Apple Knowledgebase says G92s are defective
Bumpgate You wanted proof...
IT LOOKS LIKE Nvidia was being a little less than honest in screaming about desktop cards not being defective, at least according to Apple. The Apple knowledgebase, Radar, says that G92s are bad in desktop configs as well.
Our editors don't like us to use the phrase 'lying sacks' when referring to corporate entities, or their agents, so we will refrain this time. If you are not familiar with the story so far, the summary is that Nvidia announced a 'small' batch of bad chips. We said there was more. And more. And more. Nvidia dutifully denied it. We said it wasn't confined to laptops, Nvidia again denied it. We dug up why (Parts 1, 2, 3), and they downplayed it.
This sad state of denials was at it's most fevered pitch when we said that desktop parts were affected. Nvidia really doesn't want you to know that this is happening, especially if you are a stock analyst, it may cost the firm a lot of money.
The problem is that bridging the gap between defective chips and hardware failures is a tough thing to prove unless you have access to pools of reliability and warranty data. A single incident does not a trend make. We don't have all that data. Apple, however, does. Which is why its knowledgebase is so damning to Nvidia, take a look at these two entries.
(Note: numbers have been scrubbed to protect our sources)
-----------------------------------------------------------------
XX-Sep-2008 XX:XX PM xxxxxx :
Request to qualify Nvidia G92 GPU bump material change for M86 for known Nvidia bump crack issue in order to support MPS and enhance package robustness.
APN: XXXXXXXX
06-Sep-2008 03:27 PM xxxxx :
NVIDIA will transition from using high-lead solder (95%Pb/5%Sn) to eutectic solder (63%Sn/37%Pb) flip-chip bump material for the G92 product family. During the transition period NVIDIA will be supplying both high-lead and eutectic bump until inventory is depleted. No other materials are being changed.
-----------------------------------------------------------------
radar XXXXXXX Request to qualify Nvidia G92 GPU with bump material change for K3 CTO option
--
XX-Sep-2008 XX:XX PM xxxxx :
Request to qualify Nvidia G92 GPU bump material change for K3 CTO for known Nvidia bump crack issue in order to support MPS and enhance package robustness.
G92 MXM card part number: 631-0440 (Sam memory)/ 631-0556 (Hyn memory)
XX-Sep-2008 XX:XX PM xxxxxxx :
NVIDIA will transition from using high-lead solder (95%Pb/5%Sn) to eutectic solder (63%Sn/37%Pb) flip-chip bump material for the G92 product family. During the transition period NVIDIA will be supplying both high-lead and eutectic bump until inventory is depleted. No other materials are being changed.
-----------------------------------------------------------------
For those not fully up on Apple lingo, M86 is a tower machine, not a laptop, not even a slimline anything, and is in no way related to a portable box. K3 is the iMac which is sort of related to a laptop. The G92 is a desktop part, the G92 MXM is, but these are two different parts, and the MXM version may be a desktop part in a small form factor, that is unclear.
In any case, what you are seeing is Apple saying that desktop G92s, aka 8800GT/9800GTX/GT1xx, have a "known Nvidia bump crack issue". This one is going to be really tough to spin, especially after Nvidia's track record of honesty with Apple.
So there you have it, hard evidence that desktop parts are bad. Hard evidence that G92s are bad. Just like we told you. This also is strong corroborating evidence that the entire run of Nvidia 65 and 55nm parts with high-Pb solder and low-Tg underfill are defective.
As you can see from the second radar entry, changes of this magnitude mean a significant cost for the OEM in addition to being a royal pain in the ass. You simply do not change materials like this unless you absolutely have to, period. Nvidia has to. µ
Update:
After this was written, Nvidia's Mike Hara got back to us about the the G92. We
asked him if there were any known problems with the G92. His response is as
follows:
There are no known problems with G92s. They do use the same material set, but as I stated before, and will state again, they are not defective. They are NOT subject to the same thermal designs and usage patterns that our other parts have been subjected to (in notebook systems). Essentially, they do not operate anywhere near the critical temperature that causes the other parts to have abnormal failure rates.
There you have it, according to Nvidia, there are no problems with the G92. According to Apple, there are. Read this again carefully, I will leave it to the reader to decide on what to make of it. µ

Comments
So at Apple, are they masochistic?
You think they would use nVidia chipsets and GPUs if this was such a huge issue?Hurt me plenty!
Phew/
Well finally, about time we saw it said in 'plain' english and not only circumstantial evidence.I feel relieved, for some reason.
where is the link to desktop part failure at apple?
so where is the link to the apple article or a screenshot???You point to apple homepage that clarify defect of laptop chips. Not desktop chips.
The vendetta part 25
Charlie, as much as I love reading the inquirer - I think I can safely speak for at least a few thousand people who read the inquirer when I say "we get the point".Now please next time AMD offer another cheque to handle their PR for them behind closed doors I suggest you perhaps take time to reflect on your job, which i am hoping it actually a journalist. Take a trip back through memory lane when you actually wrote stories worthy of publishing.
Typo
"There are no known problems with G92s" that we would like to share with you.He just forgot to add the second part of the sentence, an honest mistake. ;-)
Wow!
This AND the Google purchasing Valve story? Wow... The INQ is right about everything.Bang Bang Bang goes Charlie's drum
At what point do you guys start to sound hysterical about all this?We've got it. NV tried to hide the fact that they are having some problems. Nasty problems. How many stories will there be on this? I think the next story will be story number 11 or 12?
While I appreciate the coverage of a story that is pretty significant, about a less than honest vendor, what stories are you not covering while you are beating this horse?
To 'no fanboy'
The Apple article that was linked to states the following..."In July 2008, NVIDIA publicly acknowledged a higher than normal failure rate for some of their graphics processors due to a packaging defect. At that same time, NVIDIA assured Apple that Mac computers with these graphics processors were not affected."
CD just said that nVidia weren't hoenst with Apple about the situation and linked to that knowledgebase article. It's quite a funny read considering how Apple portray themselves as the saviours to their customers by saying "However, after an Apple-led investigation, Apple has determined that some MacBook Pro computers with the NVIDIA GeForce 8600M GT graphics processor may be affected."
Hooray for Apple! :D
New MacBooks Affected?
So, now that the new line of MacBooks are coming out today, apparently all with nVidia chipsets/GPUs, is this a case of Apple being locked in to a design decision made before they were aware of the scope of these failures, or do you think nVidia really has solved the problem by changing the materials alone (and not the design?) Or, are the graphics in the new MacBooks a new design that is no longer subject to the same failures? In all the 'new MacBook' hoopla over the past two weeks, I have seen NO mention of anyone anticipating a chipset/graphics disaster with Apple's bread-and-butter computer line...nvidia reference design vs apple design
I think they're talking about different situations. nvidia is probably right about all the normal 8800gts/9800 g92 cards, as they are using a good (ie expensive) cooler that keeps the cards cold enough - I'm a little worried about the original reference 8800gt though, as it got pretty hot. And ofcourse some special apple designs may have issues as only keeping as cold as originally specified and not as cold as nvidia later found out was neccesary...circumstantial?
more like downright unproven. a change of material hardly proves that all of the old units are faulty.Burn Nvidia to the ground
Nvidia killed my father, then raped my mother. ATI fanboys unite, lets declare war on all nvidia products. Nvidia is the devil, and this is the worst thing to ever happen forget about war, poverty, or anything else important. Nvidia has committed the worst crime of all time. Charlie you are my hero, you are so smart. If only everyone could be as smart as you.Honesty is right out the window...
"There are no known problems with G92s."When is Nvidia going to learn that lying and covering up problems is going to cost you in the long run?
How hard is it to admit that you screwed up and will correct the problem?
not the "smoking gun"
it's extremely common in manufacturing (of anything) to incorporate a process or material change across product lines, even though only one specific product has a problem; it simplifies many things and reduces cost by not having to manage many different processes. that these changes are applied across-the-board is not evidence that they're all bad.And yes, whenever a material or process change is made, you damn well make sure to inform your customers even if it's not a quality fix.
Of vendor choice and MXM
Regarding the difficulty of switching video cards, two problems come to mind:1) How do you de-rail and re-rail existing design/validation tracks without significant expense?
2) Sadly, the current choices for high-perf GPUs are slim. There's ATI, but honestly, ATI still has driver issues despite having superior silicon. ATI is working through them (and WILL get better), but that's a little ways into the future.
@charlie: on a side note, anything MXM sounds like a mobile part by definition (Mobile pci-eXpress Module). unless this is a quasi-mobile part getting shoe-horned into a mac-mini?
Is anyone buying NVidia parts with confidence they will last 3+ years?
After what's happened these past months, do any of you have confidence in NVidia?Would you tell the people you care about (family, friends and customers) to buy NVidia products and for them to have as much faith in them as they can when buying a TV?
ATI are not perfect, but NVidia are really gaining an image of "screw the customer when things get tough".
Some people never admit it
So first you guys said he was downright wrong, then you claimed he had no credible evidence to back him up, and now you're bored and it's supposed to be yesterday's news.In short, some people will simply never admit that Charlie was RIGHT ALL ALONG and RIGHT AGAIN to BELABOR THE POINT.
You're tired of reading this kind of article ? Don't make me laugh - nobody made you click on the link.
Go huff off somewhere else. Charlie was right, period, and he is perfectly right to continue shoving that in your face given your complete lack of acknowledgment of your mistaken beliefs.
It's a sport for some to take Charlie down, but those who play seem to be sore losers.
Fair play, know what that means ? It means you win some, and you lose some. When you win you don't crow about it, when you lose you congratulate your opponent.
But that is sooo last century, hmm ?
Er, Pascal...
I'm not saying that they *can't* be defective, I'm only saying that this is _not_ proof that they *are.*HP and nVidia screwed us
Corporate America, HP and nVidia, are screwing us over and hoping that nobody will notice. I got hit twice, once with a desktop nVidia 8600GT video card going bad and again with an expensive HP Pavilion notebook with nVidia 7600 GPU going bad. Both are totally dead, I now have worthless hardware junk and they have close to $2000 of my hard earned money. HP extended some warranties just to look good in the press, most are left in the hopeless dust. HP support is so horribly bad - don't buy anything from them. I know that there are a lot of people in the same situation as me and the list is growing.Nvidia defective GPUs
please...be aware of the facts...there are thousands like me having problems with these Nvidia defective chips...and I can assure you...HP is not covering all of its laptops models affected...http://forums11.itrc.hp.com/service/forums/bizsupport/questionanswer.do?threadId=1274587&admit=109447626+1224512935608+28353475
Why is Nvidia SO RELUCTANT to provide a comprehensive list of its defective GPUs...think about that...
We the affected customers are being victims of DAMAGE CONTROL in some sort of confidential business agreement between Nvidia and OEMs....
Mac Pros and 8800 GT
Personally I've seen several of these fail, but the only annoucement is for MacBook Pro and 8600's.Are Nvidia still denying that there are problems with 8800's!!!! I can't get any answers through my suppliers, just replacement cards which are the same models as those being shipped back!
Any update would be appreciated.