For some workstation users, there’s no option other than having exclusive access to a PC, one that’s finely tuned for their specific needs. But for many companies, dedicating a single PC to each user might put them on a path to poor ROI. It’s for that reason why virtualized workstations are beginning to take off; it’s not just less expensive, it can also be much more efficient.
This is something that AMD reiterates in the press release for its latest FirePro release, the S7100X. This GPU becomes the industry’s first hardware-virtualized GPU to support Blade servers, and each one is able to support up to 16 workstation users with “practically indistinguishable” performance from a native machine.
It’s important to note that these 16 users are not just going to be given the bare minimum, such as desktop and video acceleration; the GPU can support these users while they create their projects in tools like Siemens NX and SolidWorks. If you’re a VMware workstation user, the S7100X was made for you: the card supports both VMware’s vSphere and ESXi.
AMD doesn’t provide specifics about the card’s hardware, but does note that it carries a 100W TDP and ships with 8GB of GDDR5 on a 256-bit bus (split evenly, that’d provide 512MB to each user). Based on these specs, it’s clear that the S7100X is not going to be for higher-end users, but it’s not meant to be. What it is meant to do is provide ample performance to those that lack justification to get a dedicated machine.
This follows in the footsteps of NVIDIA’s unveiling of the Tesla M10 earlier this month, a GPU also designed to enable virtualized environments for users. Unlike AMD’s solution, NVIDIA’s targets companies needing to provide light virtualization to end users; those who won’t even understand that they’re using a virtualized application or environment. It’s for that reason why NVIDIA’s solution can support up to 64 users per GPU; the Tesla M10 is more for light use, whereas the S7100X is for getting work done.
As the FirePro S7100X isn’t going to be sold direct to consumers, pricing information is not available. AMD has already secured a design win, however, with HP’s Enterprise division. The GPU can be equipped inside of the company’s ProLiant WS460c Gen9 server. Each one of these servers can only be configured with a single S7100X, as the card’s form-factor will prevent the only other mezzanine slot from being used (the same applies to competitor solutions that use the same form-factor). That means that in addition to the 512MB of VRAM made available to each user, a total of 2TB of DDR4 memory can be tossed in as well. Who doesn’t love lots of breathing room?