SIS Xabre400/600 Based Video Cards Roundup

Part 1. November 2002:
Inno3D Xabre400
PowerColor Evil Xabre400 
Chaintech Xabre400
Soltek Xabre400


  1. Peculiarities of the video cards
  2. Test conditions and drivers, 2D quality 
  3. Optimization of the Xabre for 3DMark2001 
  4. Information on performance of such video cards from 3Digest 
  5. Conclusion

The fall that seized the Europe is now handing it in to the winter. The earth is still conquering the white attack, but soon it will surrender. Pockets of potential buyers of hi-end accelerators are also resisting losing their weight and tremble at the thought of the upcoming NV30, and ATI is said to be preparing something new able to stand against the future 3D king. 

Although the potential of High-End accelerators are intersting for many, not everyone can afford such card, and there are also quite wealthy people who assume that it's insane to spend $400-500 for a video card and it's better to get a gaming console with a decent bunch of games. There are not more than 5-6% of users out there who might buy such an expensive card, while the most will go with solutions priced at $80-100. Therefore, we'd better to have a look at new-comers from this price niche. Some of them were already reviewed, they are NVIDIA GeForce4 MX and ATI RADEON 9000 (Pro) based cards. But there is another long-liver on the market, - Xabre and, in particular, Xabre400, from SiS. 

Theoretical materials and reviews of video cards which concern functional properties of the SIS Xabre400/600 GPU

Today we will analyze some quirks and interesting aspects of operation of the Xabre and cards based on it. 

The companies the cards are coming from are well known to all of you, that is why we are heading directly into the analyses of the products. 


The cards have AGP x2/x4/x8 interface, two come with 64 MB and the Chaintech's and C.P.Technology's solutions come with 128 MB DDR SDRAM located in 8 chips on both PCB sides. 
Inno3D Xabre400 

PowerColor Evil Xabre400 

Chaintech Xabre400 

Soltek Xabre400 

Inno3D Xabre400 
Hynix memory, 4ns access time. The card operates at 250/250 (500) MHz. 

PowerColor Evil Xabre400 
Samsung memory, 4ns access time. The card operates at 250/250 (500) MHz. 

Chaintech Xabre400 
Samsung memory, 4ns access time. The card operates at 250/250 (500) MHz. 

Soltek Xabre400 
EtronTech memory, 3.5ns access time. The card operates at 275/250 (500) MHz. The chip's speed has reached the Xabre600's level, though it's not advertised. 

Comparison with the reference design, front view
Inno3D Xabre400  Reference card SIS Xabre400 

PowerColor Evil Xabre400 

Chaintech Xabre400 

Soltek Xabre400 

Comparison with the reference design, back view
Inno3D Xabre400  Reference card SIS Xabre400 

PowerColor Evil Xabre400 

Chaintech Xabre400 

Soltek Xabre400 


All the PCBs look similar and the design is almost identical to the reference one. The cards differ only in color and memory size. 

However, there is one more interesting point that distinguishes the Soltek's card: its chip works at 275 MHz instead of standard 250. If the memory worked at 275 (550) MHz, we would have a normal Xabre600. However, the chip's increased speed didn't boost up the performance with the memory working at its standard frequency, but it will be touched upon later. 

The cards have also different coolers and accessory packs. 

Here are the cooling devices: 

Inno3D Xabre400 
The most unsophisticated design. An ordinary cooler sitting on thermo glue instead of clips. 

Chaintech Xabre400 
Such cooler could be seen on GeForce4 Ti 4200 based cards; the developers used it here as well, to some reason, though the chip heats up not much and the memory doesn't need to be cooled down, especially considering that the memory on the back is not released from the heat at all. 

Soltek Xabre400, PowerColor Evil Xabre400 
Such heatsink is typical of ATI RADEON 9000 Pro based cards. It's also a classical solution, but it's more efficient than the cooler on the Inno3D Xabre400. 


The coolers are hiding the Xabre400 GPU itself: 

As I mentioned lots of times, such cards support TV-out via the SIS 301 coprocessor which is mounted on nearly all Xabre based solutions. TV-out quality is discussed in the first review of the list above. 

Now I'm going to take a peep into the boxes. 

Inno3D Xabre400 
User Guide, CD with drivers, WinDVD, S-video-to-RCA and DVI-to-d-Sub adapters, RCA extender. 

PowerColor Evil Xabre400 
User Guide, CD with drivers, PowerDVD, S-video-to-RCA adapter, RCA extender. 

Chaintech Xabre400 
User Guide, CD with drivers, WinDVD, S-video-to-RCA adapter, S-Video extender. 

Soltek Xabre400 
User Guide, CD with drivers, PowerDVD, S-video-to-RCA adapter, RCA extender. 


The video cards ship in retail packages. 

Inno3D Xabre400 
The package is typical of Inno3D. 

PowerColor Evil Xabre400 
This box is also peculiar to C.P.Technology, but it's made of thick cardboard. 

Chaintech Xabre400 
The designers tried to attract attention of users with bright colors, but a picture of some alien with a great deal of polygons is really over the head of Xabre. 

Soltek Xabre400 
Finally, greet the most stylish box which looks perfectly and underlines of the origin of Xabre (Sabre). 


That's all about peculiarities of the cards. 


Inno3D Xabre400  250/500 -> 275/540 MHz 
PowerColor Evil Xabre400  250/500 -> 273/548 MHz 
Chaintech Xabre400  250/500 -> 280/550 MHz 
Soltek Xabre400  275/500 -> 300/560 MHz 

Look at the Soltek's card: well, the core seems to be doing well, and if the memory had a good potential as well, the core's speed could be much more beneficial. 


    • in course of overclocking you must provide additional cooling, in particular, for the card (first of all, for its memory):

    • overclocking depends on a definite sample, and you shouldn't generalize the results of one card to all video cards of this mark or series. The overclocking results are not the obligatory characteristics of a video card.

Test system and drivers


  • Pentium 4 2530 MHz based computer: 
    • Intel Pentium 4 2530 MHz; 
    • ASUS P4T533 (i850E) mainboard; 
    • 512 MB 32bit RDRAM PC4200; 
    • Seagate Barracuda IV 40GB HDD; 
    • Windows XP. 
  • Athlon XP 2000+ based computer: 
    • AMD Athlon XP 2000+ (1666 MHz); 
    • Soltek 75DRV5 (VIA KT333) mainboard; 
    • 512 MB DDR SDRAM PC2700; 
    • Seagate Barracuda IV 40GB HDD. 
The test system was coupled with ViewSonic P810 (21") and ViewSonic P817 (21") monitors. 

In the tests we used SiS's drivers of ver.3.06.55. S3TC was off in the applications. 

Test results

Before we start examining 2D quality I should say that there is no a complete technique of objective estimation of this parameter because: 

  1. Almost all modern 3D accelerators can have 2D quality much dependent on a certain sample, and it's impossible to trace all cards; 
  2. 2D quality depends not only on a video card, but also on a monitor and a cable; 
  3. Besides, certain monitors do not get along with certain video cards. 
As for the tested samples, together with the ViewSonic P817 monitor and BNC Bargo cable the cards showed excellent quality at the following resolutions and frequencies: 
  • Inno3D Xabre400: 1280x1024x85Hz, 1024x768x100Hz; 
  • Chaintech Xabre400: 1280x1024x100Hz, 1024x768x100Hz; 
  • Soltek Xabre400: 1280x1024x120Hz, 1600x1200x75Hz, 1024x768x120Hz; 
  • PowerColor Evil Xabre400: 1280x1024x100Hz, 1024x768x100Hz; 

Now I'm going to estimate operation of the cards in 3D. This part of the tests consists of two subsections. In the first one we will touch upon some aspects of operation of the Xabre in the 3DMark2001: 

  • 3DMark2001 SE Pro (MadOnion/Remedy) - DirectX 8.1, Hardware TCL, multitexturing, LOW Details, DXTC OFF, double buffering, 24-bit Z buffer
Here I'm going to show you that the performance level of the Xabre in this popular packet is far-fetched. One of our readers did an interesting work on 3dmark2001se.exe having removed all remarks on "3DMark2001" and replaced it with 1DMark1913. So, such software won't recognize 3DMark in it. And here is what we obtained when dumped the Xabre400 64MB through 1DMark1913.exe. 

Game 1, Low details

It's obvious that the Xabre's software is connected to the word 3DMark and it's optimized for it. Without it the performance falls down by over 30%. But even at a such drop the Xabre is still able to fight against the RADEON 9000. 

Game 2, Low details

Although the speed's drop is smaller in the test with the Xabre being optimized for the 3DMark on, without the optimization the Xabre becomes an outsider. 

Game 3, Low details

The same situation. 

Game 4

Here, the Xabre doesn't perform well even without the 1DMark1913. The speed of operation with shaders has a long way to go. 

3D Marks

In general, the speed goes down by a bit over 20% without the optimization for 3DMark. It's quite much. Why does it take place? Let's compare the quality. 

1DMark1913 3DMark2001
Game 1

Game 3

Game 4


Although the difference could be unnoticed at fight glance, they play with light (it's well seen in Game4). In the 1DMark1913 the processing is complete, and in the 3DMark2001 the drivers treat the lighting in a specific manner. Hence the speed gain. Note that the tests in the 1DMark1913 were also carried out with the NVIDIA and ATI based cards, but there were no noticeable changes. All the tests were fulfilled in the turbo texturing mode of the Xabre where the pictures have quality inferior to the rivals. What happens when this mode is off is described in the previous article. 

When the patch for the 3DMark2001SE was announced (which was meant to improve operation with the Xabre), after the tests on the version 3.30 we wondered why the speed grew up on the same drivers in all the tests of the 3DMark2001SE? And now we can see how easily test packets and drivers can adjust. The history of such optimizations goes back to 1998 when ATI released its turbo drivers for the Rage Pro which marginally increased the performance in the 3D WinBench, and after that the packet started losing its popularity. I'm afraid the MadOnion and its 3DMark can follow it... 

And now take a look at the summary diagrams from our 3Digest so that you can have deeper understanding of how such cards perform. 

The overclocked cards are marked with red color, the sign o/c (overclocked) is followed by the frequencies reached.

Well, all the video cards worked stably in the tests, no failures, and we had no complains about their operation (certainly, we cannot guarantee operation of all cards in all games within the test suite we are traditionally using due to the "physical bounds"). And all the problems of the Xabre are mentioned in our monthly 3Digest


Before we arrive at the final conclusion let's take a look at the prices of the Xabre400 cards: $75 on average. In comparison, the RADEON 9000 is priced at $65, and the RADEON 9000 Pro at $92. The Xabre, being much more efficient, doesn't support anisotropic filtering and vertex shaders (which means a defeat in UT2003-type games), has 3D quality lower than that of other cards (due to turbo texturing), and it's dearer. However, the rumor has it that some Xabre400 solutions are already shipping somewhere at $40 in retail. I assume that at about $50 the Xabre400 based cards will find their market, especially because of the excellent TV-out. But right now, I can't recommend them at $75. Remember that we haven't mentioned yet AA quality (see the next 3Digest, a new section on the AA). 

  1. Inno3D Xabre400 - this card has not many peculiarities and it's very similar to the reference solution. Note that it comes supplied with a DVI-to-d-Sub adapter. 

  3. PowerColor Evil Xabre400 - it is also based on the reference design, though it comes complete with 128 MB memory. However, such cards don't need so much memory as the speed is primarily limited by the chip's capabilities and the memory's speed. 

  5. Chaintech Xabre400 - it also has unnecessary 128 MB memory which just affect the price. The set of accessories is as modest as that of the rest. 

  7. Soltek Xabre400 - its chip's working at 275MHz, but it brings small benefits. Besides, its speed is lower to some reason than that of the rivals, probably because of the timings of the EtronTech's memory. I don't see why they have mounted 3.5ns memory as it provides no leeway for overclocking and affects the price. Note that this card, as well as its box, is the most attractive. 

In our 3Digest you can find complete comparison characteristics of video cards of this and other classes. 


  • Satisfactory performance in 3D graphics (though the RADEON 9000 is close on the heels having richer 3D features and a lower price); 
  • Excellent build quality of all the cards; 
  • Good overclockability of the Soltek's card; 
  • Perfect TV-out of all the cards. 
  • Imperfect drivers and easy to detect optimization for the 3DMark2001SE; 
  • Inferior image quality in 3D at the expense of turbo texturing; 
  • Anisotropy unsupported; 
  • Overpriced (the price doesn't match the capabilities of the Xabre400). 
Andrew Worobyew (