Personal Supercomputing Systems Reconditioned GPGPU · Boston, MA
Request a quote

Systems

Side-by-side Specification

Both machines run the identical software image. The difference is density, power draw and where the box lives.

Dell C4130 Dell T7910
Price $1,998 $1,298
Form factor 1U rackmount Tower / deskside
GPU cards 4 × NVIDIA K80 1 × NVIDIA K80
GPUs in system 8 2
Total CUDA cores ≈20,000 ≈5,000
CPU Dual Intel Xeon E5-2680 v4, 28 cores, 2.4 GHz (3.3 GHz turbo) Dual Intel Xeon E5-2687W, 3.0 GHz
Memory 128 GB DDR4 ECC, configurable 64 GB DDR4 ECC
Storage 500 GB HDD 500 GB HDD
Power supply 1400 W 1300 W
Best for Throughput, shared lab resource Quiet single-user development
Details Dell C4130 Dell T7910

Prices are for the configurations listed above, collected in Boston. Customizations to RAM, storage or GPU count are quoted on request, and freight is extra.

Warranty: 1 month full warranty. The unit will be repaired at no cost.

In detail

Each machine

Dell C4130

Dell C4130 datasheet

This browser will not display PDFs on the page. Open the datasheet directly instead.

The C4130 rack unit running on a bench under a monitor and keyboard, the screen showing nvidia-smi listing eight Tesla K80 GPUs beside a Mandelbrot set render

The dense one. Four K80 cards, eight GPUs, just under 20,000 cores in a single rack unit, for people who need real throughput from a rack or a lab bench.

GPUs
4 × K80 (8 GPUs)
Cores
≈20,000 CUDA
Memory
128 GB ECC
Storage
500 GB HDD
View full spec

$1,998

Dell T7910

Dell T7910 datasheet

This browser will not display PDFs on the page. Open the datasheet directly instead.

The T7910 tower beside a monitor, keyboard and mouse on a workbench, the screen showing a Mandelbrot set render and terminal output

The quiet one. One K80, two GPUs, ~5,000 cores in a tower that runs on a normal wall outlet under a desk. The natural first CUDA machine for a student or a teaching lab.

GPUs
1 × K80 (2 GPUs)
Cores
≈5,000 CUDA
Memory
64 GB ECC
Storage
500 GB HDD
View full spec

$1,298

Which one

Which one to buy

Both run the identical software image, so code moves between them unchanged. The choice is about where the box lives and how many people share it.

Choose the

Dell C4130 $1,998

if

  • You have a rack or a machine room
  • A class or a group shares it
  • You want maximum throughput
  • Eight GPUs on one host matters
See the Dell C4130

Choose the

Dell T7910 $1,298

if

  • It lives in an office or classroom
  • One person uses it
  • Quiet operation matters
  • It is a first CUDA machine
See the Dell T7910

Need more than 12 GB inside a single GPU? Or neither machine can do it. Send us your requirements and we will get back to you.

Honest fit

What these machines are good at, and what they are not

Worth thirty seconds before you ask for a quote. The short version: this is Kepler-generation hardware, so the toolchain stops at CUDA 11.x and each GPU holds 12 GB.

A good fit for

  • Learning and teaching CUDA
  • Kernel and algorithm development
  • Double-precision numerical work
  • Long unattended jobs, with no hourly meter
  • Visualization, with VTK preinstalled
  • Labs on a fixed capital budget

Not the right machine for

  • Modern large-scale AI training
  • More than 12 GB inside one GPU
  • CUDA 12 or newer toolkits
  • Mixed-precision or tensor-core work
  • Quiet offices, for the C4130
  • Best throughput per watt