NVIDIA RTX PRO 4500 Blackwell 32GB GDDR7 Workstation Graphics Card

SKU:

1006596

£5,160.00 (inc VAT)

Trustpilot
  • Architecture: NVIDIA Blackwell
  • Memory: 32GB GDDR7 with ECC, 256-bit, 896 GB/s
  • CUDA cores: 10,496
  • Single-precision: 51 TFLOPS
  • Interface: PCIe 5.0 x16
  • Outputs: 4x DisplayPort 2.1b
  • Power: 200W, 1x PCIe CEM5 16-pin
  • Form factor: Dual slot, full height, blower cooler
  • Encoders: 2x NVENC (9th gen), 2x NVDEC (6th gen)

£5,160.00

1 in stock

1 in stock

In stock. Order now for delivery on Tue 6 Oct (dispatched Mon 5 Oct, the next working day, orders by 3pm).
SKU: 1006596 Category:

NVIDIA RTX PRO 4500 Blackwell 32GB GDDR7 Workstation Graphics Card

A Blackwell-generation professional graphics card with 32GB of ECC GDDR7 memory, built for AI, rendering and engineering workloads in a 200W dual-slot package.

The RTX PRO 4500 Blackwell Edition is NVIDIA’s mid-range professional card for desktop workstations. It brings the Blackwell architecture, with fifth-generation Tensor Cores and fourth-generation RT Cores, to a card that draws only 200W and fits a standard dual-slot bay. With 10,496 CUDA cores and 32GB of GDDR7 memory on a 256-bit bus, it handles large CAD assemblies, high-resolution video timelines, GPU rendering and local AI model work that would overflow a consumer card’s memory.

Memory with error correction

The 32GB frame buffer uses error-correcting code, which matters for simulation, scientific computing and any job where a silent memory error could corrupt a result. Memory bandwidth is 896 GB/s, keeping large datasets fed to the GPU.

AI and rendering performance

NVIDIA rates the card at 51 TFLOPS single-precision, 153 TFLOPS of RT Core performance and 1,617 AI TOPS with sparsity. Two ninth-generation NVENC encoders and two sixth-generation NVDEC decoders with 4:2:2 support accelerate professional video work, and the card supports DirectX 12, OpenGL 4.6, Vulkan 1.4, CUDA 12.8 and OpenCL 3.0.

Displays and fitting

Four DisplayPort 2.1b outputs drive up to four 4K displays at 165Hz or two 8K displays at 100Hz. The blower-style active cooler exhausts heat out of the case, which suits multi-GPU and rack workstations. Power comes from a single PCIe CEM5 16-pin connector. This is an OEM (small box) unit, and Punch can fit it in a new workstation with a suitable power supply and certified drivers.

Ideal for

  • CAD, BIM and product-design workstations with certified ISV drivers
  • GPU rendering and 3D content creation
  • Local AI development and inference needing 32GB of ECC memory

Specifications

  • Part number: 900-5G147-2250-000
  • Architecture: NVIDIA Blackwell
  • CUDA cores: 10,496
  • Tensor Cores: 5th generation
  • RT Cores: 4th generation
  • Single-precision performance: 51 TFLOPS
  • RT Core performance: 153 TFLOPS
  • AI performance: 1,617 TOPS (with sparsity)
  • Memory: 32GB GDDR7 with ECC
  • Memory interface: 256-bit
  • Memory bandwidth: 896 GB/s
  • System interface: PCIe 5.0 x16
  • Display outputs: 4x DisplayPort 2.1b
  • Maximum displays: 4x 3840 x 2160 at 165Hz or 2x 7680 x 4320 at 100Hz
  • Video engines: 2x NVENC (9th gen), 2x NVDEC (6th gen) with 4:2:2
  • Total board power: 200W
  • Power connector: 1x PCIe CEM5 16-pin
  • Form factor: 4.4in H x 10.5in L, dual slot, full height
  • Cooling: Active (blower)
  • Packaging: OEM, small box

Additional information

Brand

NVIDIA

GPU Class

Professional workstation

GPU Memory

32 GB

GPU memory Type

GDDR7 with ECC

GPU Cores

10,496 CUDA cores

GPU Display Outputs

4x DisplayPort 2.1b

GPU Interface

PCIe 5.0 x16

GPU Power

200W, 1x PCIe CEM5 16-pin

GPU Slot Width

Dual slot

PC Graphics

Share your thoughts!

Let us know what you think...

What others are saying

There are no contributions yet.

×

Continue as a Guest

Don't have an account? Sign Up

Trustpilot Product Reviews