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
Dual-GPU server accelerator, bare circuit board with passive heatsink and gold PCIe fingers

NVIDIA

NVIDIA Tesla K80 24GB GDDR5 graphics card

SKU CH-F49T-XB2JP

Dual-GPU accelerator with 4992 CUDA cores and 24 GB GDDR5 on a PCI Express 3.0 x16 interface

  • Memory TypeGDDR5
  • 4992 CUDA cores accelerate parallel workloads
  • 24 GB GDDR5 memory supports large datasets
  • Dual Kepler GK210 GPUs deliver combined compute power
  • PCI Express 3.0 x16 provides high-bandwidth host connection
  • 384-bit memory interface enables fast data throughput

Dual-GPU Kepler Architecture

The NVIDIA Tesla K80 pairs two GK210 processors on a single PCI Express 3.0 x16 board. Each GPU contributes 12GB of GDDR5 memory for a total of 24GB across a 384-bit interface per chip. The card delivers 4992 CUDA cores aimed at high-performance computing and deep-learning workloads in server environments.

Sustained Compute Throughput

Dual GPUs share the thermal envelope of one double-slot form factor, so clock rates settle at levels that maintain continuous throughput rather than short bursts. The 384-bit GDDR5 buses feed each Kepler engine with steady bandwidth for matrix operations and scientific simulations. Workloads that scale across both processors see linear memory capacity gains without host-side bottlenecks.

Server Installation Notes

The board occupies one PCIe 3.0 x16 slot and draws auxiliary power through standard server connectors. Passive cooling requires directed chassis airflow; insufficient front-to-back velocity is the most common cause of thermal throttling. Driver stacks must enumerate both logical GPUs correctly, and workload schedulers should pin tasks to individual devices to avoid cross-GPU latency.

Highlights

  • 4992 CUDA cores accelerate parallel workloads
  • 24 GB GDDR5 memory supports large datasets
  • Dual Kepler GK210 GPUs deliver combined compute power
  • PCI Express 3.0 x16 provides high-bandwidth host connection
  • 384-bit memory interface enables fast data throughput

Specifications

Memory TypeGDDR5

Questions about this item

Does the 24 GB figure represent one single pool of memory?

The card contains two separate GPUs, each with its own 12 GB of GDDR5 memory. The total is 24 GB but it is not a unified pool.

Will this card fit in a standard PCIe 3.0 x16 server slot?

Yes, the board uses a PCI Express 3.0 x16 edge connector and is designed for server chassis that accept full-height, dual-slot accelerators.

How many CUDA cores are available for compute workloads?

The K80 provides 4,992 CUDA cores split across its two GK210 GPUs, giving parallel throughput for HPC and deep-learning tasks.

Also in this aisle

People compared these

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