even though the score is pretty much meaningless. I really wish they would make the scale work properly, at the minute it doesn't measure anything meaningful and the scale is senseless. It had the potential to be such an incredibly useful tool.
I'm going to have to partly disagree. One of the primary advantages and purposes of WinSAT is to make it so that all Windows Applications and Games can use essentially the same benchmark tests- done by the system- and those ratings can be done independent of the Application.
As for what it measures- actually the ratings themselves are a very brief summary of the data that is collected. There are a huge number of XML files found in %SystemRoot%\Performance\WinSAT\DataStore that you can peruse to see the various measurements made that arrive at that score. Some isolated examples picked from only my Graphics WinSAT data include information about the Pixel Shader Version supported, Resolution, etc. It appears There are numerous tests that use various capabilities, such as texturing, multitexturing, alpha blending, large amounts of polygons, etc.
CPU Scores are calculated using a similar scheme. According to the assessment data it appears to record similar statistics such as total bytes throughput while performing streaming encoding and decoding and encryption/decryption (or both), compression tests, and of course standard mathematic functions and the use of various math instructions. These are recorded in what seems to be pretty good detail- I can look back to when I last ran the test on March 10th, 2010 and see that it took my CPU 732572047 clock ticks to complete the compression assessment using a single thread, and only 182810123 ticks using four threads; It even appears to have data that records the same test being run across the different cores by forcing affinity.
Seems fairly exhaustive. I say I partly disagree, though, because unfortunately no applications actually use it that way and so far it's just been used as a rating for people to compare.
This might seem like a bit of a tangent, but I think it partly explains how the jump would have occurred, particularly on a laptop. Laptops don't generally have very powerful graphics processors, and a lot of the actual Oomph for 3-D acceleration is actually taken on by the Driver software itself. Since the Driver is the thing doing all the work, new versions can make
huge improvements to the apparent speed of the thing they are allegedly driving. Even by implementing their own device-sensitive algorithm for specific capabilities can give big gains; if Direct3D cannot find the capability exposed through the Driver on the graphics card, it will just go "ok, fine," and emulate it itself. But that logic will be more or less device agnostic; if the Device driver says "nope, I don't support that Graphics Extension/Capability" in one version, DX will emulate it which will slow things down. If in a later version it says "yep, I support it" and then implements it using an algorithm that involves heavy direct work with the graphics adapter, that is going to end up faster simply because it is more "Direct".
This of course applies with standard "Graphics" which I think measures 2-D Desktop performance. There are numerous accelerator features provided by the video card for features we now take for granted like alpha blending and stretching and resizing bitmaps and whatnot. If these are implemented mostly in the driver software and that driver software happens to improve those implementations, you get faster performance.
Interesting side note: Windows 8.1 has removed the UI for any WinSAT features. That is, you cannot get it anymore.
You can still run the WinSAT tests by running 'Winsat formal' at the command prompt and reading the formal report from the dataStore folder described above.
(Apparently, my laptop now scores 3.6 for Graphics where before Windows 8.1 it scored 3.5- This is clearly the happiest day of my life.)