Contact
NVIDIA Tesla T4 16GB GDDR6 Data Centre GPU — AI Inference, Full Height, HPE R0W29A
NVIDIASH-NV-TESLA-T4-16G

NVIDIA Tesla T4 16GB GDDR6 Data Centre GPU — AI Inference, Full Height, HPE R0W29A

R 17 360
Available to order
10–15 day delivery

Pay by EFT or bank deposit — we confirm stock and the final price before you pay.

Pre-owned

Pre-owned: 90-day Server Hub limited warranty, return-to-base.

30-day returns on unopened items — you arrange the courier; we pay it on faulty or dead-on-arrival units.

Quantity
1

The NVIDIA Tesla T4 is a Turing-architecture data-centre GPU built for AI inference, video transcoding and virtual desktops. It pairs 2,560 CUDA cores and 320 Tensor Cores with 16GB of GDDR6 memory at 320GB/s, delivering up to 65 TFLOPS of FP16 and 130 TOPS of INT8 within a 70W power budget.

The card fits a PCIe 3.0 x16 slot and draws all its power from the slot, so it needs no auxiliary power cable. This unit comes with a full-height bracket.

The T4 is a common upgrade for Dell PowerEdge, HPE ProLiant and Lenovo ThinkSystem servers that need GPU acceleration without the power and cooling demands of larger cards. Typical work includes large language model and computer-vision inference, CCTV video analytics, media transcoding and NVIDIA virtual GPU desktops (vGPU licences are sold separately).

Condition: Grade A pull, removed from a working server. HPE part numbers R0W29A and P09571-001.

Specifications

Memory
16GB GDDR6

Highlights

  • 16GB GDDR6 memory at 320GB/s
  • 2,560 CUDA cores and 320 Turing Tensor Cores
  • Up to 65 TFLOPS FP16 and 130 TOPS INT8
  • 70W, powered from the PCIe slot with no auxiliary cable
  • Full-height bracket, HPE part R0W29A

Processor

CUDA Cores
2,560
Tensor Cores
320 (Turing)

Memory

Memory Type
GDDR6
Memory
16GB GDDR6
Memory Bandwidth
320GB/s

Power

Power
70W, slot-powered

General

Interface
PCIe 3.0 x16
Part Number
R0W29A / P09571-001
INT8
Up to 130 TOPS
Condition
Grade A pull
Bracket
Full height
FP16
Up to 65 TFLOPS

Related Products

The NVIDIA Tesla T4 is a 70W Turing data-centre GPU with 16GB of GDDR6, 2,560 CUDA cores and 320 Tensor Cores, delivering up to 65 TFLOPS of FP16 and 130 TOPS of INT8. It adds AI inference, video analytics and transcoding to Dell, HPE and Lenovo servers without extra power cables, and this unit has a full-height bracket.

Frequently asked questions

What is the NVIDIA T4 used for?

AI inference for large language models and computer vision, CCTV video analytics, media transcoding and NVIDIA virtual GPU desktops. Its 70W power draw makes it easy to add to an existing server.

Does the T4 need a separate power cable?

No. It draws all its power from the PCIe 3.0 x16 slot, so it needs no auxiliary power connector.

Which servers can take a Tesla T4?

It is a common GPU upgrade for Dell PowerEdge, HPE ProLiant and Lenovo ThinkSystem servers with a free PCIe x16 slot. This unit has a full-height bracket, so it suits full-height slots.

What condition is the card in?

It is a Grade A pull, removed from a working server. HPE part numbers R0W29A and P09571-001.