Benchmarks: 466MHz Digital Audio G4 from 16MB ATI Rage 128 to 128MB Radeon 9600XT

[Jethryn Freyman]

OK, so I finally swapped out the ATI Rage 128 in my 466MHz Digital Audio Power Mac G4 for the Radeon 9600 XT that came with my G5 tower. I had to disable pins 3 and 11 on the card otherwise the Mac wouldn't even power on, I used an X-ACTO knife to scratch away at the solder lines connecting the pins to their resistors (R15 and R60.)

Specs:

-466Mhz G4 (7400)
-1.5GB memory
-80GB, 7,200 rpm hard disk on IDE
-Mac OS X 10.5.8, Leopard
-2048x1152 display, connected via VGA

Anyway, the only benchmark I could run was the original, software rendering only version of Quake. OS X version 1.10 that is. I specified it using the command line to allocate up to 384MB of memory to the game if needed. Screen was running at 1600x900, the highest the ATI Rage card could display properly, scaled up to fill out all 2048x1152 pixels on my screen.

Ran it at 1280x720 first up. Scored 24.2 fps.

OpenGL accelerated Quake did not launch, gave a pixelformat error.

Swapped out the cards. Used a DVI to VGA adapter to connect the Radeon card. Set display resolution to fill the screen at 1600x900.

Ran the same timedemo at 1280x720. Scored... 18.0 fps. What?

I got a score of 17.4 when running it at 1280x720 with the actual system display set (in Display Preferences) to the native 2048x1152.

I then ran the timedemo in the OpenGL accelerated build of Quake. Scored 59.4 fps.

Perfectly buttery smooth. I think it would have played through faster but VSync may have been on or something... more tests to follow.

Whole UI of OS X is much smoother, also. Awesome upgrade!

 

[wobegong]

Nice job, though a Rage128 is positively prehistoric so that should be a very noticeable upgrade - enjoy

 

[Jethryn Freyman]

As promised:

The green bar is the G4 with original 16MB ATI Rage 128 Pro. It could only perform the first test, which was Quake in the software renderer.

The red bar is the G4 with the 128MB ATI Radeon 9600 XT. It completed all tests.

The black bar is my dual 1.8GHz G5 with 256MB ATI FireGL X3 (at standard clocks.) Added for comparison. Didn't bother completing every test.

Quake in the software rendered appears to have been running with vertical sync (VSync) on. As my monitor was at 60Hz, it was unable to go above 60 fps. In OpenGL accelerated Quake, there's the option to disable VSync. The graph doesn't show it, but the FireGL in my G5 scored 397 fps for the GLQuake test.

Halo was run with minimum details set. UT2004 was set with maximum details, sound disabled, and on DM-Antalus.