This article is more than 1 year old
AMD rattles Nvidia's cage with hardware-based GPU virtualization
15 users, one chip
VMworld 2015 AMD has used the VMworld conference in San Francisco this week to take wraps off a new, hardware-based GPU virtualization tech for virtualized workstations.
Known as the AMD Multiuser GPU, the chipmaker claims it can allow up to 15 virtualized desktops to share a single graphics processor without any loss of performance.
The goal is to provide hardware graphics acceleration for virtualized GPU-intensive workloads, including design and manufacturing and multimedia applications and GPU-accelerated computation.
"When these AMD GPUs are appropriately configured to the needs of an organization, end users get the same access to the GPU no matter their workload," AMD corporate veep Sean Burke said in a statement. "Each user is provided with the virtualized performance to design, create and execute their workflows without any one user tying up the entire GPU."
The Multiuser GPU is designed to work in environments based of VMware vSphere/ESXi 5.5 and later, and it offers accelerated drivers for DirectX 12, OpenGL 4.4, and OpenCL 2.0 – meaning users should be able to do most of what they could do on a local machine, only in a virtualized environment.
Exactly how many users can take advantage of the Multiuser GPU depends on what they're trying to do. While as many as 15 typical workers can share the same virtualized chip for Office-type applications, extremely graphics intensive workloads might only support sharing the chip between two simultaneous users, with various scenarios in between.
But another benefit of this hardware-based virtualization approach, AMD claims, is that it makes it much harder for a malicious actor to hack into the system and peep at other users' screens, as compared to software-based systems.
Mind you, a lot of this "software-based" business is a direct knock on AMD's rival, Nvidia, which doesn't have the hardware-based GPU virtualization that AMD has baked into these latest chips.
Nvidia did, however, use VMworld to roll out a new version of its Grid desktop virtualization tech based on its latest Tesla GPUs, and it's claiming it can support as many as 128 GPU-accelerated users per server, if you cram in enough of the boards.
More details on AMD's Multiuser GPU are available on the data sheet here. The company has yet to announce a firm release data for the boards, though, saying only that they're "coming soon." ®