We take cryptocurrency — so the coins you hold can buy real hardware.Pay in crypto, spend it on real hardware. Free insured shipping worldwide over $399. Plain, unbranded boxes, sent from Germany.Free insured shipping over $399. Your rate is locked for 30 minutes once checkout opens.Rate locked 30 minutes. Every item sealed, serial-checked and covered by a 24-month warranty.Sealed, 24-month warranty.

Saved items Sign in
Black full-length dual-slot graphics card with silver double-flow-through cooler and four DisplayPort outputs

NVIDIA

NVIDIA RTX PRO 6000 96GB GDDR7 AI Accelerator Card 900-5G144-2200-000

SKU CH-JKYT-Z1PS9 MPN 900-5G144-2200-000

AI accelerator with 96 GB GDDR7 memory, 24 064 CUDA cores and 4 000 AI TOPS for large-model training and rendering

  • InterfacePCI Express 5.0 x16
  • Chipset ManufacturerNVIDIA
  • GPURTX PRO 6000
  • CUDA Cores24064
  • Memory Size96GB
  • Memory Clock1792 GB/sec
  • 96 GB GDDR7 memory with ECC for large AI models
  • 8 TB/s memory bandwidth accelerates data-intensive workloads
  • 24064 CUDA cores deliver massive parallel compute power
  • 4000 AI TOPS enable fast inference and training
  • Four DisplayPort 2.1b outputs drive multiple high-resolution displays

Professional AI Accelerator

The NVIDIA RTX PRO 6000 is a desktop AI accelerator card built on the Blackwell architecture for professionals who need maximum local compute. It targets researchers, developers and content creators running large language models, complex simulations or high-resolution rendering workloads that demand extensive memory and tensor throughput.

96 GB GDDR7 with ECC

The card provides 96 GB of GDDR7 memory with ECC across a 512-bit interface delivering 1.8 TB/s bandwidth. This capacity and bandwidth combination is the decisive specification for fitting and feeding the largest models and datasets without offloading to system memory.

Suits large-model work; not for multi-GPU scaling

Buyers who need a single workstation card with the highest available VRAM for local AI training, inference or 3D rendering will match this profile. Teams planning multi-GPU clusters or servers requiring NVLink interconnects should look at products that explicitly support that scaling topology.

Highlights

  • 96 GB GDDR7 memory with ECC for large AI models
  • 8 TB/s memory bandwidth accelerates data-intensive workloads
  • 24064 CUDA cores deliver massive parallel compute power
  • 4000 AI TOPS enable fast inference and training
  • Four DisplayPort 2.1b outputs drive multiple high-resolution displays

Specifications

BrandNVIDIA
SeriesBlackwell
Part Number900-5G144-2200-000
InterfacePCI Express 5.0 x16
Chipset ManufacturerNVIDIA
GPURTX PRO 6000
CUDA Cores24064
Memory Size96GB
Memory Clock1792 GB/sec
Memory Interface512-bit
Memory TypeGDDR7 with ECC
DirectXDirectX 12
OpenGLOpenGL 4.6
DisplayPort4x DisplayPort 2.1b
CoolerDouble-flow-through
AI FeaturesAI TOPS: 4000
Tensor Cores: 5th Gen
Ray Tracing Cores: 4th Gen
NVIDIA Encoder (NVENC): 4x 9th Gen
NVIDIA Decoder (NVDEC): 4x 6th Gen
CUDA 12.8
OpenCL 3.0
DirectCompute
Auxiliary Power Connector1x PCIe CEM5 16-pin
PowerMax Power Consumption: 600W
Power Consumption600W
Dimensions (L x H)5.4” H x 12” L
Slot WidthDual Slots
FeaturesMIG Support:
>Up to 4x 24 GB
>Up to 2x 48 GB
>Up to 1x 96 GB
Shipping weight2.18 kg
Package size375 × 165 × 64 mm

Questions about this item

How does the 4000 AI TOPS figure translate to real workloads?

The 4000 AI TOPS rating indicates peak throughput for FP4 tensor operations on the 5th Gen Tensor Cores; actual inference speed depends on model precision, batch size and software optimisation.

What practical bandwidth does the PCI Express 5.0 x16 link provide?

The PCIe 5.0 x16 interface supports up to 64 GT/s raw data rate, enabling high-throughput transfers between the GPU and host for large model loading and multi-GPU scaling.

What clearance and power connections does the double-flow-through cooler require?

The card measures 5.4 in height by 12 in length and needs a single PCIe CEM5 16-pin auxiliary power connector rated for 600 W maximum draw.

New to this? Buying server hardware with crypto — what to check, what it costs to pay, and what is never asked for.

Also in this aisle

People compared these

Same shelf, same checkout — eight coins and a 30-minute rate lock.