
No
Infinity Fabric
No

Virtualization

MIG
Yes
vGPU
Yes

Media engines

NVENC engines
4
NVDEC engines
4
AV1 encode
Yes
AV1 decode
Yes

Software

CUDA
Yes
ROCm
No

Precision support

FP64
No
FP32
Yes
TF32
Yes
BF16
Yes
FP16
Yes
FP8
Yes
FP4
Yes
INT8
Yes

RTX PRO 6000 frequently asked questions

Common questions about NVIDIA RTX PRO 6000 Blackwell Server Edition pricing, specifications and workloads on HexGrid Cloud.

How much VRAM does the NVIDIA RTX PRO 6000 Blackwell have?

The NVIDIA RTX PRO 6000 Blackwell Server Edition has 96GB of GDDR7 memory with ECC and 1,597 GB/s of memory bandwidth.

Is the RTX PRO 6000 good for LLM inference?

Yes. Its 96GB memory capacity, fifth-generation Tensor Cores, and FP4 and FP8 acceleration make the RTX PRO 6000 particularly well suited to large language model inference and generative AI workloads.

Does the RTX PRO 6000 support MIG?

Yes. The RTX PRO 6000 Blackwell Server Edition supports up to four isolated MIG instances, allowing its 96GB of GPU memory and compute resources to be partitioned across workloads.

Does the RTX PRO 6000 support FP4 and FP8?

Yes. The RTX PRO 6000 Blackwell Server Edition uses fifth-generation Tensor Cores and supports both FP4 and FP8 precision for accelerated AI workloads.

Does the RTX PRO 6000 support NVLink?

No. The RTX PRO 6000 Blackwell Server Edition connects to the host over PCIe 5.0 x16 and does not support GPU-to-GPU NVLink.

What is the difference between the RTX PRO 6000 and L40S?

The RTX PRO 6000 is based on the newer Blackwell architecture and doubles GPU memory from 48GB on the L40S to 96GB. It also introduces fifth-generation Tensor Cores, FP4 acceleration, PCIe 5.0, and MIG support.

How much does an RTX PRO 6000 cost per hour?

RTX PRO 6000 cloud pricing varies by provider, region, and configuration. HexGrid Cloud lists RTX PRO 6000 instances from $1.99 per GPU hour when available.

Other GPUs on HexGrid Cloud

If the RTX PRO 6000 isn't the right fit, these are also available to deploy.

Browse every GPU we rent

Start training on the RTX PRO 6000 today

$1.99 per GPU hour, billed per minute, running in about a minute.

Specification sources