Well, let’s take a good hard look and see if there really is a difference between the KT266 and KT266A. I took a standard production Soyo K7V Dragon that uses a VIA KT266 chipset, and then a “new” KT266A based Soyo K7V Dragon Plus that uses the KT266A chipset, both with standard production BIOS’s. I found the optimal BIOS settings for both, installed an Athlon MP 1.2GHz CPU (which should be the same as a new Athlon XP), 256MB of PC2100 memory, a WD 200BB ATA/100 7200 RPM hard disk, 3.5″ floppy disk, a 10X DVD drive and installed Windows 98 along with any current drivers and my test programs. I then tested them both, changing nothing other than the mainboard. Since the ONLY thing that changed was the chipset, it should give a good valid comparison between the KT266 and KT266A chipsets.
So what are my expectations? Well, as I noted, I read the reviews on a number of different sites and did a bit of number crunching (throwing out some of the really outrageous and inconsistent scores). I came up with this: The KT266A should give about a 3~6% gain over a KT266. It will be interesting to see how close I came. On thing to remember, the variance of the test run is 3%, so a gain of less than that is really meaningless. As a rule of thumb, I prefer to use 5% as a minimum for showing any real improvement and 10% as a gain that a end user might actually even notice.
Now, I saw some really wild variance in the published results when it came to office type applications benchmarks, one may have even been close to 20%! But others were much less, so I tended ignore that one high one. Let’s look at those benchmarks first.
Gee, looks like I was right on the money for both Winstone and SYSmark 2000. You may wonder why I used SYSmark 2000 instead of 2001? Well, it was so I could show you this (not given in the SysMark 2001 results):
Interesting to see which applications in particular that show a gain and by how much. But only one comes even close to that magic 10% number I’m looking for. Now the question is if gaming type applications will give the same results.
Not quite, these two game based benchmarks actually show almost twice as much gain with a KT266A over a KT266 vs. the business applications. It still isn’t up to that magic 10% number, but 6~7% is not too bad.
You may have noticed I have not included any actual memory benchmarks anywhere here. There is a reason – over the years I have found that while doing so can be fun and can help identify why two different systems perform differently, what I really care about is how it relates to actual system performance. All that said, I do still like to run the Winbench 99 CPU mark. It not only tests CPU performance but also main memory.
Not a large percentage, but it does show that the KT266A does have better memory performance than the KT266 does. To me that sort of result is much more meaningful than looking at different individual memory operations. Remember that since the CPUmark 99 includes testing more than just the memory, the actual gain in memory performance is more than 3.57%.
The 3D Winbench CPU test is similar to Winbench, and it seems like there is a bit of a trend when you look at all the results – 3D and Multimedia scores seem to show a higher percentage gain on a whole than the business applications do.
Be the first to discuss this article!