backtop

Print E-mail del.icio.us 37 comment(s) - last by PrinceGaz.. on Nov 7 at 2:51 PM

Voltage dip causes artifacting and restarting
Isolated issues with some GeForce 8800GTX graphics cards have forced NVIDIA to stop shipment on some launch cards

A memo published by NVIDIA has circulated through GeForce 8800GTX launch partners claiming some initial G80 cards have been stopped from reaching resellers.  Parts of this memo were republished on Bit-Tech earlier today.  Specifically, the memo states that all vendor GeForce 8800GTX based cards will generate artifacts and cause the system to restart after 30 minutes to 2 hours of operation. 

The artifacts and crashing are a result of an incorrect resistor value. According to the document, a 50K ohm resistor was installed instead of a 30K ohm and results in artifacts during heavy operation.  Nevertheless, the document recommends checking the MOSFET to ensure the proper ones were installed and replace them if necessary.

Any manufacturer shipping cards based on the NVIDIA reference design with the reference bill of materials is not affected.  However, several manufacturers have come forward saying their original batch of cards are affected, including EVGA, Gigabyte and MSI.

EVGA has released a statement to DailyTech claiming "EVGA has been working very closely with NVIDIA to correct an error in an early [Build of Material] list for internal testing, sample boards and a few review units.  There is no issue with production boards and this will not delay the launch."

Several other manufacturers, including XFX and ASUS, have confirmed to DailyTech that their GeForce 8800GTX cards are unaffected and will ship at launch without problem.

NVIDIA has issued an official statement regarding the defective cards:
Some GeForce 8800 GTX boards that were built through our contract manufacturer had a simple BOM error - wrong resistor value. GeForce 8800 GTS boards are not effected by this.

These GeForce 8800 GTX boards were shipped to our Add-In-Card partners.

We have been working with them to pull these back and change the resistor to the correct value.

We believe we will still be able to hit our hard launch this week (Nov 8th) with the new GeForce 8800 GTX boards.

This is a testament to our execution as well as the execution of our Add-In-Card partners.

Adam Foat, Product PR Manger, Northern Europe.
DailyTech previously explored the performance of NVIDIA’s upcoming G80 GeForce 8800GTX. The card used did not exhibit artifact or restarting issues.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Not refurbished, still new
By UnstableHero on 11/6/2006 1:40:47 PM , Rating: 2
I think if the product didn't touch consumers then it can be worked on without it being sold as refurbished




RE: Not refurbished, still new
By aos007 on 11/6/2006 1:57:57 PM , Rating: 2
Exactly. Changing a resistor is trivially easy and can be done fast. Even though it's all manual labour, given the price of these boards, no point throwing them out. It's a bad publicity though - especially if it didn't actually hit retail. EVGA doesn't need any more bad news this year, do they? They already had trouble earlier with cards overheating (my 7600GT was one of them).


RE: Not refurbished, still new
By Misty Dingos on 11/6/2006 2:42:47 PM , Rating: 1
Replacing surface mounted components requires specialized equipment and technicians. Neither of which are cheap. And it is not a particularly fast process. Then the board has to be retested. The end result is that they may be better off doing one for one swaps. And then tossing the bad boards.


RE: Not refurbished, still new
By dwalton on 11/6/2006 3:43:35 PM , Rating: 2
What the chance that cost for repairing a bad transistor being greater than the cost of manufacturing the board?


RE: Not refurbished, still new
By peternelson on 11/6/2006 4:00:52 PM , Rating: 2
As long as it's the SAME transistor on each board, they can set that up as an automated manufacturing process for re-assembly.

Typically the robot machine will locate the component position using place markers on the pcb.

I'd say because this is routine not a one off, if they collect all defective boards together, the cost to remove wrong resistors, place replacements and solder them on would be IMO around $5, much less than the board cost, making it well worthwhile than dumping boards.


RE: Not refurbished, still new
By czarchazm on 11/6/2006 3:32:52 PM , Rating: 1
Misty Dingos is correct. The amount of money to pay a trained technician for their work and the yields afterwards (read: loss of further money) do not justify remanufacturing a possibly large batch of defective boards. Most companies use these sort of large resistor values as circuit identification tags and not as actual functional parts of the circuit. It could be possible that EVGA tried to tag their own parts for their testing machines and not realized that replacing these resistors could have caused instability.


RE: Not refurbished, still new
By peternelson on 11/6/2006 4:02:52 PM , Rating: 2
The only labour involved is to program the assembly machine with what to do, a one-time process.

Besides, the labour costs are minimal since they are working on chinese wages.


RE: Not refurbished, still new
By flatblastard on 11/7/2006 1:50:17 AM , Rating: 2
So does a robot carry all the cards over to the machine? There is more labor/cost involved than one might think.


RE: Not refurbished, still new
By JNo on 11/7/2006 4:35:17 AM , Rating: 3
Why do anyone of you give a sh1t considering none of you have to make this decision for nvidia or its board partners and will never know..?


RE: Not refurbished, still new
By mindless1 on 11/7/2006 1:21:32 PM , Rating: 2
Wrong. You mistakenly think a so-called "Trained technician" receives high pay. If it were the case, the vast majority of parts on the market couldn't exist at the price points they do. It wouldn't be surprising at all for the work to be done by someone earning less than the US minimum wage.


Refurbs?
By augiem on 11/6/2006 12:58:42 PM , Rating: 3
Hmmm, I wonder if this means there'll be a bunch of EVGA refurbs hitting newegg's shelves soon...




RE: Refurbs?
By Brandon Hill (blog) on 11/6/2006 1:18:31 PM , Rating: 2
Woot.com anyone? :-)


RE: Refurbs?
By pepsimax2k on 11/6/2006 2:13:07 PM , Rating: 2
just what i was thinking...


RE: Refurbs?
By Chillin1248 (blog) on 11/6/06, Rating: 0
RE: Refurbs?
By lennylim on 11/6/2006 2:19:18 PM , Rating: 5
quote:
Would you really want a card that shuts down after only 30 minutes of operation?

Market that as a feature : time limiting function to prevent your kid/spouse/self from spending too much time playing games.


RE: Refurbs?
By kextyn on 11/6/2006 2:47:51 PM , Rating: 2
Refurbs would have the proper resistor on it.


By RaynorWolfcastle on 11/6/2006 1:30:32 PM , Rating: 2
Why on Earth would you change a MOSFET to fix it? Wouldn't it make more sense just to replace the resistor with one of the correct value?




By Tamale on 11/6/2006 1:36:45 PM , Rating: 2
lol.. this is what i was wondering..


RE: If the problem is an incorrect resistor value
By Desslok on 11/6/2006 1:40:59 PM , Rating: 2
I believe they mean checking the MOSFET.


By Desslok on 11/6/2006 1:44:37 PM , Rating: 2
oops.

Nevermind.


RE: If the problem is an incorrect resistor value
By lennylim on 11/6/2006 2:21:28 PM , Rating: 2
Call me a cynic, but it sounds a lot better blaming a resistor than saying "our chips are flawed".


RE: If the problem is an incorrect resistor value
By dwalton on 11/6/2006 3:37:17 PM , Rating: 1
What's the chance that a flaw that's card manufacturer dependent also being a chip flaw?

Would not that require the chip flaw to be restricted to a certain batch or lot and that batch or lot being delivered to just a subset of the companies that manufacture Nvidia cards?

If so, wouldn't EVGA or every other company affected blame nvidia for the mishap like Dell blamed Sony (who wants unwarranted negative publicity and to give competitors unfair advantage)?


By Goty on 11/6/2006 5:29:08 PM , Rating: 3
I don't think it's a chip flaw, but these boards were certainly build by NVIDIA. If you read the Bit-Tech article it says that the boards were shipped to their AIB partners like this. If I were someone that was going to buy one of these cards, I think I'd have to hold off for a few weeks while any chance of getting one of these boards was erased.


sounds fun!
By Samus on 11/6/2006 4:30:35 PM , Rating: 2
I'll pick one up cheap and solder a new resistor on there for a dollar.




RE: sounds fun!
By allnighter on 11/6/2006 5:57:06 PM , Rating: 2
I'm glad someone actually picked up on a "resistor" instead "transistor" part of this problem.
Since it's a "resistor" issue - it's a non-issue, in terms of hard can it be to fix. I would not expect an delays or faulty products on a significant scale. Stuff like this is pretty common and it's been around ever since. However this time it made front page news. Nothing to get overly excited about really.


RE: sounds fun!
By Goty on 11/6/2006 6:20:35 PM , Rating: 1
Sure, try to find one tiny resistor on that entire card!


RE: sounds fun!
By allnighter on 11/6/2006 6:32:11 PM , Rating: 2
...ummm... ok, I'll restrain my self and won't comment on that.
./shake_head_in_disbelief


RE: sounds fun!
By Goty on 11/6/06, Rating: 0
RE: sounds fun!
By allnighter on 11/7/2006 3:01:43 AM , Rating: 2
Ok...obviously I have to..
Primo - fortunately Joe Consumer is not the one that will be scouring the PCB in a jolly task of locating and replacing the resistor, but more likely a manufacturer with proper schematics and equipment to take care of business quickly and efficiently.
Secondo - yes, if the push came to shove, I'm sure I'd be able to dig up information about what needs to be replaced, how and with what part. Than it'd be just a matter of having a steady hand that day. I've had an opportunity to vol mod some of my cards and never have fried a single one.
How 'bout you bro?


RE: sounds fun!
By Goty on 11/7/2006 2:26:43 PM , Rating: 2
You're still missing the point. Assuming that you could find out exactly which resistor to replace, then no, it wouldn't be a problem. The problem is that getting that information would be a pain in the ass and without that information, it'd be almost impossible to guess at which 50K Ohm resistor to replace and be correct.


RE: sounds fun!
By allnighter on 11/7/2006 2:41:43 PM , Rating: 2
Are you sure it is me missing the point here?


Unphased?
By mechBgon on 11/6/2006 1:04:29 PM , Rating: 4
I think you meant "unfazed." :)




Honesty & Quality
By crystal clear on 11/6/2006 11:20:18 PM , Rating: 2
Thats what I call Honesty & Quality-well done Nvidia.
Better to come out & say yes there is a problem & we are rectifying it.
People appreciate this approach & attitude.
&
Well done DT for a prompt response on this.




When is the official launch?
By Staples on 11/7/2006 9:16:28 AM , Rating: 2
I thought it was today but I don't see any reviews for it so I guess it is not today. Does anyone know?




Sell it as an 8600GT
By wingless on 11/6/06, Rating: -1
RE: Sell it as an 8600GT
By Warren21 on 11/6/2006 11:16:36 PM , Rating: 2
I can only say one thing to that comment: Learn2read.

It is a single bad resistor(technically not even a bad resistor, just the wrong kind) which is part of the PCB and has nothing to do with the GPU cores/G80 chips themselves. Worse comes to worst, they could dump the PCBs and save the chips/memory but even that is overkill. Just change the one resistor and bang.

As for the RMA comment these were a small bunch cards that were mostly engineering samples for reviewers. These cards will never make it into the hands of consumers, even this close to launch.


RE: Sell it as an 8600GT
By PrinceGaz on 11/7/2006 2:51:05 PM , Rating: 2
quote:
Just change the one resistor and bang.


"bang" was possibly not the best choice of words :p


"Intel is investing heavily (think gazillions of dollars and bazillions of engineering man hours) in resources to create an Intel host controllers spec in order to speed time to market of the USB 3.0 technology." -- Intel blogger Nick Knupffer











botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki