NVIDIA’s Quadro 2000 is the mid-range professional graphics solution in its Quadro line of Fermi-based cards. The Fermi architecture cards have the first professional-class graphics processing units to integrate high performance computing capabilities with advanced visualization techniques. Featuring Scalable Geometry Engine technology, the Quadro can deliver 1.3 billion triangles per second, a significant leap over previous 3D graphics limitations.
The Scalable Geometry Engine is a significant technological advance in that it allows multiple graphics processor clusters (GPCs) to create graphics triangles. Compared with previous generations, each cluster is able to create twice the number of raw polygons in a single clock cycle, leading to dramatically faster performance. Hence, drawing speed is MUCH faster on workstation using a Quadro card. With 192 CUDA parallel processing cores, the Quadro 2000 is a certified solution for leading CAD packages including Autodesk AutoCAD, 3Ds Max, Maya, Revit and Inventor. By comparison, the previous generation Quadro FX 1800 had only 64 CUDA cores.
Using the Quadro on a capable system with 3Ds Max, Revit and AutoCAD installed I was able to achieve vastly higher levels of performance than I could using my system’s original card. The big leap in productivity, for applications such as Autodesk’s that have been written for GPU acceleration, is the ability to access all of those CUDA graphics processing cores for regular operations that used to require CPU power, such as rendering and shading in 3Ds Max and line rendering in AutoCAD and Revit.
NVIDIA’s GigaThread Engine allows to 10 times faster context switching operations over previous generation GPUs. This allows applications such as 3Ds Max to achieve higher performance by permitting its functions to access GPU resources simultaneously. Designed to maximize parallel kernel execution, the GigaThread Engine allows multiple tasks to run concurrently. In the case of the Autodesk apps and the Quadra 2000 I tested, the GigaThread Engine provides dual copy engines, which allows simultaneous transfer of data into and out of the GPU in addition to either 3D processing or computing. The design of the card itself creates higher graphics processor core efficiency and provides maximum data throughput.
Another bit of homework done by NVIDIA that will help designers is the Fermi architecture’s support for ECC (error correcting code) memory. ECC protection detects and corrects single-bit soft errors due to fluctuations in voltage or interference near your computer, before they affect the system.
My Quadro 2000’s GPU supported Single-Error Correct Double-Error Detect (SECDED) ECC codes that detect and correct any single bit error in hardware as the data is accessed. In addition, SECDED ECC ensures that all double bit errors and many multibit errors are detected and reported so that the program can re-run them rather than being allowed to continue executing with bad data. Users can configure ECC memory support on the Fermi based Quadro cards to be enabled or disabled via the NVIDIA Control Panel.
All of these advancements are great news for designers who have never been able that kind of performance out of a mid-range graphics card available on the market for less than $450 before. You can expect more advancements in computing, witness NVIDIA’s announcements at this year’s Consumer Electronics Show, to continue to bring the cost of computing lower.