On Monday I previewed the 2015 NVIDIA GPU Technology Conference (GTC) calling it the biggest VDI conference of the year. Well it has more than lived up to that billing with some of the most eye-opening sessions on desktop virtualization that I have seen for years. Not only were there more VDI sessions than there were at both Citrix Synergy and VMworld last year, but the technical depth of the sessions appeared to be greater than most of those seen at VMworld and Synergy in recent years.
The decision to allocate 49 conference sessions to VDI was amply justified on Monday when NVIDIA released its take on the total addressable market for graphics virtualization workloads. NVIDIA believes that there are as many as 25M high end designers, architects and engineers will want to use NVIDIA GRID cards to run 3D CAD and graphics apps on virtualized GPUs, followed by 200M “Power Users” users apps from Esri, TeamCenter, Adobe, and Autodesk, most of which are likely to be new to VDI and RDSH representing a huge opportunity for Citrix and VMware. A further 400M “Knowledge Workers” looking for better 3D performance than CPUs alone can provide make up the rest of the likely market. Together this could mean a market for over 100M GRID cards, clearly a very significant opportunity not just for NVIDIA, Citrix, And VMware, but with many customers already asking about the availability of 3D DaaS, service providers could secure a significant piece of the new market.
This professional graphics VDI market is likely to be immune to the problems that plagued early VDI implementations. The IOPS challenges that hampered performance and increased costs of early VDI desktops have now all been solved, but even if this were not the case, the financial benefits that virtualizing professional graphics workstation is so great that had storage still cost $500 per desktop 3D VDI adoption is unlikely to slow. During a pre-conference briefing, NVIDIA execs cited a report from a large engineering firm that reported that a 1000 user implementation would save the company $30M annually. Similar qualifications were available from many of NVIDIA customers who had participated in a large early access program which attracted over 400 participants
One of the most interesting discoveries at the conference was just how tolerant of high latency WAN connections the early access program customers were. Speaking to conference speakers and attendees many reported satisfactory experience with >150 ms network latency, high enough to allow organizations to do real time collaboration on complex models from anywhere in the US to anywhere in Europe dramatically changing user productivity.
NVIDIA and VMware are so confident about VMware View’s new GRID enabled cloud desktop that it has made the service available to test from the cloud at www.nvidia.com/vmware-trygrid.