The Vendor’s Perspective
What do Nvidia and AMD have to say about why you should choose a pro graphics card instead of a consumer-grade one?
We asked Nvidia and AMD to comment. Both companies sent official statements underlining the professional certification of their Quadro and Firepro ranges. Benefits of pro vs consumer A lot of the quoted benefits are a little hard to quantify, such as the technical pre- and post-sales support. The support forums for professional cards will also be focused on issues with professional software rather than problems with the latest games. However, Nvidia also pinpointed driver features that would be beneficial to the professional user. Aside from being optimised for professional applications, Nvidia’s Quadro drivers also ship with a huge list of presets for specific applications. But there is now also increasing interest in harnessing the GPU power for realistic live rendering and even accelerated final output.
In Solidworks, the Realview mode only works with Quadros ( and AMD Firepros), not consumerlevel cards. This means the highly useful Ambient Occlusion, which operates through Realview, will only work with professional hardware. However, although Nvidia talks about Activeshade in 3ds Max being only officially supported via Nvidia Iray or Nvidia mental ray running on Quadros, it’s still compatible with consumer- grade cards. It’s a similar story with the recently released Solidworks Visualize, which is based on the Bunkspeed software that Dassault Systemes acquired when it purchased a 84 per cent stake in maker Realtime Technology AG in December 2013. This is certified with Quadros, but potentially can run on any CUDA graphics. CATIA’S Live Rendering feature, which uses Nvidia Iray, should also work with Geforce cards as well as Quadros.
However, PTC’S Creo Parametric offers a feature called Order Independent Transparency ( OIT). This was originally only available with AMD Firepro cards, but now also includes Nvidia Quadros. Either way, consumer- grade cards aren’t supported. OIT makes transparent materials look more realistic, with correct depth effects and much better performance due to the hardware acceleration. Flex the memory Both AMD and Nvidia pointed out the availability of ECC memory on the top members of their ranges, which we mentioned elsewhere in this article. Nvidia’s Quadros also support Gpudirect across compute nodes. The Geforce only supports this within a single computer. Gpudirect is primarily important for Tesla compute clusters connected via Infiniband, and allows GPU memory to be written to directly, without needing to write to system memory as an intermediary. Workstation users probably won’t have an opportunity to take advantage of this facility. However, the Quadro cards support 64x FSAA antialiasing, and with multiple cards in SLI mode this can be increased to 128x. But Maxwell- generation Geforces offer something called Multi- Frame AntiAliasing, which is aimed at games, and may not be available with professional software. Monitor support The ability to drive large screen arrays or switch between multiple desktops are key features of both Nvidia and AMD professional ranges. AMD’S Eyefinity multi-monitor system was actually introduced with its Radeon consumer range, however, Nvidia’s nview desktop management system was dropped from the Geforce range when Windows Vista arrived, and became a Quadro- only feature. Using a synchronisation card and Gen-lock signal, multiple Quadro cards can also be combined into very large video arrays. Final thoughts There are lots of features that could justify the expense of a Quadro or Firepro over a Geforce or Radeon. However, if you want to investigate using the GPU for rendering, it may not be necessary to stick with pro cards. Although some systems like Solidworks Realview explicitly won’t work with consumer graphics, other systems like Nvidia Iray might. It’s worth borrowing Geforce or Radeon hardware to try this out with your software, and some vendors, such as workstation manufacturer Boxx, have experimented with this and have had considerably positive results.
There is now also increasing interest in harnessing the GPU power for realistic live rendering and even accelerated final output