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 P40 graphics card, dual-slot full-height board with passive heatsink and gold PCI-E 3.0 x16 edge connector

NVIDIA

NVIDIA Tesla P40 24GB DDR5 graphics card

SKU CH-QJAC-ZBT28

Delivers 47 TOPS INT8 inference throughput with 24 GB DDR5 memory and 346 GB/s bandwidth in a dual-slot PCIe 3.0 x16 form factor

  • Package ContentsComes with P40 card only, refer to images
  • Fastest inference throughput with 47 TOPS INT8 performance
  • Real-time video analytics via 35 HD stream decode engine
  • 24 GB GPU memory at 346 GB/s bandwidth
  • 3840 CUDA cores deliver 12 TFLOPS single-precision
  • Dual-slot PCIe 3.0 x16 fits standard data-center servers

Inference accelerator with 24 GB memory

The NVIDIA Tesla P40 is a data-center GPU built for deep-learning inference. It delivers 47 TOPS of INT8 throughput and 12 teraflops of single-precision compute. The card targets servers that need high-density inferencing and real-time video analytics.

Dual-slot full-height form factor with PCIe 3.0 x16

The board measures 4.4 inches by 10.5 inches and occupies two slots in a rack chassis. A PCI Express 3.0 x16 edge connector supplies the host link and up to 250 W of power. One decode engine and two encode engines handle up to 35 simultaneous HD streams.

Pascal architecture limits and 346 GB/s bandwidth

The Pascal GPU provides 3840 CUDA cores and 24 GB of DDR5 memory fed by 346 GB/s of bandwidth. ECC protection and page-migration support improve reliability for long-running workloads. The 250 W thermal envelope and x16 interface define the practical ceiling for density and throughput.

Highlights

  • Fastest inference throughput with 47 TOPS INT8 performance
  • Real-time video analytics via 35 HD stream decode engine
  • 24 GB GPU memory at 346 GB/s bandwidth
  • 3840 CUDA cores deliver 12 TFLOPS single-precision
  • Dual-slot PCIe 3.0 x16 fits standard data-center servers

Specifications

BrandNVIDIA
SeriesTesla
ModelP40
Package ContentsComes with P40 card only, refer to images

Questions about this item

Will the dual-slot 10.5-inch form factor fit in a standard 2U server chassis?

The card measures 4.4 inches high by 10.5 inches long and occupies two full-height slots, so it fits most 2U chassis that accept full-height, dual-slot PCIe cards.

What auxiliary power connectors are required for the 250 W board?

The P40 draws up to 250 W and requires two 8-pin PCIe power cables from the server power supply; no power adapters are included in the package.

Can the single decode and dual encode engines handle real-time transcoding for a 35-stream security array?

Yes, the hardware video engine is rated for simultaneous decode and inference on up to 35 HD streams in real time, making it suitable for that workload.

Also in this aisle

People compared these

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