Aceitamos criptomoeda — para que as moedas que detém possam comprar material informático real.Pague em cripto, gaste-a em material informático real. Envio seguro gratuito para todo o mundo acima de 399 $. Caixas neutras, sem marca, expedidas a partir da Alemanha.Envio seguro gratuito acima de 399 $. A sua taxa fica bloqueada durante 30 minutos assim que abre a finalização da compra.Taxa bloqueada 30 minutos. Cada artigo é selado, o seu número de série é verificado, e está coberto por uma garantia de 24 meses.Selado, garantido 24 meses.

Artigos guardados Iniciar sessão
HP Tesla V100 16GB HBM2 PCIe graphics card with single 8-pin power connector

HP

HP NVIDIA Tesla V100 16GB HBM2 workstation graphics card Q2N68A

SKU CH-19MS-FKXC0

Delivers 16 GB HBM2 memory for AI and HPC workloads over a PCIe 3.0 x16 interface

  • GPUTesla V100
  • Memory TypeHBM2
  • 16GB HBM2 memory accelerates large-model training
  • 250W TDP fits standard workstation power budgets
  • PCIe 3.0 x16 interface integrates with existing servers
  • DirectX 12 and OpenGL 4.6 support professional visualisation
  • CUDA 7.0 enables GPU compute workloads

HP Tesla V100 16GB HBM2 compute card

This HP workstation graphics card carries an NVIDIA Tesla V100 GPU with 16GB of HBM2 memory. It is built for double-precision compute, AI training and high-performance simulation workloads. The board uses a PCIe 3.0 x16 interface and draws up to 250W through a single 8-pin power connector. It targets data-centre and workstation users who need dense memory bandwidth in a standard expansion slot.

Suited for dense GPU compute.

Engineers running CUDA 7.0 kernels, OpenCL 3.0 or Vulkan 1.2 workloads will benefit from the high-bandwidth memory and 250W thermal envelope. The card also supports DirectX 12 and OpenGL 4.6 for visualisation tasks. Buyers who need only display output or single-precision gaming performance should look at something else. Systems without a 250W power budget or a free 8-pin connector cannot use this board.

16GB HBM2 memory capacity

The 16GB HBM2 frame buffer is the specification that decides fit. It determines the maximum model size, dataset tile or simulation domain that fits in GPU memory without paging. Workloads exceeding this capacity will spill to host memory and lose performance. Check your application memory footprint against the 16GB limit before purchasing.

Pontos fortes

  • 16GB HBM2 memory accelerates large-model training
  • 250W TDP fits standard workstation power budgets
  • PCIe 3.0 x16 interface integrates with existing servers
  • DirectX 12 and OpenGL 4.6 support professional visualisation
  • CUDA 7.0 enables GPU compute workloads

Especificações

GPUTesla V100
Memory TypeHBM2

Perguntas sobre este artigo

How is the PCIe 3.0 x16 card installed and what step is often missed?

Insert the card into a PCIe 3.0 x16 slot and connect the 1x 8-pin power connector; the step often missed is ensuring the power cable is fully seated before booting.

What maintenance or replacement will the 16GB HBM2 card need later?

The 250W TDP means dust buildup on the heatsink will raise temperatures; clean the fins and fan annually and replace thermal paste if temperatures exceed safe limits.

What limit appears first under sustained compute load?

The 250W power limit is reached first; the GPU will throttle clocks to stay within the 1x 8-pin connector's delivery capacity during long Cuda workloads.

Também nesta secção

Os clientes também compararam

Mesma prateleira, mesma finalização da compra — oito moedas e uma taxa bloqueada durante 30 minutos.