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
Passive dual-slot GPU card with silver heatsink, gold PCIe connector visible

NVIDIA

NVIDIA 5.0 Tesla M40 12GB GDDR5 workstation graphics card

SKU CH-ZTGQ-5S7Z2 MPN 5.0

Accelerator card with 12 GB GDDR5 memory, 3,072 CUDA cores and 288 GB/s bandwidth for compute workloads

  • Memory TypeGDDR5
  • 3072 CUDA cores accelerate parallel compute workloads
  • 12 GB GDDR5 memory handles large datasets in memory
  • 288 GB/s bandwidth feeds data to cores without bottleneck
  • Passive heat sink enables silent operation in dense servers
  • 250 W power draw fits standard PCIe + 8-pin power budgets

Tesla M40 Accelerator

The NVIDIA Tesla M40 is a workstation graphics card built for compute acceleration. It delivers 3072 CUDA cores and 12 GB of GDDR5 memory across a 384-bit interface. The board connects through a PCI Express 3.0 x16 slot and draws up to 250 W via a PCIe plus 8-pin CPU power connector. A passive heat sink handles thermal dissipation in a dual-slot form factor.

Suited For Compute Workloads

This card suits users running CUDA, DirectCompute or OpenCL workloads that benefit from 7 TFLOPS of single-precision performance and 288 GB/s memory bandwidth. It also supports Shader Model 5.0, OpenGL 4.6 and DirectX 12 for compatible applications. Buyers who need active cooling, display outputs or a lower power envelope should consider other options. The passive design requires adequate chassis airflow.

12 GB Memory Headroom

The 12 GB frame buffer provides headroom for large datasets and model parameters that exceed smaller buffers. Workloads fitting within this capacity and the 384-bit bandwidth run comfortably at the rated throughput. Tasks demanding more than 12 GB of device memory or higher single-precision throughput will exceed the card's limits. The 250 W power draw also sets a firm boundary for system power budgeting.

Highlights

  • 3072 CUDA cores accelerate parallel compute workloads
  • 12 GB GDDR5 memory handles large datasets in memory
  • 288 GB/s bandwidth feeds data to cores without bottleneck
  • Passive heat sink enables silent operation in dense servers
  • 250 W power draw fits standard PCIe + 8-pin power budgets

Specifications

Memory TypeGDDR5

Questions about this item

Does my server have the PCIe + 8-pin CPU power connector this card requires?

The card draws up to 250 W and needs both a PCIe slot supply and an 8-pin CPU power connector; confirm your power supply provides that connector before installing.

Will the passive heat sink cool adequately in a chassis designed for active GPU fans?

The M40 uses a passive heat sink and relies on strong front-to-back case airflow; verify your server delivers high-volume directed air across the dual-slot 4.40” H x 10.50” L form factor.

How do I confirm the 3072 CUDA cores and 12 GB GDDR5 are recognised after driver installation?

Run nvidia-smi or the NVIDIA Control Panel; both should report 3072 CUDA cores, 12 GB GDDR5 memory and a PCI Express 3.0 x16 link at full width.

Also in this aisle

People compared these

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