The graphics card is obviously another key component to improving performance, though unfortunately the latest and greatest solutions again do not guarantee stellar performance. Having said that, anything less than a Radeon X1950 Pro is going to make life difficult when playing on the larger maps with more than 2 players. In fact, using the low quality settings at 1280x1204 saw the X1950 Pro render just 24fps on average. The ultra powerful GeForce 8800 GTX was not a great deal better, producing an average frame rate of just 31fps. The maximum in-game quality settings saw the 8800 GTX fall to an average frame rate of just 22fps.

Based on the results we recorded it is clear the GeForce 8800 GTX is being limited by something else, most likely the processor. Looking at the high quality Intel Core 2 Duo E6700 results we see that anything faster than the Radeon X1950 Pro (7900 GTX, X1950XTX, 8800 GTS and 8800 GTX) benefits greatly from higher processor frequencies. The average frame rates continued to climb when using these four graphics cards, as the processor frequency was increased. The very same thing, possibly to a greater degree, was found when testing the AMD Athlon64 X2 processors.

Therefore, it is clear that this game is heavily reliant on raw processing power, probably more so than rendering power, though they are both very important. What I did notice; and for those wondering why we only tested with 2GB's of DDR2 memory and did not use 1GB and 4GB configurations, please let me explain. In the past we have tried to prove the worth of 2GB memory kits when compared to smaller 1GB kits. While we could physically tell the difference when playing games, the average frame rates never really seemed to change, at least not enough to justify the added cost. Unfortunately, when moving from 2GB's of memory to 4GB's with Windows Vista, the same problem presents itself.

Nevertheless, using Windows XP first I did test Supreme Commander with 1GB of memory which was just plain awful and then 2GB's which, was a vast improvement. However, after having measured the performance with 1GB and then 2GB's of memory, I noticed the frame rates hardly changed, with results increasing as little as 2 ~ 3fps. Therefore, I did not want to include more pages of graphs that show near no difference in performance and rather decided to discuss this issue. After testing these two memory configurations in Windows XP it was time to compare 2GB and 4GB memory configurations in Windows Vista.

Supreme Commander will be available on the 20th of this month and if you are going to grab yourself a copy, keep in mind that this game is not even half as good with low quality settings as it is with high quality settings. Of course when the game is finally released, the game performance might be a little less taxing than what we have seen here today, though I very much doubt it. In fact I really believe that our results are some what conservative given we were never heavily engaged with the AI (Artificial Intelligence). With a rather limited amount of units on a small map I really do not believe we pushed this game anywhere near its limits.

Imagine a map two or three times bigger with more than twice as many players engaging in a non-stop battle. Surely this will be far more demanding when compared to the demo we tested here today. Another thing to keep in mind is that all our tests were run on a relatively high-end Core 2 Duo system with 2GB’s of DDR2 memory. While we are not going to test more platforms until the full version is released, it will be interesting to see how the GeForce 7900GS for example and even the GeForce 8800 GTX cards, handle themselves on slower platforms.

Back when we first looked at Company of Heroes the results were a little surprising and we found that with medium quality settings even a GeForce 7300GT paired with an Athlon64 3800+ processor could deliver reasonable performance. However, Supreme Commander is obviously a lot more demanding, as even the 7600GT was unable to handle medium quality settings. The performance differences between the 7900GS and X1950 Pro were very small as both graphics cards produced very similar results. Surprisingly, both were limited to 1024x768 with high quality settings enabled.

As with most articles of this nature, we can't fully give a conclusion because so many factors come into play. Though as it stands, we can learn a lot from this simple round of tests. First is the fact that Vista is not going to be an OS for an ultimate gaming machine anytime soon. This is something that you should consider before you throw a bunch of cash at a new computer and then pick up Vista along with it.

The Vista NVIDIA drivers used during testing were just finalized the other day, the result of a string of tireless days of bug smashing. Vista is not necessarily a harder OS to develop for, but because coders need to learn the new architecture, we can expect to wait a while before our PCs are "perfect". Even with these "final" drivers, SLI is still inoperable and may not be for another few weeks. Even at that time, nobody is going to know whether it's going to equal the performance we currently see on XP. Even with my simple one card testing, we can see there is a lot of work to be done, both with the drivers and perhaps with games developers releasing updates that better support Vista.

So, I've said it before and will say it again. Vista is a good operating system... but it's -far- from being even 90% reliable. When going from an XP machine to Vista, it feels like you just went from a Toyota Celica to a Ferrari 599 GTB Fiorano. The only problem is, that it feels like your new Ferrari borrowed your Celicas engine. We'll let the numbers speak for themselves.

In the end, if you are making the move to Vista and are a gamer, you may as well prepare to partition your hard drive to dual boot. You are not going to want to go through the hassle of making your games run smooth in Vista, or go through the trouble of tweaking to make it happen. Hopefully within the next two months, NVIDIAs drivers will be much more refined and -all- games should run as they do on XP, or at least close to it. I regret not having an ATI card on hand to perform testing there as well, but it may not be a far stretch to expect a similar experience there.