Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

z970

macrumors 68040
Original poster
Jun 2, 2017
3,591
4,546
Well, I figured that if my '03 G5 is going to have a deliciously excellent Radeon 9700 Pro, I may as well go ahead and give my dual-core the same treatment, as I've wanted to do for a long while now. So, I went ahead and bought one of these:

X1900GT_fhq.jpg


A PC Radeon X1900 GT Revision 2, with the (as far as I can tell) improved cooling system over rev. 1. Everything else is technologically identical to what was sold for PowerPC.

A long and arduous story short, it has been flashed with the retail ATI Radeon X1900 GT for G5 BIOS available at The Mac Elite, and now works like an absolute charm, even on 10.4.11. Leopard, Linux, and BSD I have not yet tried.

The only real (minor) caveat I've come across are (mostly) green pixels scattered all over the display, plus occasional thin horizontal bars seen below:

Picture 2.png


However, these were present with the original PC BIOS, which rules out the possibility of a bad flash. Beside that, the card arrived chock full of dust and debris (plus the original paste / pads) which I have not yet refurbished out, so my intuition says that this could be a sort of thermal issue as a result, or maybe some dust bunny interfering with the circuitry in some way. It is also a rather heavy card; maybe there is too much uneven weight on the PCIe slot, even with the bracket firmly seated?

At worst case though, the performance gained from this card over the stock 6600 is worth having to live with these small graphical oddities, case in point Web scrolling being even smoother over said 6600, and fluid 1080p HD YouTube playback being just ALMOST possible in TenFourFox.

All in all, even though I haven't seen anybody showcasing this particular card to work with a G5, I can safely confirm here that the Radeon X1900 GT Rev. 2 works exactly as the original, and is a great upgrade over any of the OEM cards sold with the PCIe G5s. :)
 
Have supplies of the X1950XT dried up now? People were giving them away when I flashed mine and it has never left any display artefacts or caused any glitches. Good to know there is an alternative.
 
Largely, that seems to currently be the case, yes.
 
Yes. For my case in particular, very.

Having the display connected to the 6600 in Slot 3, with the X1900 GT in Slot 1, Graphiccelerator ATI Multi Dumper wouldn't produce anything, and Graphiccelerator ATI Multi Flasher wouldn't even open (this being after 'Run Me First' is run, and a ROM is already loaded). Even now, after the Mac ROM has been flashed and the display running off the X1900, ATI Multi Flasher still won't open (with all prerequisites of course filled). Thus, rendering Graphiccelerator essentially useless, thereby forcing me to use my less than stellar only PCIe flashing solution.

(That, along with its favoritism when flashing ROMs and downright refusal to open when a certain card is installed that it doesn't like, doesn't give me a very good impression of the tool at all.)

Back to the story... I don't have a mini-tower PCIe PC. However, I've got a desktop-size OptiPlex GX620 that takes (horizontally) low profile graphics cards. So, the card would fit in its single PCIe slot. The trouble is, the bracket won't fit properly into its slot, meaning the card is forced to hang at a slight angle from the PCIe slot, leaving only the little latch on the PCIe connector from keeping the entire card from falling out. In the process, only barely being able to "stably" fit.

But there was another problem. As this computer was never designed to take calibers of GPUs that required a power cord, it wasn't able to power the card. But the G5 was, while the cord was only about a foot long.

So I ended up having to FLIP the OptiPlex vertically upside down and position the unit as up against the running G5 as much as possible in order for the cord to be able to reach the card while plugged into a separate machine. It was actually a very dangerous configuration, but the only possible route I had available (no thanks to Graphiccelerator).

Of course... at the time, I set that all up so I could back up the ROM BEFORE I tried actually confirming if ATI Multi Flasher would work with the card. Therefore, I had to devise this goofy setup to first back up the ROM with DOS 7.1, then put it back in the G5 only to be blown off by Graphiccelerator (I tried safe mode), then fashion it all BACK together to try flashing the ROM from DOS instead.

But it didn't end there. When I first pulled any video out of the card, there were artifacts and corrupted text characters in places where there should have been normal characters and text. Whether that was a problem with the card itself or the relatively unstable connection to the machine, I can't say for sure. Regardless, I had to try to work through graphically-corrupted DOS to attempt this flash, which I did end up accomplishing after another period of trial and error.

Anyway... Even given the minor reoccurring glitches with OS X, I believe this card is so far needless to say, happier when living as a G5 Edition.

All that said, it's important to keep in mind that most of my troubles stemmed from this individual card and my own lack of proper equipment. Anyone who has a Radeon X1900 GT R2 and a full-size PCIe PC should have no problems at all when flashing and using theirs.
 
Last edited:
I wonder if anyone has tried flashing one of these? https://www.ebay.com/itm/251593798054

I just did. The above are my experiences in doing so.

As to the drivers, they locked up Tiger on the first couple reboots. Afterward, they appeared to be fine.

Half the time, OpenMark wouldn't start properly. The other half it would lock up mid benchmark (cursor control remained). I'm hoping these are all just graphics library issues that OS X can't properly handle.
 
Last edited:
So, I refurbished the card yesterday (which involved opening it back up twice due to a problem with the VRAM chips not making contact with the heatsink; thermal paste is a PITA to remove). The fan is clean and lubricated, heatsink is clean, board is dusted, and the paste and pads are fresh (especially the paste).

Unfortunately, this does not seem to have helped the various pixel errors around the screen, which even in Leopard now, have also locked up the display once during Expose. Not only that, they are also starting to produce (small) artifacting in the Dock, windows, and other graphics applications. Although I continue to believe the lock ups are caused by some kind of improper libraries error in OS X 10.4 and 10.5, the artifacting is roughly comparable to what I've seen in the Late 2006 iMac, albeit on a slightly less critical scale.

I'm beginning to suspect that these issues are not caused by an improperly supported card in slot, but something I remember somebody explaining somewhere on YouTube. Going off of memory, apparently there was a required industry switch to lead-free solder sometime around 2003 or 2004, but the new lead-free solder mixtures were very new and rather unvetted, resulting in brittle solder that was prone to disconnecting over time, thereby explaining the consumer electronics board failures in the mid 2000s (Xbox 360, MacBook Pro, iMac, Nvidia and Radeon GPUs, etc.) thus also explaining how reflowing a graphics card can (temporarily) fix this. Of course, the concurrent and prominent capacitor plague plays a role here, but often times these issues are usually caused by one of the two, either way producing similar results.

Or something like that, at least... Seems plausible to me.

-

On a slightly more interesting note...

Picture 2.png


For some reason, Leopard seems to think the X1900 GT is an X1900 XT. Not only that, after QuartzGL and the rest are enabled via AuroraAccelerator, QuartzGL turns up in System Profiler as "Supported", which is not something I believe I have seen before on other cards.

Very interesting.
 
the artifacting/locking up is because your card is either failing or is being clocked out of spec (or both)

its nothing to do with OS X's drivers (especially if the artifacting was present before flashing)

(it shows up as a X1900 XT in OS X because it shares/uses the same framebuffer personality as the X1900 XT found in the first generation Mac Pro, note its actual name is "ATY,RadeonX1900" its System profiler thats sticking the ATI Radeon X1900 XT there)

if you want something even weirder, 7240 comes back as the Device ID of an X1950 XT, (actual X1900 G5's are like this too) so go figure that one out...
 
if you want something even weirder, 7240 comes back as the Device ID of an X1950 XT, (actual X1900 G5's are like this too) so go figure that one out...

Seems fair as I have a 1950 in mine masquerading as a 1900. I wonder if the original Mac cards were downclocked 1950s?
 
  • Like
Reactions: LightBulbFun
Being that I've got an MDD now, I bought a cheapo 9800 Pro for the G5, as the 9700 can just go into the G4 as an upgrade over its 9600.

So I'm trying to flash it in 10.5.8, and true to form, Graphiccelerator won't even open, and neither will ATI Multi Flasher. It's the same behavior in both safe mode and normal mode.

What a useless utility.
 
If you see it with the PC bios, as said, it may be failing or some thermal issue.
A good clean up and new thermal paste to see it solves.
 
The card isn't failing, and its heatsink is merely warm.

Graphiccelerator shouldn't be able to detect either, anyway.
 
So I'm trying to flash it in 10.5.8, and true to form, Graphiccelerator won't even open, and neither will ATI Multi Flasher. It's the same behavior in both safe mode and normal mode.

What a useless utility.

have you setup the tools properly?

I have found Gaphiccelerator and ATI multi flasher to work very well for me and has been very robust in a number of weird and wacky situations iv thrown it in

I suspect the issue lies in something on your end
 
have you setup the tools properly?

I have found Gaphiccelerator and ATI multi flasher to work very well for me and has been very robust in a number of weird and wacky situations iv thrown it in

I suspect the issue lies in something on your end

I followed the Read Me exactly as instructed. I ran Run Me First, ran ATI Multi Dumper, which then dumped the original ROM as normal, and then I launched the Graphiccelerator application as told. The icon appears in the Dock, then immediately closes. Same for ATI Multi Flasher.

Maybe it doesn't like the machine ('03 G5)? What am I missing?
 
I followed the Read Me exactly as instructed. I ran Run Me First, ran ATI Multi Dumper, which then dumped the original ROM as normal, and then I launched the Graphiccelerator application as told. The icon appears in the Dock, then immediately closes. Same for ATI Multi Flasher.

Maybe it doesn't like the machine ('03 G5)? What am I missing?

you have moved everything out of the DMG and into a folder on the root hard disk right?

(ATI multi flasher wont launch until you install a ROM into it)
 
Yes. For my case in particular, very.

Having the display connected to the 6600 in Slot 3, with the X1900 GT in Slot 1, Graphiccelerator ATI Multi Dumper wouldn't produce anything, and Graphiccelerator ATI Multi Flasher wouldn't even open (this being after 'Run Me First' is run, and a ROM is already loaded). Even now, after the Mac ROM has been flashed and the display running off the X1900, ATI Multi Flasher still won't open (with all prerequisites of course filled). Thus, rendering Graphiccelerator essentially useless, thereby forcing me to use my less than stellar only PCIe flashing solution.

(That, along with its favoritism when flashing ROMs and downright refusal to open when a certain card is installed that it doesn't like, doesn't give me a very good impression of the tool at all.)

Back to the story... I don't have a mini-tower PCIe PC. However, I've got a desktop-size OptiPlex GX620 that takes (horizontally) low profile graphics cards. So, the card would fit in its single PCIe slot. The trouble is, the bracket won't fit properly into its slot, meaning the card is forced to hang at a slight angle from the PCIe slot, leaving only the little latch on the PCIe connector from keeping the entire card from falling out. In the process, only barely being able to "stably" fit.

But there was another problem. As this computer was never designed to take calibers of GPUs that required a power cord, it wasn't able to power the card. But the G5 was, while the cord was only about a foot long.

So I ended up having to FLIP the OptiPlex vertically upside down and position the unit as up against the running G5 as much as possible in order for the cord to be able to reach the card while plugged into a separate machine. It was actually a very dangerous configuration, but the only possible route I had available (no thanks to Graphiccelerator).

Of course... at the time, I set that all up so I could back up the ROM BEFORE I tried actually confirming if ATI Multi Flasher would work with the card. Therefore, I had to devise this goofy setup to first back up the ROM with DOS 7.1, then put it back in the G5 only to be blown off by Graphiccelerator (I tried safe mode), then fashion it all BACK together to try flashing the ROM from DOS instead.

But it didn't end there. When I first pulled any video out of the card, there were artifacts and corrupted text characters in places where there should have been normal characters and text. Whether that was a problem with the card itself or the relatively unstable connection to the machine, I can't say for sure. Regardless, I had to try to work through graphically-corrupted DOS to attempt this flash, which I did end up accomplishing after another period of trial and error.

Anyway... Even given the minor reoccurring glitches with OS X, I believe this card is so far needless to say, happier when living as a G5 Edition.

All that said, it's important to keep in mind that most of my troubles stemmed from this individual card and my own lack of proper equipment. Anyone who has a Radeon X1900 GT R2 and a full-size PCIe PC should have no problems at all when flashing and using theirs.

The connecting of the card to another computer to power it while flashing it was something that I had to do with my Quadro 4500, as I didn't have a PC with a power connector for the card at the time. I just moved the PC next to my Mac Pro and powered the card that way. It was a bit of a kludge, but it worked out.
 
hmm. I've got one of those exact same cards, (haven't checked in a long while, but that thing used to run really nice, no green pixels or bars across the line either ;) nice little cards those.
 
@VanneDC What's funny is that the seller does not remember any artifacting from the last time he tested it, although he didn't try it immediately prior to the actual listing or shipment.

I have reason to suspect the physically unstable configuration I had to use when flashing it may have electrically damaged something.

If only Graphiccelerator actually worked, I could have avoided all that trouble and potentially foregone the hypothetical damage inflicted.
 
  • Like
Reactions: VanneDC
Being that I've got an MDD now, I bought a cheapo 9800 Pro for the G5, as the 9700 can just go into the G4 as an upgrade over its 9600.

So I'm trying to flash it in 10.5.8, and true to form, Graphiccelerator won't even open, and neither will ATI Multi Flasher. It's the same behavior in both safe mode and normal mode.

What a useless utility.

Now this is interesting.

I decided to take a look at the system log after failing to launch Graphiccelerator. Turns out that it's complaining about a problem with JavaVM.framework every time it opens, then closes.

I remember installing Java 6 and 7 on this OS, so I just download the switcher scripts, and run the Java 5 switcher. It doesn't work, which it has been seen to sometimes exhibit.

So then I just get ahold of a fresh JavaVM.framework from another machine's install, replace the folder, and sure enough, Graphiccelerator starts up.

So then, one flash & reboot later:

Picture 1.png


However, that still doesn't explain its failure to open when I tried flashing the X1900. That was attempted in Tiger, which I remember had an untouched vanilla JavaVM.framework. Granted, it would have been nice if the application was coded to at least tell the less-inclined user to check the system log if any errors are encountered.

That said, I do not remember checking its system log when I tried / failed, so that's potentially up for debate.

-

On the upside, I suppose we can walk away from all this now knowing that Graphiccelerator is NOT compatible with Java 6 and 7. It's one or the other.

Which is a good thing, because now nobody else has to go through this trial like I did in order to put two and two together.
 
After stumbling across this resource while in the middle of my badass Pentium III project, I thought back to what happened with the X1900.

If what the article says is indeed what went on, then maybe it was not a good idea to use an '08 Mac Pro GPU cable with this card, or at least not when the cable was plugged into the G5. If it does in fact damage certain components, then that explains the artifacting that the seller never experienced.

You learn something new every day...
 
what?? the g5 and intel mac pro gpu cables are identical??? arent they? ive inter swapped them lots before.. :D i dont think that was your issue. its possible some just had poor solder contacts, id probably put it down to that. 👍 hows the availabily of those x1900GSS/GT cards these days? maybe grab another and see if you can make it work without the artifacts?

Lol just checked on the bay, they are cheap as chips, there are 3 available from Germany at between 25 bucks and 40 bucks.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.