System setup, IQ, and notes
Hardware
- S3 DeltaChrome S8, 256MB, AGP8X, 300/600
- ATI Radeon 9800SE AIW, 128MB, AGP8X, 380/680
- AMD Athlon 64 3400+, 2200MHz
- EPoX 8HDA3+ S754 VIA K8T800 Motherboard (12/12 BIOS)
- Corsair XMS3500C2, 2 x 256MB, 2-2-2-6 @ DDR400
- Samcheer 420w PSU
- AMD reference cooler
- IBM 120GXP 40GB Hard Drive
- Dell P991 19" flat-faced CRT monitor
Software
- Windows XP Professional w/SP1
- S3 6.14.100.1597-15.06.33.4 drivers
- ATI CATALYST 3.9
- VIA Hyperion v4.51 chipset drivers
- DirectX 9.0b Runtime
- 3DMark 2001SE v330
- 3DMark03 v340
- AquaMark 3
- Splinter Cell (Beyond3D Demo)
- Unreal Tournament 2003 Retail (patched up to 2225) (HEXUS custom benchmark)
- X2: The Threat Rolling Demo
- Quake III PR1.30
Notes
S3's midrange DeltaChrome will be compared to another card that supplies much more than 3D grunt. The Radeon 9800SE AIW is perhaps the finest current example of a card that can do it all at a palatable price. The 9800SE's 380/680 engine and memory clocks don't quite tell the whole picture. It's a strict 4x1 setup with a 256-bit memory bus, so it'll possess inferior fillrate capabilities but should eke out a bandwidth advantage over our test DeltaChrome S8. The S8's 256MB of onboard memory appears to be, like most midrange cards, an attempt to woo the customer. We've previously shown that it is only beneficial on cards with huge memory bandwidth.
2D quality
2D quality was excellent. Clarity at 1600x1200 @ 75Hz was as good as anything that I've personally tested from ATI or NVIDIA. 3D accelerators spend most of their time in standard 2D mode, so having excellent image quality is a must, especially if one of the main intended uses lies in media work. We'd have no qualms in using the S8 on a day-to-day basis. Compare this is with awful 2D from, say, the Intel Extreme Graphics 2.
Noise
The small GPU fan's sound profile will most likely be masked in systems that employ more than one cooling fan. The test system used a quiet CPU and PSU fan and no further fan-assisted cooling. The DeltaChrome's fan could barely be picked out over the system's sound. We'd not be surprised to see this model use a large, passive heatsink.
3D image quality
3D quality used to be barely considered until it became apparent that drivers weren't necessarily sticking to the reference image quality. Both NVIDIA and ATI have been guilty of putting benchmark performance above pure IQ. Increasing gaming performance whilst not degrading IQ can be achieved via the use of optimised drivers, but we're now left to examine screenshots in an attempt to verify that no driver trickery has been used. ATI's DX9 implementation and driver set are both good, so we sought to compare the DeltaChrome's standard image quality against the 9800SE AIW's. AquaMark3 and Splinter Cell were used.
- DeltaChrome S8 - 1024x768 no AA or AF (click to enlarge)
- 9800SE AIW - 1024x768 no AA or AF (click to enlarge)
And now Splinter Cell.
- DeltaChrome S8 - 1024x768 no AA or AF (click to enlarge)
- 9800SE AIW - 1024x768 no AA or AF (click to enlarge)
The images have undergone JPEG compression but show little or no deviation from one another. Having carefully checked a few other games it appears as if S3's default image quality is reasonably good. We found a few visual artifacts in UT2003 and X² - the latter benchmark often appeared to shimmer the image heavily in certain places. 3DMark03's visual quality didn't appear to be as good as ATI's, too.
Testing
Our biggest compliant is with the lack of OpenGL adjustment. Antialiasing, too, only worked at 1024x768. Higher resolutions would cause the benchmarks to fall back to the desktop. Anistropic filtering, however, performed as expected. In view of the fact that AA was limited and our usual testing is to place a heavy image enhancement burden on the cards (4x FSAA and 8x AF at 1024 and 1600 resolutions), we decided to forego FSAA / AF analysis. We hope that a revised driver set will offer options comparable to ATI and NVIDIA cards. In view of this, benchmarks were carried out at 1024x768, 1280x1024, and 1600x1200, respectively.
