So, what can we really conclude from all of this? Unfortunately it is not quite as informative as I had hoped it would be. The limitations of the Winstone and Content Creation benchmarks makes the data useful only for a specific type of computer usage. However, we can come to the following conclusions with some degree of certainty.
- There are apparently some situations where putting the fastest drive as the boot drive will not result in the best system performance, particularly where the user is more concerned about application wait time than boot or load time (and the system is configured properly). This is something for vendors and technicians to keep in mind when dealing with a customer with limited funds for upgrading. One needs to clearly identify how the system is being used, and configure the system accordingly.
- If building a system from scratch, the best performance can be achieved by using two fast drives rather than a single drive by configuring the system as suggested in Organizing Your Windows System.
- RAID 0 does not appear to provide any significant benefit vs. just installing two fast drives and spreading the access across them with proper system configuration. One caveat to this is that since we cannot accurately measure application load time with the benchmarks used, there may be a benefit with RAID 0 in that sense, but there isn’t any solid evidence to prove it at this point. There are also some drawbacks to RAID 0, including additional cost and inability to separate data from OS/applications for ease of maintenance.
- If we can assume these two benchmarks emulate normal usage with regards to application mix and resource usage, there seems to be absolutely no benefit in putting two IDE drives on separate channels. There may very well be situations where bandwidth is an issue, but usually that will include DVD or CD-RW devices, which probably would be better off using a separate IDE channel from the boot and data drives anyway. This could make cabling a bit easier, since hard drives are typically physically close to one another, while other devices may be farther away.
The limitations of the tests do not provide any information about whether a system with insufficient RAM would benefit by putting the swap file on a secondary drive, or whether that would interfere with the operation of the applications. As mentioned at the outset, these results do not reflect any performance impact on boot time with the various configurations, while application load time is only minimally tested and cannot be broken out from the operational performance.
One complaint that I am almost certain to get is that I did not use more benchmarks. While I have SysMark 2000 and 2001 available, there is insufficient information available to determine what these tools are really doing. This would make it much more difficult to interpret the results in any meaningful way. I honestly don’t really know how anyone is able to come to any reasonable conclusions about various components without this background information, but perhaps they are much more knowledgeable than I. I also must admit that I simply do not play many games on the computer, so I would be extremely dishonest if I tried to run such benchmarks and assume I understand how gamers use their systems.
On the other hand, part of the reason for this article is my extreme interest in benchmarks in general and the results that have come from some very stringent testing for other purposes. As I gather information about the various benchmarks, and run these same tests with them, we may gain some additional insight not only into how a system can be configured better, but what kind of information we can really tell from the various benchmarks that are available.
Be the first to discuss this article!