Is there a list of known dodgy ATI Radeon chips?

[LukeDavis]

Hello again! (People are likely going to get tired of seeing posts from me lol)

I'd like to try and compile a list of known bad AMD/ATI Chipsets and graphics processors.

While I do not possess the knowledge of some of the brilliant minds on this forum, I can at least mention a few chips I've encountered as being troublesome.

1. ATI 216-0752001 (RS880M/RX280M) found inside 2 HP laptops, one I've since deemed screwed, and swapped the board to an intel one, the other HP seems to be less hateful and runs fine.

2. ATI Mobility Radeon X1600 and the MacBook Pro, I believe these chips are used inside the MacBook Pro A1211 and A1150, I don't really see a point in trying to rescue these MacBooks as you can get a similar one with a Patched 8600M instead.

3. ATI Mobility Radeon X1800 - while I've Only heard horror stories about this chip coming from Dell XPS owners I'm not sure if I should or shouldn't include this chip as I'm not an expert.... so I'll put this one as a "maybe"


Asside from that I don't know if any other chips to be careful of.. feel free to chime in and hopefully this information is useful to others who collect old computer junk and not just me!

Once again I thank your for your time reading

[piernov]

There's no definitive list as it's not like Nvidia where the root causes of the design defect are known. All of them can fail, but some are known to have an extremely high failure rate especially in some systems, some of them have a more moderate failure rate or last longer but eventually die however they are used and some of them are mostly fine.

From my experience the very high failure rate is especially apparent on the GPUs from the Radeon HD 4000 (3000 could be included too but I see them less often) to the Radeon HD 7000 so that's why I often say 2008 to 2012. That's TeraScale/TeraScale 2 architecture.
The mobile ICs when used at high temperatures fail within less than a few years, which is what you see with iMacs and MacBooks from that era. But of course the failure can be seen on a lot of other all-in-ones and laptops. And desktop graphics cards, especially higher-end ones, fail as well but maybe less often.
Mobile ICs part numbers roughly fall into the 216-06xxxxx to 216-083xxxx range
The RS780/RS880 northbridges fall into that range too. From what I've seen they can last a bit longer but these days it's rare to see one that hasn't failed yet. They have part number 21x-06740xx for RS780 and 21x-075200x for RS880 (desktop starts with 215, mobile starts with 216).

But as you said, earlier ones commonly fail too, for example the Mobility Radeon HD 2000 series (first to use TeraScale architecture in R600 core) found in the 2007/2008 iMacs are known to be a problem, but I think to a lesser extent than the 2009-2011 iMacs. As for the X1000 series, I've seen dead ones (both desktop and mobile), but I've also seen a lot of working ones. There were also failure of the X600 in the iMac G5 but these were also plagued by bad capacitors and failing displays so they can easily be misdiagnosed.
The RS690 northbridges seem to not cause too many issues.
These have part numbers starting with 215 (desktop) or 216 (mobile), no hyphen, followed by some letters and numbers, can't tell a range as I don't know what the naming scheme means.
That's also when the high-end GPUs started pulling a significant amount of power (like going from 50W to 200W for the highest-end desktop part) and requiring good active cooling, which wasn't always well executed.

I don't remember seeing failed Radeon 9000, but I haven't seen too many of these. It looks like that's when they switched to FC-BGA starting from the Radeon 9500 with the R300 core.
Before that, GPU themselves rarely failed. If you see a random Rage or Radeon 7000 GPU it's likely fine, but on graphics card there's a small chance to have bad VRAM or bad capacitors.

An interesting thing to note though it the failure observed on the iBook G3, which, for once, I think is due to bad soldering between the IC and the PCB.

And then newer ones fail too, but it doesn't seem to be as bad (starting from GCN architecture). There was still a noticeable failure rate of the high-end desktop graphics card likely caused by a dead GPU from the Rx 200 and Rx 300 era, but then it seemed like the trend has faded away, and nowadays you get a lot of power delivery and VRAM failures instead, as well as PCB damage.

[LukeDavis]

There's no definitive list as it's not like Nvidia where the root causes of the design defect are known. All of them can fail, but some are known to have an extremely high failure rate especially in some systems, some of them have a more moderate failure rate or last longer but eventually die however they are used and some of them are mostly fine.

From my experience the very high failure rate is especially apparent on the GPUs from the Radeon HD 4000 (3000 could be included too but I see them less often) to the Radeon HD 7000 so that's why I often say 2008 to 2012. That's TeraScale/TeraScale 2 architecture.
The mobile ICs when used at high temperatures fail within less than a few years, which is what you see with iMacs and MacBooks from that era. But of course the failure can be seen on a lot of other all-in-ones and laptops. And desktop graphics cards, especially higher-end ones, fail as well but maybe less often.
Mobile ICs part numbers roughly fall into the 216-06xxxxx to 216-083xxxx range
The RS780/RS880 northbridges fall into that range too. From what I've seen they can last a bit longer but these days it's rare to see one that hasn't failed yet. They have part number 21x-06740xx for RS780 and 21x-075200x for RS880 (desktop starts with 215, mobile starts with 216).

But as you said, earlier ones commonly fail too, for example the Mobility Radeon HD 2000 series (first to use TeraScale architecture in R600 core) found in the 2007/2008 iMacs are known to be a problem, but I think to a lesser extent than the 2009-2011 iMacs. As for the X1000 series, I've seen dead ones (both desktop and mobile), but I've also seen a lot of working ones. There were also failure of the X600 in the iMac G5 but these were also plagued by bad capacitors and failing displays so they can easily be misdiagnosed.
The RS690 northbridges seem to not cause too many issues.
These have part numbers starting with 215 (desktop) or 216 (mobile), no hyphen, followed by some letters and numbers, can't tell a range as I don't know what the naming scheme means.
That's also when the high-end GPUs started pulling a significant amount of power (like going from 50W to 200W for the highest-end desktop part) and requiring good active cooling, which wasn't always well executed.

I don't remember seeing failed Radeon 9000, but I haven't seen too many of these. It looks like that's when they switched to FC-BGA starting from the Radeon 9500 with the R300 core.
Before that, GPU themselves rarely failed. If you see a random Rage or Radeon 7000 GPU it's likely fine, but on graphics card there's a small chance to have bad VRAM or bad capacitors.

An interesting thing to note though it the failure observed on the iBook G3, which, for once, I think is due to bad soldering between the IC and the PCB.

And then newer ones fail too, but it doesn't seem to be as bad (starting from GCN architecture). There was still a noticeable failure rate of the high-end desktop graphics card likely caused by a dead GPU from the Rx 200 and Rx 300 era, but then it seemed like the trend has faded away, and nowadays you get a lot of power delivery and VRAM failures instead, as well as PCB damage.

To add on to your thesis about the iBook G3 I actually have an IBM Thinkpad T42 which the fault ended up being the solder balls under the MR7500 chip, it's interesting to note that the T4X series failed from the board being flexed a lot...same applies to some Sony Vaio's. And I believe a Reball with lead solder should (knock on wood) prevent any future issues I just make sure these machines aren't transported with the lids open, my T42 is running beautifully now.

In regards to the X1800 in my Dell i believe it's failure was down to multiple factor's, I don't think it's the VRAM solder like most people claim. I believe unfortunately it is the Chip itself .. thankfully many people who own working examples have modded the fans to run at full and replaced the thermal compound which dramatically boosts the M2010s reliability.

Did the iBook G4s suffer any issues? I have one that's had some sort of fizzy drink spilt on it but that's otherwise largely a fixable issue.

And you'll be happy to know I've finally dumped the AMD boards out of both those HPs (I've left the G62 alone because it works for now) so you'll happily never have to hear me moaning about it ever again.

The Dell I think is the only Machine that worries me reliability wize. I discovered one of my Alienwares has an NV41 Quadro which is "okay" I think. And the other one has MXM which basically means I have free reign over what cards to use.

Once again, thanks for explaining things in such detail and having to put up with my Problematic machines for so long Your knowledge I have to say is the best! 🙂

[piernov]

The iBook G4 also has reliability issues but apparently it's caused by bad solder joints on the GPU VCore buck controller U28 as opposed to the GPU itself (see attached document). Funny how the MacBook Pro Retina 15" 2012 suffers from a similar issue even though it's a completely different design…

[LukeDavis]

The iBook G4 also has reliability issues but apparently it's caused by bad solder joints on the GPU VCore buck controller U28 as opposed to the GPU itself (see attached document). Funny how the MacBook Pro Retina 15" 2012 suffers from a similar issue even though it's a completely different design…

Interesting, I'll definitely check that out and see if it ever becomes a problem. Ironically I own the same retina and it works beautifully it was basically brand new internally (I probably should repaste it as it's likely never been repasted previously)

Do you know much about the NV41? According to techpowerup there was only "one" GPU under the NV42 number (6800 Ultra I believe) either they didn't find any others or that's the only chip. Gonna say it's as unreliable as the latter due to it basically being the architecture of "most" I'd the 6000 series. But as it's a 2004 laptop I'm keeping my fingers crossed.

According to some Clevo owners the Quadros in the D900T fair Better than other cards. It's just upkeep you gotta keep ontop of and since the M7700 is based on the same system it should be relatively the same. (Exept my motherboard has a large hole.blown in the edge thanks to a very angry trace catching fire.

a new board is expensive but I don't have much of a choice I think somebody offered to take a look at it for me but I'd be shocked if they're able to fix it...even if the system powers on the DJ board on the front of the laptop will never run again as most of its components were. Consumed by the fire.