An NVIDIA Tesla is a PCI-E card that features one or more GPUs that are designed to run GPU-specific applications that have been written in CUDA, NVIDIA's parallel processing programing language.

Unlike NVIDIA's GeForce and Quadro graphics cards Tesla cards do not have any graphics outputs, so they cannot be connected to a monitor. However, Tesla cards have much more RAM than a graphics card, which is essential so that everything can be loaded into the frame buffer rather than much slower system memory. In addition, the memory on a Tesla card supports ECC, making them much more reliable for processing complex calculations that may take hours or days to complete.

As modern NVIDIA GPUs have thousands of cores, any task that can be highly parallelised will potentially run much faster on an Tesla card than on a CPU alone, delivering cluster performance in a single workstation.

The table below lists some of the workstation applications that will benefit from the addition of a Tesla card.

Bioinformatics Computational Finance Computational Fluid Dynamics
Data Science, Analytics, and Databases Deep Learning Defence and Intelligence
Electronic Design Automation Imaging and Computer Visions Materials Science
Media and Entertainment Medical Imaging Molecular Dynamics
Numerical Analytics Physics Quantum Chemistry
Oil and Gas/Seismic Structural Mechanics Visualization and Docking
Weather and Climate
NVIDIA Tesla Example

NVIDIA Tesla Performance

Many engineering applications such as those running CFD type simulations are now GPU accelerated. Reducing the time taken to run such simulations will help bring your products to market faster.

ANSYS Mechanical Performance

The graph below shows the potential performance gain seen in ANSYS Mechanical running a direct sparse solver with the numbers quoted being the number of jobs completed per day.

Xeon E5 CPU 8-Cores


Tesla K40


ANSYS Fluent Performance

Pressure based coupled flow solvers in ANSYS Fluent are GPU accelerated; the graph below shows how many seconds different hardware configurations take to calculate a double-precision pressure-based coupled solver.

AMG Solver Time

CPU cluster with 24 Xeon E5 CPUs each with 6-cores


GPU cluster with 48 Tesla K40 cards


Total Solution Time

CPU cluster with 24 Xeon E5 CPUs each with 6-cores


GPU cluster with 48 Tesla K40 cards


Tesla Solutions

Scan is an NVIDIA Preferred Solution Provider and stocks a wide of NVIDIA Tesla cards available in a wide range of form factors in our award-winning workstations and servers.

Tesla GPU Accelerator Cards

NVIDIA Tesla K20

Tesla K20
With 2496 cores and 5GB of RAM the Tesla K20 is the entry-level GPU accelerator for workstations, yet still outperforms CPUs by a significant margin.

NVIDIA Tesla K40

Tesla K40
This mid-range Tesla card has 2880 cores and 12GB of memory allowing you to load much more complex models and simulations than the K20.

NVIDIA Tesla K80

Tesla K80
The flagship Tesla K80 has two GPUs, each with 2496 cores for a total of 4992 per card and sports 12GB of memory per GPU, making it to deliver unparalleled performance.

Tesla Card Specifications

Specification Tesla K20 Tesla K40 Tesla K80
Peak double-precision floating point performance (per card) 1.17 teraFLOPS 1.43 teraFLOPS 2.91 teraFLOPS
Peak single-precision floating point performance (per card) 3.52 teraFLOPS 4.29 teraFLOPS 8.74 teraFLOPS
Cores (GPU / Card) 2496 2880 2496 / 4992
GPU GK110 GK110B 2 x GK210
Memory (GPU / Card) 5GB 12GB 12GB / 24GB
Memory Bandwidth (ECC disabled - GPU / Card) 208GB/sec 288GB/sec 240GB/sec / 480GB/sec

Tesla Workstations

The 3XS GW-Multi is our series of powerful and reliable workstations with NVIDIA Tesla cards. Prices start from as little as £3799 ex VAT and the configurations are fully customisable to meet your requirements.

Deep Learning Workstations and Servers

Deep learning uses the power of neural networks to teach computers how to detect patterns and concepts in data, such as translating languages, image recognition and trend analysis. Scan 3XS Deep Learning systems use NVIDIA GPUs to speed up this highly computationally-intensive process, providing you with the fastest possible research so you can explore multiple network architectures and manage and collect datasets to speed up the delivery of data to your customers.