System setup and notes
Here's a quick rundown of our test system.
- DFI Xabre X400-T2 64MB (250/250)
- Radeon 9000 (250/200)
- AMD XP2700+ CPU (2.16GHz / 166FSB)
- MSI K7N2 nForce2 motherboard (8x AGP) run synchronously at DDR333
- 512MB (2 x 256MB Corsair Twinx) PC3200LL memory in dual channel mode 2-5-2-2 timings (SPD)
- Taisol 760 cooler
- IBM 61.4GB Hard Drive
- Pioneer 105 DVD/RW
Software
- Windows XP Professional Build 2600.xpclient.010817-1148
- nForce 2.0 drivers
- SiS 3.08 drivers
- ATi CATALYST 3.1 drivers (6292)
- 3DMark 2001SE v330
- UT2003 Demo
- Comanche 4 benchmark
- Serious Sam 2 Demo
Notes
I'll be comparing the Xabre 400 to a card that's vying for your attention in the same price bracket. The Xabre 400 retails for around £60 and the Radeon 9000 can be had for roughly the same price from various online vendors. The Radeon 9000 is a little ahead of the SiS offering in the specification stakes. Offering Vertex Shader 1.1 and Pixel Shader 1.4 by hardware, it's a little more futureproof. We must note that it has the same 4 pixel pipelines as the Xabre 400, so its fillrate is the same at 1GPixel/s for single-texturing instances, but its multi-texturing fillrate is halve that of the Xabre's 2GTexels. It only has one texture unit per pipe compared to the Xabre's two. Lastly, the Xabre has a little more bandwidth than the Radeon's 6.4GB/s, but ATi's bandwidth-saving techniques are extremely efficient. This should be a fairly even match up given the respective specifications of both budget-based cards.
Benchmarks will be conducted at 1024x768x32, 1280x1024x32, and 1600x1280x32 to give us an idea of how the cards cope with varying degrees of fillrate limitations. We'll be benchmarking the Xabre with both the TexTurbo set to the default 3 and it set manually to 0. If you want to compare like for like, I'd suggest looking at the TexTurbo 0 results.
Using the supplied overclocking tool on the driver CD, we managed to get the DFI stable at 280MHz core and 265MHz memory. Given that 4ns RAM was supplied, we had hoped to go a little higher. Your overclocking mileage may vary considerably. You shouldn't infer that every card has the same potential.
