Q
Get started Bring yourself up to speed with our introductory content.

What can admins use as a data center GPU?

There is no formal standard to differentiate consumer and data center GPUs. Still, you should evaluate processing speeds and hardware requirements before GPU installation.

There is no formal industry benchmark that defines consumer and data center GPUs. However, there are hardware and...

processing speed differences between GPU offerings that separate data center and PC-based use cases.

A traditional consumer GPU, such as the Nvidia GeForce GTX Titan Xp, clocks in at up to 0.38 teraflops (Tflops) in 64-bit floating point processing, while the enterprise-class Nvidia Tesla V100 runs as high as 7.8 Tflops. Each GPU has different architectures, number of Compute Unified Device Architecture (CUDA) cores, memory bandwidth and overall functionality.

Vendors classify GPUs by use case and processing speed, so you should identify which workload you need the GPU for and then research the best options. Consumer GPUs are suitable for data centers with low bandwidth and power requirements, testing software, or installation in off-the-shelf PCs.

You should investigate high-end GPUs, however, if you deal with image classification, big data processing, the internet of things or machine learning. You can maximize your hardware investment if your server includes a GPU expansion chassis. Data center GPU cards can provide ample graphics processing capabilities for the most demanding workloads.

Nvidia's data center GPU offerings include the Quadro and Tesla product lines. With Nvidia's recent end-user licensing agreement updates, which ban the use of consumer GPUs in the data center, your selection might be limited to high-end GPUs -- especially if the GeForce and Titan drivers and software aren't licensed for data center use.

Implement preinstalled GPUs in the data center

An increasing number of servers are available with preinstalled and preconfigured GPUs. This enables GPU introduction to be a regular part of the server technology refresh cycle.

There are many server models with various data center GPU options available for preinstallation. For example, the Tesla V100 is available on the Dell EMC PowerEdge C4140 and PowerEdge R740, while the Tesla P40 is preinstalled on the Lenovo NeXtScale nx360, plus the ThinkSystem SD530 and Lenovo D2.

For non-Nvidia offerings, the AMD FirePro S7100X is available on the Dell PowerEdge M630 and the HPE ProLiant WS460c Gen9.

Your selection of preinstalled GPUs will depend on what server hardware is already in your data center and your current processing needs. Someone on your team should be familiar with graphics libraries and languages, such as OpenGL or CUDA, so they can program the GPUs and update any associated software.

This was last published in December 2018

Dig Deeper on Data center server infrastructure and OSes

Have a question for an expert?

Please add a title for your question

Get answers from a TechTarget expert on whatever's puzzling you.

You will be able to add details on the next page.

Join the conversation

1 comment

Send me notifications when other members comment.

Please create a username to comment.

How do you evaluate GPUs for your data center?
Cancel

-ADS BY GOOGLE

SearchWindowsServer

SearchServerVirtualization

SearchCloudComputing

Close