MacBench's 2D performance results were nearly identical for the two cards when installed in the PowerCenter Pro (210MHz 604E) at thousands colors/1024x768. However at higher color depths or with faster CPUs that could 'feed' more data to the cards, the Rage128 was approximately 20% faster at 16-Bit tests and about 25% faster at 32-Bit tests in MacBench than the Game Rocket.

MacBench 5.0 Tests:
I ran MacBench 5.0 Graphics and Pub Graphics (requires 1152x870 mode) tests for all three machines. I've graphed the results by machine.

MacBench Graphics Primitives Tests:

Since I ran a complete series of MacBench graphics tests I examined the Quickdraw primitives performance (too long a list to show here in a graph). What I found interesting was the wide spread of scores on some specific low-level Quickdraw function tests, especially with the slowest CPU tested. For example the results with the 210MHz 604E of the PowerCenter Pro running 1152x870, millions colors the major function performance differences were:

• Copybits Transparent: Rage128 was 1526% faster than the Game Rocket
• FramePoly: Game Rocket was 95% faster and RagePro was 15% faster than Rage128.
• EraseRect: Game Rocket was 307% faster than the Rage128
• FillRect: GameRocket was 80% faster than the Rage128
• FrameRect: Game Rocket was 202% faster than the Rage128
• PaintRect: Game Rocket was 337% faster and RagePro was 17% faster than the Rage128
• EraseRgn: Game Rocket was 304% faster than the Rage128
• FillRgn: Game Rocket was 78% faster than the Rage128

The MacBench Graphics and Pub. Graphics tests run a scripted series of Quickdraw calls that simulate actual applications like Photoshop, Word, etc. running (in fact you can see this during the tests). The weighting of the functions above would depend on how often they were used in applications. Erasing, filling and painting rectangles and polygons would be fairly common I'd guess, but apparently the applications scripted tests of the Graphics and Pub Graphics tests in MacBench must not use them often based on the scores it returned.

I also ran some informal scrolling tests in Photoshop and Word with the card in the 66MHz PCI slot of the B&W G3/400 at 1152x870, millions colors. Both cards were within a few seconds or so of each other, with the Rage128 always a bit faster. Many simple tests were within what I consider the margin of error for stopwatch tests, but overall 'feel' of the cards in most applications made me realize the Rage128 is a faster card by a bit. It's hard to 'see' a 20% difference in actual use.

The 2D MacBench Graphics/Pub Graphics scores showed the Rage128 was significantly faster both test modes (as much as 20-25%) and it was the fastest 2D Mac Graphics card I've tested to date. The Rage128's more advanced engine appeared to have more performance potential, especially in 32-Bit mode. The 565 Pixel format workaround in thousands colors (16-bit) mode for the Banshee processor may result in less than the full color range for that mode, but imaging professionals would be running millions colors anyway so this is not a major issue. Considering everything, I rated 2D performance of each card as follows

• ATI Rage128: 9.5
• IXMicro Game Rocket: 8.5

The next page covers 3D performance in non-game applications.