In an article on its Web site datelined yesterday, Computer World reports that Devil Mountain Software, a software firm based in Florida, is claiming that its XPNet survey of over 23,000 PCs shows that Windows 7 PCs are suffering from excessive use of memory (RAM) by the system:
Citing data from Devil Mountain Software’s community-based Exo.performance.network (XPnet), Craig Barth, the company’s chief technology officer, said that new metrics reveal an unsettling trend. On average, 86% of Windows 7 machines in the XPnet pool are regularly consuming 90%-95% of their available RAM, resulting in slow-downs as the systems were forced to increasingly turn to disk-based virtual memory to handle tasks
Now, this sounds pretty bad, especially since, according to the company’s figures, only 40% of PCs running Windows XP are experiencing “memory saturation”; furthermore, the typical machine in the sample running Windows 7 has 3.3 GB of memory, compared to 1.7GB for Windows XP machines.
I have not been able to track down an analysis that describes these how these figures are measured. But I do want to caution you about jumping to conclusions. The company does say that it does not know whether the increased memory usage is due to the activity of Windows itself, or to greater usage by applications. There is another possibility. I suspect that what is being measured is the amount of memory the system reports as being free. Contrary to what one might at first think, it is not necessarily better for that figure to be larger. Remember that “free” memory is just sitting there — it’s not doing anything useful at all. Many modern systems (I believe including Windows 7) will use available free memory for anticipatory caching: “read ahead” buffering for disk I/O, for example. If this is done intelligently, it will improve system performance. The Acer laptop, running Ubuntu Linux, on which I am writing this post, is currently using about 103 MB of its 1 GB of memory for such caching. The real question is whether the system is sufficiently memory constrained that it is swapping or paging to/from the disk. I have checked my laptop, and it has done exactly one I/O operation by the virtual memory system in the last 30 seconds.
The report also cites higher CPU utilization with Windows 7 machines. This, too, can be a red herring. I have, from time to time, run background, low priority “grid” applications like SETI@Home or Folding@Home on my desktop machine (which also runs Ubuntu Linux). When these applications are active, the CPU utilization on the machine is essentially always 100%; and the effect on an interactive user is negligible, because the applications’ execution priority is set so that they only get a CPU time slice if nothing else wants the processor. Put another way, if your CPU utilization is 50%, then half of the time you machine is sitting there thinking to itself.
Those of you who have read this blog before probably have noticed that I am not exactly a cheerleader for Microsoft. And it may be, if we could see all the analysis behind this report, that there really is a problem. But on the evidence presented to date, I’m skeptical.