Having fun with the Voodoo5, I tried out a couple things to stress it out such as running a timedemo at 1600x1200x32. It was a stable card and performed great. Temperatures were nowhere close to the temperatures the Voodoo3 reached thanks to active cooling. I would expect a card that draws power directly from the power supply to get really hot but it wasn't the case this time. After running the demos, I reached out and touched the PCB. It was surprising warm enough to touch. The 2D is great on the Voodoo5, noticeably better than the GeForce; the text on the screen appeared sharp even at 1600x1200. The 350MHz RAMDAC lets you run at very high refresh rates and resolutions. Video acceleration is present but in the same way as the Voodoo3. There is no hardware motion compensation for DVD's so playing them with a Voodoo5 can be more taxing to your processor than would a computer with a hardware decoder.

Both the VSA-100 chips do not share the 64MB of onboard memory; each chip has an allocated amount of 32mb for texture storage. The two chips share the same frame buffer so each has access to the memory needed for rendering the frames. The high fillrate is made possible by the SLI configuration of the chips and the bandwidth they have. There's more information on the Memory Subsystem in the preview so check it out if you have time but we've got to get moving along to the benchmarks!

Batter up! All tests were run twice in Quake 3 with the highest score being recorded. All eye candy was turned on and vsync was off. The only program running while benchmarking was Explorer. I've decided to pit it against a GeForce DDR because they are at similar price points ~$250 and also because we do not have any GeForce 2 cards yet. The power gamers will opt for the Voodoo5 6000 when it comes out and that's the card we will use to test against the GeForce 2.