When all is said and done, we always come down to this page right here. Cold, hard numbers that tell us the real story of whether a product is worthy or not. But with these new 8x AGP boards hitting the streets, we have to ask ourselves if it really matters. I figure that the best way to see the answer to this is to run the tests on this new 8x board and also on a reference Ti4200 board from nVidia. Both boards will be run at factory speeds so we should see firsthand if it matters. Theoretically, the doubled bandwidth should help out in heavy graphics situations, but the numbers should tell the tale.
But before we jump into the tests, lets take a look at what we'll be running these boards on:
Motherboard: EPoX 8K9A2 (KT400 chipset)
Processor: Athlon XP 1800+ @ 1870MHz (Thoroughbred)
Memory: 512MB Crucial PC2700 DDR
Display: Hitachi SuperScan 814 21" CRT
Hard Drive: Seagate Barracuda IV 40GB
The tests I ran were very simple so that anyone can perform them for themselves. They consisted of the 3DMark2001SE from Futuremark (formerly MadOnion) and a pair of benchmarks available in the Unreal Tournament 2003 Demo. The tests were run on the reference Ti4200 board, the Prolink Ti4200-8x board and also on a Prolink MX440-8x board just for comparison sake. Drivers used were the Detonator 40.52 on all boards tested.
With the release of newer video cards with greater amounts of bandwidth, the 3DMark benchmark is beginning to show its age. A new version is due out soon, but since it is not readily available we will make due with what we have in hand.
While not a huge difference in performance numbers, we can still see nearly a 600 point increase when comparing the 8x version with the original 4x variety. While it does show a beneficial gain, there isn't enough here to make me jump up and down for joy.
One item of note, however, is the Nature benchmark. This particular benchmark has thrown video cards for loops since it first came out. The amount of data that it throws at the card is generally staggering to older boards and can cause all sorts of slow, jerky graphics. Here is what we ended up with in this test:
Now the difference between the 4x and 8x AGP busses are beginning to show a bit. While the amount of frames per second differ by less then 8FPS, this equates to roughly a 12.3% increase with the 8x board in place. Now things are getting a bit interesting.
Oh, and for those noting the lack of a result for the MX440 board, remember that the Nature test requires hardware support for DirectX 8 or higher and the MX series of video boards do not have this capability.
Unreal Tournament 2003 Demo
When you download the UT2003 Demo, it comes complete with a pair of built-in benchmarks. The first is called a Flyby Benchmark and consists of two demos that record a virtual tour of the levels. It is similar to being in Spectator Mode in the Quake series of games. During the virtual tour, it records the frames per second of each map and then gives you an average.
The second test is called the Botmatch Benchmark and consists of another pair of demos; this time with bots having a fragfest. Since there is movement involved with the characters on screen, this test will have a more drastic effect on the frame rates. After the two demos have run, the program again calculates the average frames per second and displays the result.
So how did the contestants fare in this endeavor? I'm glad you asked.
UT2003 Demo Flyby Benchmark
We see a pattern here similar to the overall benchmarks from the 3DMark utility; an improvement but not a large one. Of course, it does well for bragging rights to be able to run the UT2003 benchmarks at nearly 150FPS.
UT2003 Demo Botmatch Benchmark
Well, the difference here isn't worth much at all. We'll call this test nearly a tie in terms of performance. It looks as though the Unreal Tournament game just doesn't have the strength to fully stress out today's modern video cards.