ECS K7S6A with SiS745 in the test: The successor to the SiS735 is here

ECS K7S6A with SiS745 in the test: The successor to the SiS735 is here

Test system

The test system still consists of well-known components. As a CPU, the XP 1700+ provided to us by AMD does its job again, the graphics work is done by the Inno3D Tornado GeForce2 Ultra. The drivers remained unchanged. It was only the RAM that had to be given a lot of attention this time. More on this a little further down in the section 'Spoiled for choice'.

  • Processor
    • AMD Athlon XP 1700+
  • Motherboard
    • K7S6A (Revision 1.0)
    • Asus A7N266-C (Revision 1.03; Bios 1001.B)
    • EPoX 8KHA + (Revision 2.0; Bios 2.11. 2001)
  • RAM
    • 2x 256MB PC2100 Apacer DDR RAM CL2
    • 1x 256MB PC2700 Apacer DDR RAM CL2,5
  • Graphics card
  • Peripherals
    • Western Digital Caviar AC310100B
    • Teac 532E
  • Other
    • Logitech Mouse Man Wheel Optical (USB)
    • Microsoft Natural Keyboard Pro (PS2)
  • Software
    • Windows 2000Professional ServicePack 2
    • nVidia Detonator 23.11
    • Via 4in1 Version 4.37a
    • nForce UDP DriverPackage 1.0
    • SiS Driver Package 1.09
    • DirectX 8.1


The identical picture for the selected benchmarks. Except for SiSoft Sandra 2002, we have not considered any new features.

  • DirectX 7.0
    • 3DMark 2000 1.1
    • Unreal Tournament 4.36 (utbench)
  • DirectX 8.0
    • 3DMark 2001
  • OpenGL
    • Quake3Arena (Q3bench)
    • Vulpine GL Mark
  • Memory performance
    • SiSoft Sandra 2001/2002
  • Other
    • WinACE 2.03
    • Cinebench 4D
    • Lame 3.91

The agony the RAM choice

But even before the first benchmark, as we had feared, the memory used was a major hurdle in our way. First and foremost, the fact that we had carried out the last mainboard tests with 512MB Ram and this time unfortunately only a 256MB DDR333 module (Apacer with Nanya chips, CL2.5) was available to us was less serious but still disturbing. However, so that we can continue to include the previous test results in the evaluation, we have given the board for the comparison PC2100 to PC2700 with 256MB each and for the comparison with the competing boards with 512MB PC2100. The second problem should prove to be during the first fine adjustment of the memory timing of the DDR333 RAM. Because all negative comments about the current PC2700 modules seemed to be true. The computer booted with an FSB of 133MHz and a RAM clock of 166MHz only with the 'Safe Mode' settings and the slowest settings that the BIOS provided. The tightening of just a single variable made the PC linger in an endless loop before the BIOS prompt.

The something older Apacer CL2DDR266 RAM bars with Nanya chips, however, caused fewer problems. The individual module could be operated with the sharpest timing with almost no problem. Only the setting CL2 was vehemently denied by the board. A known problem when rummaging through the relevant forums. When we put a second, identical module on the board for the test with 512MB, this time the 'RAS Active' set to five turned out to be a hurdle. Here the K7S6A only wanted to boot again when the timing was set to the slower value six. Not a good omen for the apparently somewhat overwhelmed memory controller.

The following table provides information about the best possible memory timings we use.

256MB DDR333 256MB DDR266 512MB DDR266 clock 166Mhz 133Mhz 133Mhz Profile Safe Mode Ultra Mode Ultra Mode RAS Precharge 4 2 2 RAS Active 7 5 6 CLAS Latency 2,5 2,5 2,5 Refresh Cycle tRAS + tRP + 1t tRAS + tRP + 1t tRAS + tRP + 1t

As you can easily see, the K7S6A with DDR333 memory was not allowed to throw itself into the course with the most optimal settings. Against the competition running on absolute ultra settings (8KHA +, A7N266-C) it should certainly not be easy for the board.

The SiS745 on ECS K7S6A
The SiS745 on ECS K7S6A

On the next page: SiSoft Sandra