Nvidia Corp. yesterday added seven graphics cards to its Quadro line for professionals, including three that support a new technology designed to enable users to run multiple design and animation applications in virtual machines with their own dedicated Quadro cards on a single workstation.
Nvidia said the SLI Multi-OS virtualization platform can be used to move rendering work that currently is done, albeit slowly, by virtualization software back onto the Quadro graphics processing units, or GPUs. And the cards that support SLI Multi-OS — the Quadro FX 3800, FX 4800 and FX 5800 — are more powerful than most of Nvidia's consumer graphics cards aimed at gamers and other PC enthusiasts.
The cheapest of the three, the $1,199 FX 3800, has 192 processor cores and 1GB of video memory, while the most powerful, the $3,499 FX 5800, comes with 240 processor cores and 4GB of video RAM.
The cards and the SLI Multi-OS technology work with Parallels Inc.'s Workstation Extreme virtualization software, which also was introduced on Monday. The combined technologies support virtualization of Windows and Linux applications and are aimed at engineers who use computer-aided design and manufacturing tools as well as digital animators who use software such as Pixar's RenderMan Pro.
Using SLI Multi-OS isn't cheap: Each application that is run in virtual mode requires its own dedicated graphics card. But, Nvidia contends, that approach will still prove to be less expensive than having users run each app on separate workstations, as many do now.
Nvidia said SLI Multi-OS also could be used in conjunction with virtualization tools from vendors other than Parallels, including market leader VMware Inc. and Microsoft Corp. But an Nvidia spokeswoman declined to comment on when other vendors would support the new technology.
All seven of the new Quadro cards support CUDA, an architecture designed by Nvidia for offloading graphics processing work from the CPU to the GPU. The goal in doing so is to speed up applications that can be rewritten for parallel processing so they can take advantage of the numerous processing cores in Nvidia's cards. Central processors, in comparison, typically have a maximum of eight cores now, apart from very high-end server chips.
Nvidia has even rolled out its own line of CUDA-based systems, the Tesla Personal Supercomputer, to demonstrate the architecture's potential. And it is planning to invest as much as $5 million in software vendors that write applications to take advantage of CUDA.
But CUDA has potential competition. ATI Technologies Inc. is working on a similar technology called ATI Stream that the subsidiary of Advanced Micro Devices Inc. announced last November.