SiS Xabre 600

Date: 01.12.2002

Tests

We did the first test on a quite powerful PC of the following configuration:

Processor Intel Pentium 4 2.8 GHz;
Memory RDRAM PC800, 512 MB;
Motherboard Intel 850EMV on the i850E chipset
Sound card Creative Sound Blaster Audigy;

While choosing the test setup, we were guided by the intention to test both the video card's potentials and the declared Vertexlizer Engine which if used on a powerful system should demonstrate a substantial advantage of the vertex shader block. As far as the tests are concerned, we used the following benchmarking suites:

3DMark 2001 SE Pro (Direct3D, synthetic test on the base of the MaxFX core)
Codecult Creatures Benchmark (Direct3D, a complex test that efficiently involves the AGP part)
Unreal Tournament 2003 Demo (Direct3D, real-life stress test benchmark);
Return to Castle Wolfenstein (OpenGL, on the Quake 3 engine).

Let's consider performance using the synthetic benchmarks: The first in the queue is 3DMark 2001 SE:

As the resolution goes up, the Xabre 600 starts taking a noticeable lead over Xabre 400. We'll analyze what contributes to this through comparison of separate components of the suite:

First off, note the fill rate which is one of the most important parameters especially for the old games having no support for shaders, T&L and other new technologies, that is games which shift solely the rasterization onto the accelerator.

At the multi-texturing speed with and without it, the Xabre 600 is much faster than Xabre 400. On the average, the card displays approximately a 20% performance boost. We believe this is related primarily to the increased chip and memory speeds, that is, through overclocking the Xabre 400 you might get some similarity in results for this test.

Another important benchmark is the vertex shader performance. As you remember from the introduction to this review, SiS claimed to have added the Vertexlizer system into Xabre 600, which is meant to distribute vertex shader loads between the graphics chip and the CPU. Unlike the pixel shaders, vertex shaders can be quite easily handled by the CPU. In the end, this gave a not too bad result after all - the speed of handling the vertex shaders in Xabre 600 is often much higher than in GeForce 3. Subsequently, we'll try to prepare a detailed research on Xabre 600 vertex shader performance vs. the processor used.

The last noteworthy benchmark is Nature in which both vertex and pixel shaders as well as many other secondary rendering technologies are in abundance. Evidently, Xabre 600 is not strong enough to contend with nVidia's high-end products, but it does make a big difference from the past generation products already at the start, and as the resolution goes up, the performance difference may be as high as 50%.

Now on to the Codecult benchmark. This is an extremely resource-hungry benchmark, very critical to the memory capacity and other system characteristics. If you look at the upper graph of common results, you can't help seeing that it's the 128 MB video cards that are ahead of the pack at performance. Specifically to Xabre 600, there is an evident lag past GeForce 3. We won't bring in the benchmarking results for Xabre 400 any further, since they are visibly worse, and the benchmark runs extremely slow with this card. In this connection, it's worth recalling a recent scandal when SiS was blamed for overstating 3DMark 2001 scores. The experiment was very simple - colleagues from the iXBT site removed any mention of the 3DMark test, renamed the *.exe file and started tests. The results were much lower than those produced at the original test setup, which led to a conclusion that SiS had somehow modified the driver to get higher scores on this popular benchmark.

Finally, the gaming benchmarks - Unreal Tournament 2003 Demo and Return to Castle Wolfenstein. Here you can't cheat the counter that easy. There is an interesting evidence - in RTCW, the card in some scenes attains a 3-fold (and even higher) lead over Xabre 400, especially at higher resolutions, but in Unreal Tournament it's not much better at FPS than Xabre 400.

Wrapping it up with the benchmarks, it's worth noting that things differs greatly in different games and benchmarks. We can claim with confidence that Xabre 600 is faster than Xabre 400, and in most benchmarks the attained lead is quite substantial. Along with that, at performance the card sometimes is close to the much more expensive GeForce 3, in some benchmarks it showed even a bit better than GF3. This is a definite bonus for a budget product.

Findings

Hardly can we call Xabre 600 a new product. This is more likely an evolutionary step forward from the Xabre 400, which paves the way for the brand-new Xabre II. The software-driven vertex shader handling is of a certain interest, but only tests of commercial versions of the boards on debugged drivers will tell how justified it is. The good performance coupled with the low price gives every reason to believe that Xabre 600 may be a real contender to nVidia's and ATi's budget solutions.

Pros:

Excellent 2D quality;
Support for two monitors;
Operational stability;
Low price.

Cons:

Outdated switching between TV standards;
Illegible captions under the jumpers;
No thermopaste under the cooler.

SiS Xabre 600

Tests

Findings

Read more on this topic: