I know there are plenty of random tweaks floating around for Windows 10 and I thought it would be interesting to benchmark a few of them with 3dMark05 on one of the older machines I have lying around.
Now almost all of these adjustments are generally pretty situational, so I didn’t really expect to see any performance improvements. For example DPC Enhancer is supposed to be for buggy drivers and while I know that it does change the power settings for graphics cards on PCs with multiple cards, I know nothing else about how it works, it’s entirely a black box. So while I’ve found it to be useful in making my gaming machine noticeably more responsive on the desktop, it’s not really a performance tweak.
Still, the results are interesting, first, the numbers.
Each test was run six times, and the final result was the average of all the scores. I rebooted between changes.
It’s interesting that disabling paging resulted in an improvement, but probably not that surprising. Normally it’s the sort of change you make to prevent the computer starting to “stutter” after extended usage. Paging is also caused by IO and some people don’t want their programs to be pushed out for file accesses. It’s also worth noting that on Windows 10, disabling paging also disables memory compression. Despite what the chart says, this change was only 1% or so. It’s not an effective performance tweak.
DPC Enhancer was also useful, which was somewhat unexpected given the test machine. As mentioned previously I always thought of it as more utilitarian. Among other things the test machine doesn’t have the kind of hardware that would benefit from this tool, being a fairly old machine with a simple configuration.
Dynamic Ticks was also a surprise, while I did expect that it could make the computer slightly more responsive to input, I didn’t expect that it would actually improve the benchmarks. Generally when a test is running you don’t expect the processor to be going at anything less than the maximum anyway.
In the end, these are situational tweaks for specific issues or for specific applications that would benefit from having them. Not for performance. However, seeing the effects they have on straight performance is still worthwhile since we can know if they’re actively detrimental or not in other cases.