Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Business Benchmarks
In my last look at productivity benchmarks, it became obvious that using Winstone to measure performance has a few tricks. Winstone is very disk limited, and not all south bridges are created equal. In order to get around this, I ran each benchmark twice, once with a the normal South Bridge, and once using a Promise Ultra-100 PCI IDE controller. Doing this lets me isolate the I/O “fog factor” – because its the same for all players, and we can see the effect of the memory/AGP controller on system performance. But remember, the overall performance indicator is achieved by using the motherboard’s I/O controller, unless of course you are going to run out and buy an IDE controller to use. I will refer to these two graphs as Normalized (using the Promise Ultra-100) and System (using the on-board IDE)
Normalised Performance
| KT7A Default | KT7A | KA266 | 8K7A | |
| Average | 56.8 | 57.8 | 59.2 | 57.4 |
| Minimum | 56.3 | 57.2 | 59.1 | 56.4 |
| Maximum | 57.5 | 58.3 | 59.3 | 58.4 |
It would appear that Business Winstone is not influence by memory latency or bandwidth at all. There is very little in it. And this really is as expected.
| KT7A Default | KT7A | KA266 | 8K7A | |
| Average | 67.1 | 73.5 | 70.0 | 70.7 |
| Minimum | 67.0 | 73.2 | 69.7 | 70.5 |
| Maximum | 67.4 | 73.7 | 70.5 | 70.9 |
The tweaked boards eke out a tiny gain in performance, but not much. Again it doesn’t appear that the tweaks are doing much for performance. This is as expected. Remember that the caveat I expressed was that most (>90%) of all memory accesses are satisfied from cache.
Overall System Performance
As I mentioned previously, the I/O performance plays a significant role in Winstone scores. So the scores here are to determine how the motherboards perform as a system. In an update to my KA266 vs. KT7A article, I measured and commented upon the poor performance of the ALi drivers under Windows 200. It’s not great, but VIA does appear to be stealing a march on ALi.
| KT7A Default | KT7A | KA266 | 8K7A | |
| Average | 51.5 | 55.4 | 53.6 | 55.1 |
| Minimum | 50.9 | 54.7 | 53.2 | 54.8 |
| Maximum | 52.2 | 55.9 | 53.9 | 55.4 |
| KT7A Default | KT7A | KA266 | 8K7A | |
| Average | 66.9 | 70 | 67.2 | 66.7 |
| Minimum | 66.3 | 69.6 | 66.2 | 66.0 |
| Maximum | 67.2 | 70.2 | 67.9 | 68.4 |
To put things in perspective, I’ve included two graphs that illustrate the performance gains from using the Promise Utra-100 controller versus the on-board IDE. Note: these graphs aren’t to scale!
In both cases the VIA 686B equipped KT7A and 8K7A gain about the same performance boost, around 4 Winstone points. The ALi gains much more in the Business software scores, which we suspect to be more I/O bound. Content Creation scores gain only half that amount.
In the wash though, there isn’t enough performance difference to choose a winner.
Pages: « Prev 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Next »
Be the first to discuss this article!