As you’ll probably grasp from reading the previous pages we’re not overly excited with what Nvidia’s latest brings to the table. That's mostly due to the fact that it isn't a real next generation product anymore. ATI stole most of Nvidia's thunder with the launch of their Radeon 9700 Pro a few months ago. But we’ll let you be the judge as we slowly come around to the performance benchmarks, as the proof of the pudding is usually in the eating. From the looks of the GeForce FX we expected a lot more, based on all of the marketing hype that we’ve been fed by Nvidia we expected a product with innovative features that would clearly set it apart from ATi's offerings. Unfortunately featurewise it isn't very impressive and some things actually strike us as if the product is rushed out, instead of fully developed. That cooling solution for example goes to show that Nvidia is perfectly capable of doing ASIC design but should have contracted a third party to develop their cooling solution as I don’t see many design wins for the GeForce FX from the OEMs, as they will not be happy with the size, cost and noise production of such a heatsink.
Fig 3. A look at the GeForce FX' core, generating up to 50-watts of heat, requiring a rather large heatsink.
Before we delve into benchmarking results we wanted to explain what we measured and why. For starters we’re not going to bore you with comparing the GeForce FX to every videocard under the sun but will rather pit it against it’s archrival, the ATi Radeon 9700 Pro as that’s what Nvidia intended and what ATi expected from the onset. First point to note is that we find it pointless to use benchmarks that are several years old, for example Quake III Arena has long since been past the expiration date and does not include support for many of the new features found in the latest Direct X or the GeForce FX and Radeon 9700 Pro. We've used the below noted system configuration for all benchmarks with both cards:
- Intel Pentium 4, 3.06GHz with HyperThreading
- Asus P4T533-C, Intel 850e chipset motherboard
- Nvidia GeForce FX 5800 Ultra, GeForce FX 5800
- ATi Radeon 9700 Pro, Radeon 9700
- 2 x 256MB Samsung PC1066-32 RDRAM memory
- IBM 120GXP, 120GB, 7200-rpm harddisk
- Windows XP Pro with SP1
- ATi Catalyst 3.0 Drivers, version 7.81
- Nvidia Detonator Drivers, version 42.63
We wanted to use a latest generation game engine which supports many of the new features and look at how well both cards perform at high resolutions with all the graphics settings at maximum, as that is what most people will be using these cards for anyway. We've used Epic’s Unreal Tournament 2003 for that purpose, which, apart from being a fantastic game known to cause several of our staff members to miss one of more deadlines, has all the support built in and generally is very demanding of the 3D graphics card. But we’ll start off with some basic 3Dmark2001SE benchmarks to see how well the GeForce FX does in this de-facto 3D benchmark after which we’ll move forward with the UT2003 benchmarks and try to find the performance gap over the Radeon 9700 Pro that Nvidia claims for the GeForce FX. Just in case you’re wondering, we managed to find it, although it's a tad bit different story than Nvidia pitched it.