Nvidia reaches high on graphics performance with Tesla K80

Nvidia reaches high on graphics performance with Tesla K80

Nvidia's in 2016 will release graphics products with the NV-Link interconnect, which is five times faster than PCI-Express 3.0

Nvidia's PC graphics chips may draw all the attention, but supercomputing chips are driving the company's GPU technology ahead.

The company on Monday announced an updated Tesla K80 graphics card, which is its fastest graphics product to date. Tesla is used in some of the world's fastest supercomputers to solve pressing scientific questions.

The K80 is based on technology similar to that used in GeForce PC graphics cards, but with some memory and throughput tweaks. The graphics card can be used by engineering companies to simulate visual models, or by oil and gas companies for geological research to find the best drilling sites. Tesla has also been used in servers to deliver virtual desktops to remote clients.

The K80 delivers 8.74 teraflops of single-precision performance compared to 5 teraflops on Nvidia's flagship GeForce GTX 980 desktop graphics card. The K80 also has two times the performance and memory bandwidth of its predecessor, the K40, which was announced around the same time last year.

In 2016, Nvidia is promising even faster graphics products based on Pascal architecture, a new on-chip communication technology. Through a new interconnect called NV-Link. GPUs will be able to exchange data faster, which will facilitate better performance. Nvidia says NV-Link is five times faster than the PCI-Express 3.0 interconnect, which is used in most servers and PCs.

Nvidia's Tesla is one of many server accelerators available. Intel sells the Xeon Phi chip, while Advanced Micro Devices offers FirePro cards for servers and workstations. Software needs to be written specifically to take advantage of these processors, and Nvidia offers its CUDA parallel programming framework for Tesla.

One knock on GPUs has been the amount of power they chew up in processing. As a result, they have to be used for specific tasks.

The K80 chip is broken up into two graphics processors, which share the balance of processing. Combined, the chip has 4,992 processing cores, 24GB of GDDR5 memory and memory bandwidth of 480GBps. A technology called GPU Boost can kick up performance for specific applications in bursts to deliver better results.

Hewlett-Packard, Dell, Quanta and Cray will use the K80 GPUs in servers. Servers from HP and Dell will run on Intel's Xeon CPUs.

Dell's PowerEdge C4130 is a 1U server that looks more like an appliance and will be able to accommodate up to two K80 cards. A number of C4130 servers could be used as cabinets to build a supercomputer, said Jimmy Pike, chief architect for Dell's enterprise solutions group.

Agam Shah covers PCs, tablets, servers, chips and semiconductors for IDG News Service. Follow Agam on Twitter at @agamsh. Agam's e-mail address is

Join the CIO New Zealand group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.

Join the CIO New Zealand newsletter!

Error: Please check your email address.

Tags Graphics boardsservershardware systemsComponentsnvidia

More about AdvancedCrayDellHPIDGIntelNewsNvidia

Show Comments