ECS K7S6A with SiS745 in the test: The successor to the SiS735 is here
- 1 Preface
- The chipset: SiS745
- Sense or nonsense: DDR333?
- 2 Scope of delivery and documentation
- 3 Equipment
- 4 Stability
- 5 Test system
- The agony of RAM choice
- 10 Conclusion K7S6A
As always, SiSoft Sandra 2001 should be the Form an introduction to our benchmark course. However, it has often been found in the past that the pure streaming performance shown here does not always necessarily have to represent the subsequent performance graduation. Sandra largely ignores the latency times, which are extremely influential in many applications.
The previous highs of the nForce415 in 128-bit dual-channel operation are still untouched in front. And the KT266A can clearly stand out against the K7S6A in DDR333 and 266 operation. The different operating modes, on the other hand, are rather few. TheFPU performance is visibly unimpressed by sharper timing or the difference between 133 or 166Mhz RAM clock. Only in the ALU rating does the K7S6A easily stand out from the actually 'faster' DDR333 with its conventional DDR technology (266Mhz) and fast timing. A first sign that neither the CPU is using the theoretical bandwidth, nor can the RAM module, which is switched extremely slowly for reasons of stability, achieve new top values.
SiSoft Sandra 2002
For the first time this time, we also included the current version from SiSoft in our benchmarks. However, since we could not use any comparison values here, the K7S6A only had to sound out the internal competition for the time being.
Here the result looks a bit gloomy for the DDR333 configuration. Sandra 2002 seems to pay more attention to faster clock cycles than the 2001 version. Accordingly, the 'PC2700-K7S6A', which is running on the low-timing flame, is slightly behind.
The packer had WinACE We already had one or two surprises and so we were excited about the first practical result this time as well. In addition to the pure streaming capability, short latency times are particularly important here. The K7S6A should get some problems here in DDR333 operation. We chose around 360MB bitmaps, documents and already compressed MP3s as the 'packing template'.
And howexpected. The performance of the K7S6A with our DDR333 bar turns out to be sobering. At 11:22, the board is almost three minutes behind the competition from KT266A and nForce415 with ultra-fast-timed DDR266. And its own competition with supposedly slower RAM also takes a good two minutes from the board.
How much the settings of the RAM timing affect the overall performance in memory-intensive applications, WinACE should have presented more than impressively. The assumption that current modules do not even come close to the potential of the PC2700 technology seems to be true one hundred percent. Whether the K7S6A has an allergy to the Apacer bar and whether other manufacturers' products might have done a little better, we cannot exactly determine, due to the lack of comparison options. What is certain, however, is that there can be no talk of a guaranteed additional performance of a DDR333 system. The theory that the manufacturers have thrown both chipsets and RAM prematurely on the market for marketing reasons and at the expense of 'true speed' (timing) seems more plausible than ever.
With the help of the Lame 3.91 encoder, a 154MB wave file (For the statisticians: Tchaikovsky - Capriccio Italy) in the Audiograbber 1.80 had to be converted into an MP3 with a variable bit rate using the MTRH method. The other settings were 'joint stereo' and a high quality.
The second 'packer' also revealed a similar picture. None of the RAM configurations of the K7S6A can do anything to the crown of the DDR266 chipsets KT266A and nForce. At 1:18 minutes, the Elitegroup mainboard with 256 and 512MB DDR266 is four resp.nine seconds behind the competition. The K7S6A with 256MB DDR333 again at the end of the field.
This rather compact benchmark is based on the ray tracing program Cinema 4D and gives a good overview of the rendering properties of the CPU chipset memory subsystems.
All in all, no surprise here either. The order remains unchanged. The alleged leader DDR333 is only in last place thanks to an immature module. Depending on the load on the graphics card, the gap to the fastest board is up to 6%.
On the next page: Quake3Arena