This site is linked from the Windows Experience Index site on Wikipedia. You can upload your own score and it gives you a badge to put on your site. Mine came out as 3 but WordPress won’t let me paste style snippets.
Now the the recent unpleasantness seems to be behind us I thought I would knuckle down and start reviewing the Acer Ferrari 1000 that Microsoft and AMD sent me. One of the things that seems to be the cool thing to do when you get a new laptop running Windows Vista is to check out the Windows Experience Index.
What is the Windows Experience Index?
The Windows Experience Index is a number which rates the hardware and software configuration on your machine as it relates to its ability to run Windows Vista. The one number you get is actually a base score which is actually the lowest of a number of sub-component scores. Wikipedia has a great entry with several links which describes which tests are performed in order to determine the scores.
If you have a Vista machine handy you can get in and look at the Windows Experience Index by clicking on Start and typing “performance” into the search box, then selecting “Performance Information and Tools” from the results (usually the second in the list).
One thing that I will council you on is that sometimes the WEI is way wrong for whatever reason and you need to get it to recalculate. Ed Bott’s post reminded me of this as he re-ran the score on his Ferrari 5000 review machine.
How did the Ferrari 1000 score?
For example on the Ferrari 1000 the base score is 2.8 which I thought was pretty low, but if we look at the break down we can see what the problem is.
As you can see under the Graphics component it is saying that there is a sub-score of 2.8 which is the lowest so that becomes the base score. Since I’ve got three machines (laptops) on my desk here running Vista I thought that I would do a side by side comparison to see if we can figure out what kind of configuration settings make a difference. The machines are:
So lets look at processor performance first. On the processor the Acer scored 4.7 which is on an AMD Turion 64 X2 mobile processor (dual core jobbie). I compared this with the Fujitsu (4.7) and the Dell (4.7) both of which were running Intel Centrino Duo processors. Since they all came out the same I can probably take from that that the tests were really designed to use the 32-bit processors, it would be nice for WEI to be updated at some point in the future to leverage the unique capabilities of 64-bit processors but that would invalidate comparisons between 32-bit and 64-bit machines (although 32-bit machines are an endangered species anyway).
Next up is the memory performance and this is where I saw some differences between the Microsoft supplied machine and the Dell and Fujitsu that I already had. The Ferrari clocked in at 5.3 for memory performance where both the Fujitsu and the Dell came in at 4.7. I wonder if this is down to the 64-bit processor and some optimised instructions and it would be interesting to compare to the Core2 processor (anyone want to provide their stats here?).
Graphics performance is actually comprised of two numbers, what I call your standard desktop/shell graphics performance and your gaming graphics. On the Ferrari the desktop graphics performance came in at 2.8, where as the Fujitsu score 3.4 and the Dell 3.6. In my experience with Vista the drivers that you are using make a lot of difference. On the Acer I am using the configuration that was supplied by Microsoft and AMD which I suspect are problably the approved out of the box drivers that came with Vista for this particular ATI card. On the Fujitsu I am running the bog standard Intel 945GM drivers for Vista and on the Dell I am running some hacked NVIDIA drivers from LaptopVideo2Go.com. So at the moment the NVIDIA drivers are giving better performance but I wonder whether updating the ATI drivers would yield better results – I won’t do it now half way through the review but I will retouch on this later and try the latest drivers from ATI (if I can find them).
On the gaming graphics score the Acer, Fujitsu and Dell scored 3.1, 3.0 and 3.2 respectively. So they are all pretty close to each other there. Both the Acer and the Dell support up to 512MB of RAM allocated to the task of doing high-end 3D graphics and both of them have 256MB discreet and then share the rest with the operating system. The strange thing about the ATI driver was that it said it was actually capable of sharing up to 895MB of RAM for graphics from the OS – so I am actually quite suspect on that driver right now.
Still – this card is going to be OK for most gaming tasks and can certainly hold its head high in laptop circles (I am sure some of the desktop GPUs would really leave it to shame and I have heard some good things about the Dell M90’s from co-workers).
Finally lets look at disk performance. Disk performance is often one of those things that is overlooked when configuring a laptop, and if you do anything that requires lots of disk I/O like running virtual machines, compiling code or reading 3D models and textures then you are going to want a fast disk. Both the Acer and the Dell are configured with 7200RPM drives and they scored 4.8 and 4.9 respectively which is OK I suppose. The Fujitsu only scored 4.4, but it only has a 5600RPM drive (or whatever) so whatever is tested there obviously doesn’t stress the metal too much (I expected much lower for the Fujitsu).
Overall the Ferrari 1000 probably sits somewhere between the Fujitsu and the Dell in terms of performance but if you have memory I/O intensive applications then the Ferrari is probably going to get the job done better. It would be interesting to compare these results to the Core2 chips from Intel but I think that AMD has a reasonable offering here which was actually news to me.
What do I think about WEI?
When Microsoft invented WEI they basically generated the profits for hardware vendors for years to come. Even within Readify reporting your WEI has become the new status symbol and people are actually spending serious time comparing their ratings and optimising their machines – its yet another geek status symbol and I don’t think it’ll be long before we see WEI status proudly displayed on the front of computers and geeks sporting WEI t-shirts.
Obviously the devil is in the detail so you really need to look at the sub-component scores and if you are actually going to go out and buy hardware to optimise your system then I suggest reading up on exactly what the scores are testing and figure out if you really need that. Despite what some people say, Vista will run fine on existing computers. I used to run Vista on an Inspiron 6000 which was quite good, it just didn’t support Aero (or at least it didn’t during the BETA) and I needed that for some of my development work.
From the development perspective it is interesting to speculate how we could use the WEI statistics to optimise the behaviour of our software. Deepak Kapoor has an example on his blog on how to get access to WEI information using WMI and C#. How much more complex would your code get if you started optimising for WEI stats along with your program logic – probably something best left to the runtime.
When I read this article by Calvin Austin, this image popped into my head and I couldn’t help but put it down on paper.
Years ago I was talking to a Microsoft employee who mentioned that sometimes he thinks that there is an inverse reality sphere around Redmond, but I think it is a general problem in this industry that we believe that 2 + 2 = 5, hrm, maybe the Java folks have raced ahead and implemented the “because” operator
Platforms.Rank = “Java” !! Platforms.Options == PlatformOptions.DotNet && PlatformOptions.Ruby;
Sometimes I kill me, I really do