Grid 2.0 doubles up on users and performance with Maxwell GPUs

Grid isn't just for gaming—Nvidia's graphics-streaming solution is also targeted at business looking to run their professional 3D apps in the cloud. At VMWorld in San Francisco, Nvidia pulled back the curtain on Grid 2.0 and a pair of Maxwell-based Tesla cards—one for rack systems, the other for blade servers. The company also announced support for Linux guest virtual machines.

With Grid 2.0, Nvidia says it's doubled the number of users per Grid server. Now, 128 users can use virtualized 3D apps at a time. The company also claims those users will enjoy "twice the application performance" of the previous version. Each Grid server can use multiple GPUs, with up to 16 users per GPU, so eight GPUs would be needed to support the maximum number of concurrent users.

Previous Grid-compatible cards were built with Kepler GPUs (PDF), but the hardware behind Grid 2.0 is all Maxwell-based. The Tesla M60 has a pair of Maxwell GPUs with 4,096 CUDA cores and 16GB of GDDR5 memory, and it slips into a standard PCI Express slot. The blade-server-focused Tesla M6 has a single Maxwell GPU with 1,536 CUDA cores and 8GB of GDDR5 memory in an MXM package.

Nvidia has partnered with several big server names, including HP, Dell, and Cisco, to make Grid products available. The Grid 2.0 platform will arrive September 15, and interested businesses can sign up for a 90 day trial.

Tip: You can use the A/Z keys to walk threads.
View options

No comments in this discussion yet.