Radeon 7000 Flashing Woes

[carbide]

Hello everyone.

For MARCHintosh I had been hoping to get a few Radeon 7000s I'd picked up flashed to work on my beige G3 tower and my power mac 9500. Of the three I picked up, two of them are identical Diamond models that appear to be more-or-less reference cards with 64MB of RAM and the third was a 32MB reference card.

My "plan" (such as it was), was to keep one of the Diamond cards with its stock ROM chip and flash a reduced ROM file for testing (and/or to use in a PC if I ever do a turn-of-the-century build), replace the ROM chip on the second Diamond card with a 128KB ROM chip (I'm not terrible with a soldering iron), and simply flash the 32MB card with a Mac ROM since it had a 128KB ROM chip to begin with. Predictably, the wheels fell off my plan almost immediately.

I started with a quick sanity check of all three cards. I have a G4 Sawtooth with its stock ATI Rage AGP card, so I dropped the Diamond cards into an open PCI slot (not at the same time) and booted it to make sure they'd at least be detected. In both cases, they popped up as a VGA-comptible controller with a vendor ID of ATi(0x1002) and a device ID of 0x5159. Then I tried the third card - the G4 booted to a flat grey screen sans apple logo and sat there until I turned it off. I backed up all three ROM files using an SPI programmer attached to a Linux VM then on a lark padded one of the Diamond 64KB ROMs to 128KB and flashed it to third card. Lo and behold, the G4 booted and it showed as an ATi card the same as the others (side note - is that information stored in ROM or elsewhere?)

Happy that all three cars at least worked enough to show up in OS X, I decided to try flashing the third card with the v226 ROM from the Mac Elite site using the Graphiccelerator tool and ATI multiflasher. I thought it was a fairly foolproof plan, no mucking about with a reduced ROM and whatnot and using the same tools that everyone else was. It disagreed though, and after flashing and rebooting I was in exactly the same situation as with its stock ROM - grey screen and no boot progress.

So, I set it aside and tried flashing one of the reduced ROMs to one of the Diamond cards. I got a warning from multiflasher that the "card ROM size (0010000) < new ROM size (0001DE00)". I went ahead and tried flashing anyway since a) I knew from the SPI programmer and looking up the ROM's part number that the ROM was 64KB and b) the rom file I was flashing was only 32KB. It flashed the ROM, then reported an error at the end. I tried booting with it installed and while the G4 booted, it didn't show up in the system profiler afterwards. I restored the backed up ROM from linux box and re-tested to make sure I hadn't killed it (I hadn't), then I tried flashing the reduced ROM from the SPI programmer (after padding it to 64KB) but it had the same issue when trying to boot.

To make matters more interesting, I took a closer look at the 32MB card and I think it may actually be from a Mac of some sort as it has a sticker on the back with an apple-ish part number, though the label is damaged. I skimmed though the dump of its ROM file after that discovery and it definitely has some lines with "AAPL" in it that are similar to the full v226 ROM, though the two files are not identical by a long shot.

I was thoroughly confused at this point, and some cursory searching didn't show anyone with my exact sequence of issues here or elsewhere, so I'm hoping that maybe someone can point out something I've done wrong? Some obvious step I failed a spot check on? I've included pictures of all three cards, a picture of the back of the suspected Mac card, and a dump of that card's ROM if anyone is interested.

TL;DR: Tried flashing some Radeons and it's all gone wrong, what do?

 

[Powerbase]

That 'Mac' Radeon was probably pulled from an Xserve.

 

[Phipli]

One of my Radeon 7000s suffered from early 2000s capacitor plague. It worked, but the image shimmered. It was a 7000 from an xServe, perhaps yours is the same but worse?

I removed a couple of capacitors and measured them... they had literally 0 uF capacitance. I replaced them all and now it works properly.

 

[treellama]

I never had much success with the reduced ROMs. I bought some 128K rom chips and successfully replaced one on a 64 MB radeon 7000 a few weeks ago, and flashed the retail 226 ROM on it. I'm not running Mac OS X though. Looks just like the silver heatsink ones on the left side of your picture.

 

[History_SE30_Dude]

What did you try flashing it in? I never had great luck flashing in a Mac, if I recall the Mac flasher was only compatible with a very limited amount of flash chips. Another issue I remember having was flashing with the appropriate clock settings for memory and GPU sometime cards were very finicky. I was pretty active on Smalldogs/MacElite way back when.

 

[Powerbase]

What did you try flashing it in? I never had great luck flashing in a Mac, if I recall the Mac flasher was only compatible with a very limited amount of flash chips. Another issue I remember having was flashing with the appropriate clock settings for memory and GPU sometime cards were very finicky. I was pretty active on Smalldogs/MacElite way back when.

I didn't even know there was a Mac-native way to flash them. I did mine back in the day, and the only way back then was the dos flasher way.

 

[carbide]

That 'Mac' Radeon was probably pulled from an Xserve.

That would explain it. Anything special about those that I may need to be aware of?

One of my Radeon 7000s suffered from early 2000s capacitor plague. It worked, but the image shimmered. It was a 7000 from an xServe, perhaps yours is the same but worse?

I suppose that's possible. I did a quick visual check of all the boards before slotting them in and didn't notice anything unusual around the caps. Do you know if these need to also be replaced with electrolytics, or can these be replaced with tantalums like logic board caps?

What did you try flashing it in?

I tried the ATI MultiFlasher included with the Graphiccelerator download I grabbed from the Garden (https://macintoshgarden.org/apps/graphiccelerator) and using flashrom from a Linux VM connected to an CH341A programmer (see output from the attached screenshot).

I've got a few other things I'm going to try doing, hopefully starting tonight:
- Replace the ROM chip on Diamond #2 with a 128KB model and flash with the 226 retail ROM file.
- Find a heatsink for the PC motherboard I found in my parts bin with PCI slots on it and do some more sanity checks of the cards (incl. flashing the suspected Xserve card with a PC ROM)
- I'll test the diamond card with replaced ROM chip with the padded PC ROM as well, and if that works I'll probably replace the ROM chip on Diamond #1 too.
- Break out a magnifying glass and take a closer look at the caps on the Xserve card

Side note: when testing these cards post-flash, I've been first testing them with the Rage AGP card still in the G4 and with monitor plugged into its VGA port. I've also tried booting it with just the (suspected) Xserve card and no Rage card with the result being the monitor never coming out of power save. Any other permutations of that configuration I should test? Should I try just dumping the Xserve card into the 9500 along side its existing card and seeing what happens? It has an iXMicro 3D of some sort in it right now.

 

[Phipli]

I suppose that's possible. I did a quick visual check of all the boards before slotting them in and didn't notice anything unusual around the caps. Do you know if these need to also be replaced with electrolytics, or can these be replaced with tantalums like logic board caps?

No, this is a different thing. The failing caps from 68k (and some early PPCs) is a quality / age thing. In the early 2000s there was some industrial espionage that went wrong. Someone stole an electrolyte formula, but it was incorrect and resulted in very early failure. Caps that suffer from what's known as Capacitor Plague failed very early, within a small number of years. They fail dry - no visual symptoms. They just stop being capacitors. I'd replace them with electrolytics to be honest, in this case. Good ones. But you probably could use tantalums.

 

[treellama]

What did you try flashing it in? I never had great luck flashing in a Mac, if I recall the Mac flasher was only compatible with a very limited amount of flash chips. Another issue I remember having was flashing with the appropriate clock settings for memory and GPU sometime cards were very finicky. I was pretty active on Smalldogs/MacElite way back when.

I have flashed two with Multiflasher on a G5 running Mac OS X. One a 32 MB PC card that already had a 128K rom chip, another one of those 64 MB ones that are on ebay for $25 or so. I had to put a bigger ROM chip on that last one, I got the part number from bbraun's radeon 7000 page.

 

[Unknown_K]

One of my Radeon 7000s suffered from early 2000s capacitor plague. It worked, but the image shimmered. It was a 7000 from an xServe, perhaps yours is the same but worse?

I removed a couple of capacitors and measured them... they had literally 0 uF capacitance. I replaced them all and now it works properly.

Probably got baked from the heat.

 

[Phipli]

Probably got baked from the heat.

Erm... Are you trying to tell me that the well document thing called capacitor plague didn't happen?

That all capacitors on a graphics card that runs pretty cool in a well cooled case, away from the PSU and processors, were baked to significantly over their design temperature, uniformly over the whole board?

I have to disagree, given it is from right in the middle of the era of well documented issues, and is the only time I've ever seen every single electrolytic cap on a board have effectively zero capacitance.

 

[Unknown_K]

Erm... Are you trying to tell me that the well document thing called capacitor plague didn't happen?

That all capacitors on a graphics card that runs pretty cool in a well cooled case, away from the PSU and processors, were baked to significantly over their design temperature, uniformly over the whole board?

I have to disagree, given it is from right in the middle of the era of well documented issues, and is the only time I've ever seen every single electrolytic cap on a board have effectively zero capacitance.

No, it did happen but mostly on motherboard capacitors. Those SMT capacitors on video cards (before they switched to solid polymer) were liquid filled and putting them into compact servers boards that ran hot (G5 Xserve for example) just causes them to cook. Same thing happens to large motherboard capacitors next to the CPU heatsink on older Athlon XP/P4 machines. You have unstable ingredients that won't stand up to heat.

 

[carbide]

Sooooo.... I'm even more confused now.

I was able to remove and replace the ROM chip on Diamond card #2 with no issues. Flashed it with the 226 ROM and.... the exact same thing started happening that did with the Xserve card. G4 powered up, moved no further. I flashed it with the 208 ROM I found here - http://www.synack.net/~bbraun/ati7k.html - and the same thing happened.

At this point, curiosity got the better of me and I tossed it into my 9500, but the exact same thing happened. Computer bonged, existing display adapter never sent a signal to the monitor. I pulled it and flashed my padded version of its stock ROM to it, put it back in the 9500, and it immediately booted as normal. This is what system profiler reports on it (yes, I know it's a mobile pic, but transferring a screenshot would have been a pain. Setting up a Netatalk server is on the To-Do list):


I'm open to any suggestions, crazy ideas, etc.

 

[History_SE30_Dude]

Have you tried any of these others? There is several different reduced ones. http://themacelite.wikidot.com/wikidownloads2

 

[Powerbase]

Just as an aside, I've always wanted to find a Radeon 9100 pic card and try flashing that. You can never find the things though.

 

[carbide]

Have you tried any of these others? There is several different reduced ones. http://themacelite.wikidot.com/wikidownloads2

The v226 ROM I'm using is from there and I've tried two out of three of the reduced ROMs listed. I haven't tried the 208 reduced one, I may give that a go tonight.

As for testing any of them in a PC, I haven't found the LGA-1366 mounting bracket for my spare heatsink, though I didn't have much time to do it last night after doing the ROM surgery on Diamond#2. Hopefully I'll have better luck tonight.

 

[croissantking]

No, it did happen but mostly on motherboard capacitors. Those SMT capacitors on video cards (before they switched to solid polymer) were liquid filled and putting them into compact servers boards that ran hot (G5 Xserve for example) just causes them to cook. Same thing happens to large motherboard capacitors next to the CPU heatsink on older Athlon XP/P4 machines. You have unstable ingredients that won't stand up to heat.

FYI I bought the same type of card as @Phipli, and it had the identical problem - a shimmering picture that was solved with a recap.

The card had a Sun ROM, and appeared to be new old stock - no signs of use whatsoever. I’d vote for ATI having used a bad batch of capacitors rather than it having been a heat issue, as in my case, I literally think it had never been used.

 

[Unknown_K]

FYI I bought the same type of card as @Phipli, and it had the identical problem - a shimmering picture that was solved with a recap.

The card had a Sun ROM, and appeared to be new old stock - no signs of use whatsoever. I’d vote for ATI having used a bad batch of capacitors rather than it having been a heat issue, as in my case, I literally think it had never been used.

Could be, there are more than one way these things die.

 

[joevt]

So, I set it aside and tried flashing one of the reduced ROMs to one of the Diamond cards. I got a warning from multiflasher that the "card ROM size (0010000) < new ROM size (0001DE00)". I went ahead and tried flashing anyway since a) I knew from the SPI programmer and looking up the ROM's part number that the ROM was 64KB and b) the rom file I was flashing was only 32KB. It flashed the ROM, then reported an error at the end. I tried booting with it installed and while the G4 booted, it didn't show up in the system profiler afterwards. I restored the backed up ROM from linux box and re-tested to make sure I hadn't killed it (I hadn't), then I tried flashing the reduced ROM from the SPI programmer (after padding it to 64KB) but it had the same issue when trying to boot.

card ROM size 0010000 is 64K.
new ROM size 001DE00 is 119K.
Doesn't that mean you were trying to put a 119K rom on to a 64K rom chip?

 

[carbide]

card ROM size 0010000 is 64K.
new ROM size 001DE00 is 119K.
Doesn't that mean you were trying to put a 119K rom on to a 64K rom chip?

It means I'm bad at reading the calculator I was using to convert hex to decimal. Having said that though the reduced ROM files I tested were 64KB and 28KB, so the error is.... in error anyway?

Unless Graphiccelerator was failing to pass the correct ROM file to ATI MultiFlasher and was trying to pass it the full sized file. Bah. I was planning on retesting by using the flashing tools from OS 9 instead of 10.4 anyway, maybe that'll change something.