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
NVIDIA Tesla M40 24GB passive GPU accelerator card with large aluminium heatsink and single 8-pin power connector

NVIDIA

NVIDIA Tesla M40 24GB GDDR5 PCIe 3.0 x16 Passive Accelerator

SKU CH-SJVW-265MD

A passive PCIe 3.0 x16 accelerator with 3072 CUDA cores, 24 GB GDDR5 memory on a 384-bit bus and 250 W TDP for machine-learning workloads

  • Memory InterfacePCI Express 3.0 x16
  • 24 GB GDDR5 memory supports large neural-network datasets
  • 3,072 CUDA cores deliver 7 TFLOPS single-precision throughput
  • 288 GB/s bandwidth feeds compute-intensive workloads
  • Passive 250 W thermal design suits dense server racks
  • PCIe 3.0 x16 interface connects to standard accelerator slots

NVIDIA Tesla M40 24GB Accelerator

This passive PCIe 3.0 x16 card targets machine-learning workloads. It delivers 3072 CUDA cores and 24GB of GDDR5 memory for large-model training and inference. A 250 W TDP requires robust server airflow. No display outputs are present.

24GB GDDR5 on 384-bit Bus

The single spec that decides fit is the 24GB frame buffer paired with a 384-bit interface at 6 GHz. That combination yields 288 GB/s bandwidth, feeding the 7 TFLOPS FP32 engine. Workloads that exceed smaller buffers will run without paging.

Dual 8-Pin Power in Dense Servers

Two 8-pin CPU power connectors supply the 250 W draw, so the PSU must provide both leads. Passive cooling moves heat into chassis airflow, freeing PCIe slots that active fans would block. The x16 electrical link matches standard server backplanes.

Highlights

  • 24 GB GDDR5 memory supports large neural-network datasets
  • 3,072 CUDA cores deliver 7 TFLOPS single-precision throughput
  • 288 GB/s bandwidth feeds compute-intensive workloads
  • Passive 250 W thermal design suits dense server racks
  • PCIe 3.0 x16 interface connects to standard accelerator slots

Specifications

BrandNvidia
ModelNvidia Tesla M40
Memory InterfacePCI Express 3.0 x16

Questions about this item

Will the passive cooler fit in a standard server chassis?

The card uses a passive heatsink and requires forced airflow from the chassis fans; verify your server provides sufficient air velocity across the full-length, dual-slot form factor.

What auxiliary power cable is required?

The board draws up to 250 W and is fed by a single 8-pin CPU power connector; ensure your power supply includes this cable.

Can the 24 GB frame buffer handle large training batches?

With 24 GB of GDDR5 on a 384-bit bus delivering 288 GB/s bandwidth, the card can hold sizable models and batches for inference or mixed-precision training workloads.

Also in this aisle

People compared these

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