Is This Typical GPU Failure ?

[mmartell]

Neighbour gave me an iMac Core 2 Duo White gratis which was basically stuck at Apple logo which would eventually disappear leaving a white screen. The white screen has a type of herringbone pattern in it while the Apple which is overlayed seems proper.

After fiddling some this unit seems to have multiple issues which doesn't sit right. The hard drive spins but system will not boot to any OS that is presumably on it. The dvd rom makes some god awful noises but will boot Win7 installation disk and Memtest disc but not the OSX Leopard that came with it. The aforementioned discs have video errors and Memtest indicates errors on both 512MB sticks.

The power supply looked good as did the caps on the logic board which were removed for a solid cleaning.

If I were to guess I would say there are some bad caps on the MB at minimum. What you think ?

[reaper57]

Re: Is This Typical GPU Failure ?

Looks like the gpu. Check the caps and when you are sure they are good, install windows, but leave the video drivers for last. If it gives you a BSOD, when you are installing them, then it's probably the gpu.

[SteveNielsen]

Re: Is This Typical GPU Failure ?

I would test the RAM in another computer.

[reaper57]

Re: Is This Typical GPU Failure ?

I would test the RAM in another computer.

Yeah, I forgot about that. If it has onboard graphics it's using the ram as vram.

[SteveNielsen]

Re: Is This Typical GPU Failure ?

Even if it isn't check it because it threw up errors in the iMac.

[Th3_uN1Qu3]

Re: Is This Typical GPU Failure ?

If the memory tests bad with Memtest then definitely replace that first. See where that gets you. Though i have to say that pattern looks more like a GPU than a memory failure.

[mmartell]

Re: Is This Typical GPU Failure ?

The memory errors strike me as a false positive only because it would seem rare for BOTH to fail before the owner says to himself "hmm there seems to be a problem here." But I will try to find another stick or even a device to swap this ram into to rule out first.

[RJARRRPCGP]

Re: Is This Typical GPU Failure ?

Good chance of being fractured solder joints, if you can't detect bad caps.

[mmartell]

Re: Is This Typical GPU Failure ?

Yeah, I forgot about that. If it has onboard graphics it's using the ram as vram.

It's not onboard, it's replaceable but prohibitive and has dedicated ram.

[mmartell]

Re: Is This Typical GPU Failure ?

Good chance of being fractured solder joints, if you can't detect bad caps.

Yes but it's a pretty big board lol. And of course once it's installed you have very limited access to the front side and NONE to the backside which is actually the top side.

[mmartell]

Re: Is This Typical GPU Failure ?

I procured another 1GB stick this evening and symptoms are identical. Same on screen artifacts and lots of errors in Memtest.

[reaper57]

Re: Is This Typical GPU Failure ?

Is the video card on this one removable or is it soldered to the motherboard?

[mmartell]

Re: Is This Typical GPU Failure ?

Removeable card with an MXM interface.

[Th3_uN1Qu3]

Re: Is This Typical GPU Failure ?

That isn't too bad then... What card is it?

[mmartell]

Re: Is This Typical GPU Failure ?

Nvidia 7300 GT or the 7600GT probably the former.

It's bad because this is common enough that they are hard to come by.

[Th3_uN1Qu3]

Re: Is This Typical GPU Failure ?

I think you can upgrade or even replace it with an ATI. I know there are several cards floating around that are compatible with iMacs. Either way, avoid the 8000 series and i wouldn't replace it with the same model, either. The 7600s held up a bit better, but the 7300 all failed. It wasn't so bad as with the 8400 and 8600, but they were still crap chips.

[mmartell]

Re: Is This Typical GPU Failure ?

From what I've read can only use one of the two cards listed in previous post. Others which fit simply won't boot. I think the chip is fine but needs reballing. It's in a location where I can't really put heat on it just to test that theory.

[mmartell]

Re: Is This Typical GPU Failure ?

Well Puppy 5.01 loads up nicely and works properly. Does this have to do with loading a basic video driver ?

Now I'm wondering if one of the gpu ram chips is bad and not the nvidia chip. But tbh I don't know how all that stuff works...

[Th3_uN1Qu3]

Re: Is This Typical GPU Failure ?

From what I've read can only use one of the two cards listed in previous post. Others which fit simply won't boot. I think the chip is fine but needs reballing. It's in a location where I can't really put heat on it just to test that theory.

I don't know about compatibility, never been much into Macs, and this is one of the reasons. They're generally a pain in the ass to take apart.

However, trust me when i'm telling you it is a waste of money reballing this chip. I understand that you do not have the required equipment. If you are going to pay someone for a job, at least have him do a proper job and install a brand new BGA chip. That way you will never have to worry about it again.

Well Puppy 5.01 loads up nicely and works properly. Does this have to do with loading a basic video driver ?

Yes, absolutely, and you may get a couple months of use like that before it dies completely.

Now I'm wondering if one of the gpu ram chips is bad and not the nvidia chip. But tbh I don't know how all that stuff works...

On those generations i've never seen bad VRAM. Never ever ever, not even once. Doubt anyone else has seen that, either. It's the GPU, period.

The chips die due to a (now) well-known fabrication defect (or rather several engineering mistakes piling up and leading to a big oopsie) that plagued all nvidia BGAs made from 2005 to 2009, and perhaps even earlier. Some held up better than others due to lower heat output, but ALL of them have suffered from it.

I have here a Dell D620 with the nVidia Quadro NVS110M which is basically a rebadged 7300 Go with DDR1 memory instead of DDR3 (ouch), and calling itself a "workstation" chip. It had never been worked on before. Sometimes you wonder where people keep these things.

I know the GPU was bad as i did another one of those a couple years back, and it took two reflows and stupidly high temperatures to get it to work - and who knows how long it lasted. I played with it a bit and it would work fine if kept warm - think "warm" as in leaving the heatsink off the GPU entirely. But it also crapped out immediately after being allowed to cool, so this seemed maybe like actual bad soldering instead of a bad GPU... maybe.

I did briefly think about reballing it but i was convinced i was wasting my time, and i was right. After reflowing it with all the proper temperatures and time it showed a nice clean Dell logo on boot-up - but after the "Starting Windows" screen it just turned blank. It worked in Safe Mode and with the driver uninstalled.

I then took just my hot air station and heated the die. It booted fine with the driver. Worked for a few quick tests. Then after i shut it down, it would go to a blank screen again after loading Windows. Got mad at it and cranked up the hot air temperature to max, and blasted the die with the hot air for about 4 minutes, then immediately took a heatsink and pressed that against the die (just slight pressure, not bending the whole board), holding it until it cooled down.

It works now, but nobody knows how long it will. I have also undervolted it to 0.84v from 1.03v and replaced the thermal pad with 3mm thick copper (there's a lot of distance between GPU and heatsink on the D620) but it doesn't help much, as the fan in the D620 is programmed to run only when the notebook is about to catch fire. If i max out the fan manually the GPU sits at 55C in FurMark, which is out of the danger zone, but that is without loading the CPU as well. A completely stock cooled D620 easily creeps up into the high 80s and even low 90s when loaded heavily, and if only the GPU is loaded, it allows the GPU to go up to 76C before turning on the fan. These affected nvidia chips will die (just a matter of time) if exposed to temperatures above 60C as reported by their internal sensor. And keeping them under 60C is in most cases impossibile.

Reflow, reball, pointing a mini torch at it, putting it in the oven... it's all the same for these crappy nvidias, as anything you can do to them with heat resolders the connections on the inside of the chip, only to have them fail again due to their poor materials choice and high operating temperatures. It is completely pointless to reball such a chip. I have said it several times before and others have said it too. I've had chips pulled from boards with other causes of failure die after 3 months, and i've had four reworked dv6000 boards die just sitting on a shelf unused for 2 months or so. These nvidia chips were that bad.

If you want a solution worth paying for, replace with new chip manufactured 2010 which uses a revised material set and also runs significantly cooler. They haven't been in production since, there are chips with newer datecodes but most of them are fake. I've been burned too. Otherwise you're better off cooking it yourself, you'll get a few months if you're lucky.

[RJARRRPCGP]

Re: Is This Typical GPU Failure ?

Either way, avoid the 8000 series and i wouldn't replace it with the same model, either. The 7600s held up a bit better, but the 7300 all failed. It wasn't so bad as with the 8400 and 8600, but they were still crap chips.

At least with desktop versions, 7-series are notorious for having FZ caps!

So I get a not very good feeling when you say that 7 series is better than 8 series, lol.

But, thank God I never heard of those caps being on laptop boards.

Even though 8400 and 8600 look like they're lousy at best. (probably 8500, too) (not Radeon 8500)
Thus, leaving only 8800s, which are for a desktop, lol.

For Nvidia, I suggest to get a 9-series or later.