AMD Graphics Chips Could Last 10X To 100X Longer 150
An anonymous reader writes "According to a research report out of UCLA, released this morning, NVidia's high-lead bump packaging could last anywhere from 1/10th to 1/100th as long as AMD's advanced eutectic bump approach. (TG Daily has picked up the claim.) NVidia is currently in the midst of a $200M recall of bad GPUs, and the report suggests that the issue could be much deeper than NVidia's PR department would have us believe." The report lends credence to the strident claims of the Inquirer's Charlie Demerjian, which we discussed a month back.
Whatever. (Score:2, Informative)
I have tried three different ATI cards - all three ATI cards up and died on me within one year of usage, and I don't even play any graphics-heavy FPS games. After my most recent ATI card croaked (A Radeon X550, I think) I switched to Nvidia. So far my current Nvidia card has outlasted any ATI cards I've ever owned.
Re:More data please! (Score:4, Informative)
I have an old (circa 1980) roll of rosin-core eutectic solder, that I don't use any more because of the lead.
I also have a recently purchased some mgchemicals 4900-112G [mgchemicals.com], it is 96.3Sn, 0.7Cu, 3Ag, with a "no clean" flux. It works ok with my old soldering ironing, flows nicely, no idea how it does with tin whiskers. I'm not getting a lot of trouble with cold joints, and I do push my luck (lots of free-hand work, for instance, in-place soldering of LEDs for under-cabinet lights).
Re:Fans? (Score:3, Informative)
Fans can be replaced. I have replaced the fans on video cards on a number of occasions.
Re:Sweet! (Score:5, Informative)
"Eutectic" is a materials science word; it means (more or less, and I'm refreshing my memory from Wikipedia) a mixture (alloy) that does not separate/segregate into its original metals when it freezes; it has the lowest melting point, and passes immediately from liquid to solid phase. If, say, you have a solder that has more lead than the eutectic mix, when it freezes, it will segregate into (tiny) bits of lead and a eutectic remainder as it cools.
The advantage of a eutectic mix is that the melting point is lower, and when it is melted, it is all melted, and flows nicely. There are probably some caveats and quid-pro-quos for how it behaves in contact with other metals, which will certainly go ever-so-slightly into solution and change things.
Re:So Do nVidias last 3 months, or ATIs 30-300 yea (Score:5, Informative)
Most companies offer at least a year long warranty; if they have significant failures in that year, like 10-100x higher than normal, that may put too much pressure on their warranty policy.
And let's not forget nVidia's partners in selling cards (you know, all the non-nVidia nVidia cards). Those people may see high failure rates of nVidia parts, and all of a sudden using another chipset just got a heckuva lot more attractive.
So, the moral of the story is, there is no set 'time' that a card will die. It's not like after 10 months all of them will just conk out. But if there are higher failure rates than normal in their warranty period, not to mention harm done to their reputation, it could end up costing them greatly.
Re:Sweet! (Score:5, Informative)
There's also hypereutectic which means the opposite, obviously. A lot of pistons are cast with a hypereutectic alloy to keep costs down (forging is expensive) while increasing the melting point.
Sensationalist crap, On my /.? (Score:3, Informative)
Comment removed (Score:3, Informative)
Re:Down to the drivers (Score:5, Informative)
1] Download full CCC installer.
2] Don't install the CCC, just the drivers.
3] Download a copy of ATI Tray Tools [guru3d.com].
Re:So how can they sell these in Europe? (Score:3, Informative)
In the RoHS directive there is an exception for high-lead solder used in flip-chip style packages.
TFA mention a mixed scenario where you mix high-lead and eutectic solder. Not sure if that is excempt and also not sure that this combination have been used by nVidia.
Re:So Do nVidias last 3 months, or ATIs 30-300 yea (Score:2, Informative)
Practically speaking, using an exponential distribution means this:
If you can expect a card to last (on average) one year when it's new, then, given that it's N months old, you can still expect it to last one year. An exponential distribution has no history.
It works surprisingly well.
Re:More data please! (Score:3, Informative)
I use 70/30 and 60/40 both at home and at work. I have some 70/30 circuits I soldered up in 1974 that are still working.
High-tin solders are harder to work with: it doesn't flow as easily and doesn't seem to be willing to bead up on a pad, so if you try and self-locate a small package -- a BGA or LLP -- using solder, it won't: it'll just bridge all over the place. Thankfully, at work we provide engineering samples, not commercial stuff, so we don't have to worry about RoHS and can keep using leaded solder.
Tin whiskers are a worry: there are documented cases of them destroying valuable equipment. But it's the future of electronics, so we better get busy figuring out if tin/copper or tin/silver or others can reduce the tendency of tin to whisker in the first place.
Re:And if you already own a card? (Score:4, Informative)
That's the whole point of this fiasco.
NVidia began to update their manufacturing process in the middle of the life cycle of several chips.
They switched to eutectic pads and a new underfill material (which has a lower glassification temperature).
They stuck with high-lead solder because the bumps are laid out very early on and changing them would be a major undertaking.
Basically, this causes shit to break down faster because the parts are now much more susceptible to thermal stress.
NVidia knew there were problems.
Laptop parts started failing at very high rates.
OEMs knew about the failures in November of 2007, or earlier.
NVidia blamed OEM designs creating thermal issues.
NVidia offered to foot half of the bill (replacements, handling customers, fixes, etc.) with DELL and HP.
DELL and HP jumped at the chance to have the massive bill cut in half (this kind of offer is unheard of).
NVidia's "fix" was to crank up the fans with a BIOS update.
OEMs found desktop parts were failing at alarming rates as well.
OEMs were forbidden from speaking out about the real issue (lumped in there with that "we'll pay half" deal).
OEMs find out that their designs do meet NVidia's recommended thermal and electrical constraints.
NVidia continues to sell existing bad cards that are on shelves, and makes no mention of any of this to customers or retailers.
NVidia switches away from high-lead solder, completing the updates to it's manufacturing process, and may now be pumping out good parts.
But these new parts have not had any power distribution / control changes to accommodate the new solder material.
These new parts will likely have higher-than-normal failure rates as well.
NVidia does NOT designate the new parts in any way on the box.
NVidia does not designate the new parts in any (official) way on the actual hardware.
News comes out (last week) that NVidia based chipsets (motherboard chipsets, e.g. nForce) are bad too.
Re:To Quote from 'Count Zero'... (Score:3, Informative)
But to prove my point, electromigration happens, and IS an issue within ANY time span.
You point is valid. On-point, even. But what makes Nvidia's alleged misdeeds significant is that electromigration, along with other factors, makes the interconnects in the 8xxx series GPUs fail in an unreasonably-short time span. Without elaborate external mitigation strategies*, within warranty.
And that's the other factor in the significance in this story: Nvidia is alleged** to have made a point of downplaying, denying, avoiding, and misleading all participants about the significance of their substrate and solder choices in the 8xxx series. An example? Blaming a laptop manufacturer for designing laptops with ineffective GPU cooling--shifting the blame for GPU failure to poor system-level heat management. In point of fact, the manufacturer in question (HP) appears to have followed Nvidia design guidance, but that doesn't seem to have deterred Nvidia PR.
Executive summary: it appears (facts yet to be confirmed, wait for outcome of the class action suit) that Nvidia made poor materials choices in the 8xxx series, leading to premature parts failure in consumer use, leading to high warranty rates, leading to furious PR spin, leading to class action lawsuit.
*An example: HP laptop BIOS changes to "fix this heat problem" which force cooling fans on all the time [hp.com]:
So the fan runs all the time, and your battery runs down faster than it should, but at least that may buy a few months until the laptop's out of warranty, at which GPU failure isn't the manufacturer's problem.
**I hate using lawyer-speak, especially since I ain't one, but this whole debate is entering the realm of the courts, so I think everyone involved needs to be clear what they know versus what they've read or heard. Everything I've cited is in the latter category. There's your ObDisclaimer.