NVIDIA shows signs ... [2008 - 2017]

[neliz]

edit: toning it down..

on the wireless. yeah.. we had a couple of those two, replace wireless module and strike any key when ready.. I can safely blame INTEL for selling those wireless modules to HP ..
oh sh*T .. I ment COMPAL.

But seriously, nv sold a failing part and it's now looking into how and IF it can actually recover those costs from suppliers.

 

[trinibwoy]

.....lots of CAPS and ranting.....

What's your point?

BTW you can't do crowd control if there's no crowd....

 

[neliz]

What's your point?

BTW you can't do crowd control if there's no crowd....

Sorry man.. I'm too excited my wife's expecting a baby in a couple of days..

my point being that all we officially know is that nV is taking the blame so far and it's looking into it's supply chain to see if it can recover any costs. But wouldn't an official statement mention it if they were 100% sure it's not their fault? taking a ~150M cut on your forecast isn't made by some guy from the warranty department

 

[INKster]

Just had the opportunity to test three different BIOS'es on a HP dv2500 series laptop equipped with a Geforce 8400M GS GPU.

- The original factory-spec F.09 BIOS had a moderate-to-high fan speed about 50% of the time (hardly annoying, but perfectly audible nonetheless).
- The next one, F.13, lowered all the fan noise to the point of almost complete silence 90% of the time (tested it under load, and it still didn't make as much noise as the previous one at idle...).
- The latest BIOS version, F.2B (dated May 23, 2008) again raised the fan speed, but it's not nearly as loud as the F.09, and the speedup is a little more progressive, instead of the old "sudden kick".

The most noticeable difference to the F.13 BIOS is ironically the cooler temperature of the left palm-rest area, where the hard-drive resides, just below the ExpressCard/54 slot (the one i had here was a 5400 rpm Fujitsu SATA model).


Since the BIOS is dated from this past May, and the original factory BIOS already had high speeds, could this have anything to do with underestimated temperature targets of the intermediate lower-fan-speed F.13 BIOS by HP itself ?
I mean, last May this issue with Nvidia GPU's was completely unheard of, and yet HP restored the fan behavior back to the original laptop's factory-supplied defaults...

 

[trinibwoy]

Sorry man.. I'm too excited my wife's expecting a baby in a couple of days..

Oh, congrats!

my point being that all we officially know is that nV is taking the blame so far and it's looking into it's supply chain to see if it can recover any costs. But wouldn't an official statement mention it if they were 100% sure it's not their fault? taking a ~150M cut on your forecast isn't made by some guy from the warranty department

Yeah that I agree with completely and I don't think it's up for debate. But the premise of the thread is Nvidia's overall health. I don't think this particular issue has much bearing on their long term fortunes if it's a manufacturing defect isolated to one OEM and SKU.

The distinction here is that if there's something inherently flawed in Nvidia's QC process then it could have an impact on their business with other OEM's as well which would be a much more serious issue. Hence the niggling about whether it's something wrong with Nvidia's designs or procedures or just poor work by some 3rd party on a specific product.

 

[Geo]

Well, let's touch base again with the actual statement.

certain die/packaging material set

That doesn't sound like design or TSMC to me. That sounds like substrates and package mounting to me.

This is why I originally pointed at the possibilty that the explosion of volume in a short amount of time of their notebook business may have played a role here. It seems likely to me that they either had to bring online new capacity at existing vendors, or new vendors, or both, for those steps *after* TSMC and before delivery to OEMs, and it seems to me that's exactly where they are pointing with that wording.

The problem with that theory is the rumour-mill is claiming that nvidia is bad-mouthing TSMC behind the scenes, which wouldn't fit with what I just wrote.

 

[digitalwanderer]

B mf, B!

 

[Geo]

B mf, B!

"Brian, my friend, Brilliant!" ?

 

[digitalwanderer]

Yeah, that's it. You know how chummy me and Brian Burke are.

 

[AlphaWolf]

Charlie at the inq is now claiming its all G84 and G86 gpus.

 

[ShaidarHaran]

Nvidia doesn't manufacture anything. The "component" is built by third parties as far as I know. They're also not responsible for designing laptop cooling apparatus either.

I never mentioned manufacturing as it's irrelevant. NV is the IP-holder, the marketer, the master of disaster, yada yada. They are responsible for how their products perform in the market, in all categories. End of discussion.

Ironically, I feel the same about your position. silent_guy's input above seems to be in line with the way I think I went down. Your entire premise is that taking the $200m charge is an admission of culpability. That's a pretty pedestrian approach to the matter IMO.

I didn't realize occam's razor had gone out of fashion...

Exactly. The OEM-Nvidia situation is completely analagous to the Nvidia-Supplier relationship yet you have some weird myopic inability to recognize that.

I don't see the conflict...

With respect to the HP wireless issue it's definitely a hardware / manufacturing problem. A friend of mine's mother bought a DV6000 a while ago and got hit with same problem. HP took all the heat although it could have been some other third party's shoddy work to blame.

Through ~4k units I can account for ~100 WLAN card replacements. I was off by a factor of two or so in my previous estimation. What were you saying?

 

[ShaidarHaran]

food for thought

or FUD, if you think Charlie's a liar or simply being misled

 

[3dilettante]

I thought some blamed thermal cycling for the Xbox 360's RROD issue.
That wouldn't be an Nvidia-only problem, though perhaps both Nvidia and Microsoft were on the line with regards to thermals.

 

[Geo]

food for thought

or FUD, if you think Charlie's a liar or simply being misled

Okay, either Charlie or I (and maybe it's me) has a serious misunderstanding of what a substrate is as it relates to gpus. My understanding is it's added after the chip leaves TSMC. Further, that the viddy IHVs multi-source substrate procurements/mounting with the chip in that post-TSMC stage and there has been a history of shortages in the past.

Which one of us is right? The reason it matters is if I'm right, then its nonsense to say all G84 and G86 are bad using the logic he's using.

Edit:

http://www.tomshardware.com/news/flip,305.html
http://www.pcb007.com/anm/templates/article.aspx?articleid=17695&zoneid=4&v=
http://www.pcb007.com/anm/templates/article.aspx?articleid=17701&zoneid=4&v=
http://www.techpowerup.com/index.php?64812
http://forum.beyond3d.com/showpost.php?p=267337&postcount=22

Okay, I'm feeling better I'm not the one "unclear on the concept" here.

 

[Karoshi]

Okay, either Charlie or I (and maybe it's me) has a serious misunderstanding of what a substrate is as it relates to gpus. My understanding is it's added after the chip leaves TSMC. Further, that the viddy IHVs multi-source substrate procurements/mounting with the chip in that post-TSMC stage and there has been a history of shortages in the past.

Which one of us is right? The reason it matters is if I'm right, then its nonsense to say all G84 and G86 are bad using the logic he's using.

http://www.tomshardware.com/news/flip,305.html
http://www.pcb007.com/anm/templates/article.aspx?articleid=17695&zoneid=4&v=
http://www.pcb007.com/anm/templates/article.aspx?articleid=17701&zoneid=4&v=

Good point. Searched Wiki for substrate, the only relevant entry is:
http://en.wikipedia.org/wiki/Wafer_(electronics)

So it seems its the core's lower layer crystal (the pizza bread, where you put the 45nm anchovies, before you cut out the square servings), not the (green) mini pcb with the micro SMDs, pins and heatspreader attached where you then put the square servings on.

 

[AlNom]

hum... yes, substrate in EE terms would mean the silicon wafer, which is doped and etched (at least that's how I was taught ).

 

[Geo]

The Techpowerup article I linked after seems to be the most relevant. Tho that does say that TSMC provides "bump processing" having to do with the substrates. But if the substrates themselves are provided by someone other than TSMC, then I guess you could have some finger pointing about whether it's materials or QC by TSMC.

. . . industry sources in Taiwan believe the problem is most likely related to either the solder bumping process used by one or more of NVIDIA's manufacturing partners or the company's PCB substrate supplier(s).

Sources in Taiwan tell that the defective parts were the GeForce 8500M series mobile GPUs launched sometime in 2007. The problem was caused by related bump processing. Taiwan Semiconductor Manufacturing Company (TSMC), Advanced Semiconductor Engineering (ASE) and Siliconware Precision Industrial (SPIL) all provide bump processing services to NVIDIA.

Both ASE and SPIL denied knowing anything about the issue because the defective chips are older generation products.

 

[Geo]

hum... yes, substrate in EE terms would mean the silicon wafer, which is doped and etched (at least that's how I was taught ).

Sigh. There may be multiple "substrates" in the finished product that gets shipped to the OEMs. Maybe that's the confusion here.

When I've listened to viddy IHVs conference calls in the past, my impression is they called wafers "wafers" and reserved "substrates" for packaging stuff. But again, maybe I was just misunderstanding.

 

[Sxotty]

One random thought, though it is likely illegal somehow. Nvidia could take the charge stock goes down, buyback goes up, then recoup the charge from suppliers

Of course like I said that might be illegal I don't know all the ins and outs of that sort of law.

BTW I read charlies piece and it is complete crap, but that isn't unusual. His argument is desktop GPUs stay hot all the time as people leave their computers on instead of going hot-cold-hot like in laptops that have better throttling and are turned off more often.

 

[neliz]

BTW I read charlies piece and it is complete crap, but that isn't unusual.

Though he fired that shot back in April 2007 about the G84 and G86 gpu's being bad and actually hit something (a good year after the claim). Does that necessitate all the numerous times he spams the link?


oh... the link: http://www.theinquirer.net/gb/inquirer/news/2007/04/12/there-are-no-mobile-g846-problems

There is also no truth to to the rumour that notebook vendors who picked G8x for Santa Rosa based designs are now delaying those until they get the not-needing-a-fix A03 parts.

I just checked the quickspecs on the HP business notebooks and all new models are either released with intel IGP or a discrete ATI part (though 90% of the old discrete notebook GPU's are nV). This might change when the complete lineup is revealed in August though.