Price: £1,240 plus VAT
The new Quadro FX cards from NVidia are based on
the company’s GeForce FX architecture. They take over the top spot in their workstation graphics line from the Quadro 4 series. The latter cards are still being sold, but have been cut in price to make way for the new FX cards (meaning that there are some bargains to be had
in the mid-range).
The card on test is the Quadro FX 2000, top of the line above the Quadro FX 1000 card – a similar beast, but with fewer features and smaller price tag. Physically, the FX cards mark a departure for NVidia, who will be making and selling the cards themselves with the help of a yet-to-be-named partner company. They’re big and fat, taking two slots (AGP and adjacent PCI), though the PCI slot isn’t actually used. This is similar to the
3DLabs Wildcat cards, which also engulf a neighbouring PCI slot to allow space for GPU-cooling equipment.
The Quadro FX 2000 has, quite simply, the most outrageous cooling system we’ve ever seen on a workstation graphics card. The GeForce FX has a similar scheme, but with a plastic cowl to help channel air over the GPU heatsink. As a result, the Quadro FX also requires a dedicated power supply – very unusual for a graphics card, or any card for that matter. You’ll need to make sure that your system has at
least a 350 Watt power supply and a spare standard-size power connector (the same type as found on hard drives) in order to give the card the juice it needs.
The new architecture doubles the pipeline-count of previous Quadro designs from four to eight. And while NVidia claims that this doubles the performance, it doesn’t necessarily mean double the real-world speed. Comparing the the card to the Wildcat 7110, the FX 2000 comes close in most tests, losing out in the Maya Fluid Effects scene (though this may be a driver limitation as it’s also slower than the Quadro 4 980 GXL), but beating it in the Maya Spheres test. As it costs £40 more than the Wildcat, you’d certainly expect it to do well.
Where this card will really sing is in applications that make use of its new shader programming system (vertex and pixel). NVidia says that the new system, called ‘Cg’, is much more powerful and much simpler to use for programmers wishing to take advantage of the raw processing power of the new cards.
And this is the rub. Unless an application comes with Cg shaders, you have to write them yourself. For production facilities that have the resources to be able to do this, and for it to actually benefit their workflow, the cards are impressive.
The real-time graphics possibilities on offer are incredible: we’re talking real-time shadow-casting and self-shadowing; lighting (not simple OpenGL lighting, but fully featured software lighting); volumetrics; hair and fur; and, of course, anti-aliasing.
If you don’t fall into this category, you’re getting perhaps only 60 per cent of what the FX cards are capable of. They are faster than the previous Quadro range, but not by a massive amount in standard 3D applications such as NewTek LightWave and Alias|Wavefront Maya. Bear this in mind when considering them as an upgrade. It might be more advantageous to upgrade your workstation rather than blow well over a grand just on OpenGL graphics. For advanced visualization and real-time shader writing, though, NVidia has things sewn up.