Flashed AGP Radeon 92x0 issues - anybody else?

[Franklinstein]

A while back I picked up some low profile Radeon 9200-series graphics cards for Cube use, both 128MB but one a 9200 and the other a 9250, may or may not be the not-recommended (for reasons currently unknown to me) LE or SE versions. Recently I decided to try to flash them. I got the 9250 to (mostly) work but I botched the resistor move on the 9200 and I haven't tried to fix it (lifted pad).

The problems I'm having are that regardless of which flash I use (either the Tome or the "unknown" reduced 9250 flash from the MacElite downloads area): there's a permanent second display at 640x480 regardless of whether anything's plugged in or not; sometimes the main display is stuck at 256 colors until a reboot which usually fixes it; and OS 9 and OS X both misreport and disagree on the amount of VRAM on the card (OS 9 says 64MB, OS X says 256MB).

Anybody else notice similar problems with these cards? When the card works, it works pretty well (both OS 9 and X), but sometimes it takes a couple reboots to get it to cooperate.

 

[Franklinstein]

OS 9 may be a driver problem because if there are no drivers it's fine (except for the occasional 256 color thing) but when the 9250 drivers load the screen starts getting crazy and it's unusable. Without the drivers, though, there's no acceleration so it's kind of useless for dual-booting. Maybe I need to break out ResEdit or something.
But I may not bother: I ran XBench and the numbers were disappointing, at least for OpenGL performance: between a GF3 Ti500 (flashed from PC), GF2 MX (Apple OE), Radeon 7200 or whatever (SGRAM type, Apple OE), and the 9250, the 7200 had the highest frames/second, followed by the GF2, then 9250, then GF3, which I found odd. The other tests were all over the place but the 9250 was underwhelming across the board and even where it was good, it never outperformed anything by better than 10% and that high score was on the text scrolling UI test. All tested on the same 1GHz Cube in 10.3.9. Maybe it's just a quirk of XBench's graphics tests (some cards work better at different resolutions/bit depths vs. whatever the standardized OpenGL test box uses) but it's still a disappointing result.

So if performance improvement is marginal at best over what I already have that works as-is, I don't think it's worth the effort to try to keep mucking about with this one. Even the Apple GF3 may not be worth the premium over the GF2 but I may try more extensive testing before I make that determination.