With this last PC upgrade, I had the opportunity to try out and benchmark a number of different video cards including a dual-card SLI system.

I tried 2 different CPU's, several combinations of system memory, and 3 different video cards. Since this is a ton of stuff, I wont bore you (or me) with all of the results.

To compare the systems, I used three programs: 3DMark2006, Aquamark3, and Half-Life 2: Lost Coast. Aquamark is an older program and the company that wrote it is out of business. Since actual games are better benchmarks than test programs, I tried to use Battlefield2 but was unable to get it to work. Half-Life: Lost Coast did work but did not show much change between systems so the main benchmark turned out to be 3DMark06.

The goal of these tests was to answer two questions:
Q1: How do these cards compare to each other?
Q2: Is the new Ahtlon64/X1800 an improvement from the system that I used to have?

Testing

My base system was an AthlonXP 2500+, 1GB of 333DDR, and a Radeon X800 XT card in an AGP slot.

My new system is an Athlon64 3000+. Memory was 1GB at 333 or 400MHZ or 2GB. Video cards were a single X1600, dual SLI X1600's and a single X1800 XT, all using a PCI-Express slot.

Using 3DMark, I tested two combinations. One with no anti-aliasing and "optimal" filtering and a second setting with 4x AA and 4x Anisotropic filtering.

All tests were done at 1280x1024 resolution on Windows XP SP4. All the tests shown here used 1GB of system memory (a pair of 512MB sticks).

(I should also note that I experimented with 2GB of system memory. Unfortunately I did not have 2GB of matching memory so I combined 1GB 333MHz with 1 GB of 400MHz. The results were a mess and very slow because the system struggled to find an appropriate memory speed so I did not include the results here.)

3DMark2006 Tests

These tests show that CPU score didnt change much between the AthlonXP 2500+ and the Athlon64 3000+, which was a little surprise.

The single X1600 was comparable to my original X800 however the dual X1600 system was a huge improvement. Better yet was the X1800 that I eventually got.

Changing memory speeds from 400MHz down to 333MHz also had a small but noticeable effect.

These tests are more important because they mimic actual gameplay but they match the conclusions from the synthetic tests.

There is little CPU change. The X1600 is about the same as the X800 (except it ran the HDR tests) and the dual card and X1800 systems are much better. Memory speed did have an impact.

This final table is a numeric recap of the previous charts. Each column shows the percentage increase/decrease over the base system in the first column. The one exception are the HDR tests. Since the base system would not run these tests, the columns actually compare with the second column or single X1600 at 400MHz.

Aquamark3 Tests

The final tests were with Aquamark. Aquamark is DirectX 8 I think which is a bit dated. However I have data going back several video cards and CPU versions so it is interesting to see Moore's Law in effect. You can also see the shading differences (texels per second or TPS) with the X1600 which has a about half the transistors of the X1800.

Closing

Although the real test of a system can only be found by playing your favorite game(s), these benchmarks were helpful to me.

I planed to use a dual SLI X1600 setup in this new PC and I bought two cards with passive heatsinks and no fans for peace and quiet. Unfortunately, one of the cards fell apart within a week. The single X1600 worked fine for Day of Defeat:Source but it really struggled with Oblivion.

Thanks to the wonderful customer support at Newegg.com, I was able to return the cards and replace them with a single X1800. The tests show this single card is better than the previous pair and I can confirm this with my Oblivion experience.

Now that these baseline benchmarks are complete, I will attempt to do some overclocking.

Since I had heard that the fans on the X1800 were terrible, I bought a water cooing system. It turns out that the X1800 fans are not bad at all so I havent felt the need to try the cooler yet.

Having said this, a water cooler will allow me to experiment with overclocking both the CPU and the video card to see how that effects performance. (Another benefit of the X1800 is a heat sensor that facilitates overclocking and is not present on the X1600.)