[Rumor] More confirmation; NVIDIA G92/G92b bad

ThraxThrax 🐌Austin, TX Icrontian
edited August 2008 in Science & Tech
On August 12 The Inquirer made the damning accusation that the famed G92 and G94 cores powering the 8800GS/GT/GTS and the 9600GT were faulty across the line. This was not the first time The Inquirer had made this claim, but it appeared that NVIDIA AIBs were slipping notes on napkins to corroborate.

The buzz was that four board partners had come forward to identify a burgeoning trend in the death of these very popular NVIDIA GPUs. The trend had apparently been identified relatively early, but the failure rate has been steadily and alarmingly climbing. While big green remained mum on the situation, TheInq went so far as to say that it was a deliberate cover-up to save face in light of looming disaster.

The crisis, if true, would join the brutal beating NVIDIA is taking with the G84 and G86 mobile chips. NVIDIA's initial and stubborn admission of fault with the mobile cores suggested that a shipment to HP was the lone ranger in an otherwise-healthy line. Dell, Lenovo and other first tier OEMs have since come out with word that HP was certainly not the only manufacturer to receive a bad shipment. NVIDIA's lack of disclosure for the mobile market tends suspicion regarding their honesty in the desktop's.

TheInq is back to say that, not only are the G84, G86, G92 and G94 are bad, but so are their 55nm counterparts. The Inquirer has cited an NVIDIA Product Change Notification which identifies that the package soldering has been switched from High Pb to Eutectic Solder.

The change is unusual because there is typically little reason to alter the surface mounting material. Meanwhile, the solder technique has been identified as the point of failure in the mobile arena. The PCN also notes that the "Planned Implementation Date" is for the end of July, with shipments of the newly-altered G92 components on August 17.

The PCN notes that the change in solder material is to protect "supply and robustness." The choice in words is curious as a change to the stem the tide of mass failure does indeed fall under the domain of an improvement in product robustness.

Comments

  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited August 2008
    This could be very, very serious for Nvidia.
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited August 2008
    I wonder if there's a way to identify parts made under the new soldering process? I'd like to give my desktop a much needed upgrade to the now affordable 8800GT, but not if it's a ticking time bomb.
  • _k_k P-Town, Texas Icrontian
    edited August 2008
    My question is how knows someone with a G92 core failing on them, because I haven't heard of anyone. I know there is data backing this stuff up but what is everyone seeing currently in the field.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited August 2008
    There is lots of anecdotal information concerning failed laptop Nvidia graphics chips (forums), but I'm not reading about failing 8800 and 9600 video cards. Anecdotal information can be very misleading, but it's all I have to go by.
  • mas0nmas0n howdy Icrontian
    edited August 2008
    My understanding is that the desktop parts should take longer to fail because the issue is accelerated by heat cycling. The typical laptop is turned off and allowed to cool much more often than the typical desktop.

    Solution: GPU2 24/7
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited August 2008
    I have over 170 8800GT's and only one has had an issue... which is pretty good for how much we use them. PC systems under full load of gaming all day :)
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited August 2008
    I have over 170 8800GT's and only one has had an issue... which is pretty good for how much we use them. PC systems under full load of gaming all day
    Well, that's very encouraging. It's anecdotal, but nevertheless that's a fairly large sampling of cards.

    The supposition is that thermal cycling, not heat in and of itself, rather the repeated transition from cool to hot, will cause the solder/solder substrate to deteriorate. Those 170 video cards of yours, well, are the machines turned off at night (cooling)? How long have the cards been in use?
Sign In or Register to comment.