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
Full-height dual-slot graphics card with silver passive heatsink, no display outputs, single 8-pin power connector

NVIDIA

NVIDIA Tesla P40 24GB GDDR5 passive workstation graphics card

SKU CH-1ABX-MG4HY

Passive 24 GB GDDR5 accelerator with 3840 CUDA cores delivering 12 TFLOPS single-precision and 47 TOPS integer performance for dense inference workloads

  • Memory TypeGDDR5
  • 24 GB GDDR5 memory enables large-model inference
  • 12 Tflops single-precision compute accelerates deep-learning workloads
  • 47 TOPS integer performance speeds quantised AI tasks
  • Passive heatsink design suits dense server deployments
  • PCIe 3.0 x16 interface fits standard workstation slots

NVIDIA Tesla P40 24GB GDDR5 passive compute card

The NVIDIA Tesla P40 is a workstation graphics card built for data-centre compute workloads. It delivers 24 GB of GDDR5 memory across a 384-bit bus providing 346 GB/s bandwidth. With 3840 CUDA cores it reaches 12 Tflops peak single-precision and 47 TOPS integer performance at boost clocks. The card uses a PCI Express 3.0 x16 interface and draws up to 250 W.

Suited for dense inference

This passive dual-slot full-height board requires strong directed airflow from the chassis; it is not suited for quiet workstations or cases without server-grade cooling. Buyers who need display outputs for visualisation should look elsewhere because the card has no video connectors. Systems limited to a single 8-pin PCIe power cable cannot use the included adapter without two PCIe 8-pin sources.

24 GB GDDR5 memory capacity decides fit

The 24 GB frame buffer is the single specification that determines whether this card can hold your model or dataset. Workloads exceeding this capacity will require multi-GPU configurations or accelerators with larger memory. The 346 GB/s bandwidth and 3840 CUDA cores scale with that capacity for throughput-bound tasks.

Highlights

  • 24 GB GDDR5 memory enables large-model inference
  • 12 Tflops single-precision compute accelerates deep-learning workloads
  • 47 TOPS integer performance speeds quantised AI tasks
  • Passive heatsink design suits dense server deployments
  • PCIe 3.0 x16 interface fits standard workstation slots

Specifications

Memory TypeGDDR5

Questions about this item

Will the passive heatsink need cleaning or thermal-paste replacement over time?

The card uses a passive heatsink with no moving parts, so routine cleaning is limited to removing dust from the fins. Thermal-paste replacement is not a scheduled task and would only be considered if temperatures rise unexpectedly after years of continuous operation.

What thermal limit appears first when the 250 W GPU runs sustained workloads?

The passive cooler must dissipate the full 250 W TDP without airflow of its own, so the chassis must supply enough directed air to keep the heatsink below its thermal ceiling; otherwise the GPU will throttle to protect itself.

How much heat and power does the card produce when running all day at full load?

At sustained load the GPU draws up to 250 W and converts virtually all of that into heat inside the case, requiring the server chassis to move that much thermal energy continuously.

Also in this aisle

People compared these

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