Epic FAIL: NVIDIA?

A recent announcement from HP has created a bit of hubbub in regards to the manufacturing processes over at NVIDIA. HP has stated that some 38 models of their Slimline desktops have GPUs that could fail. Add this on to the large number of GPUs in laptops that are failing, and the green guys have a bit of a PR disaster on their hands. Let me shed a little light on this subject before you start running out to replace every 8000-series GPU in your house and office.

Charlie over at the Inq began blasting the FUD big time when NVIDIA announced that it would be spending some $200 Million on replacing faulty GPUs. He heard from industry insiders that there were many more problems than just the notebook GPUs, and went as far as to say that all G84 and G86 chips are bad. He then revised this a few days later to include every chip that NVIDIA made in the past 2 years. The bonding ASIC that couples the GPU to the packaging was basically not the correct one for the kind of thermal stress that NVIDIA claimed it could handle, and thus the chips are essentially falling apart in certain situations.

If you’ll read down, he says that all chips are bad and that even desktop chips are failing. HP’s Slimline announcement should, in theory, justify those claims. However, if anyone has ever seen one of these Slimlines in person, they can hardly be considered Desktop PCs.

See, the problem in all reality is that NVIDIA told everyone (even me) that their chips were good to over 100°C, and everyone ran out and designed to keep them in that envelope. Anyone in their right mind should know that there is no chip that can really handle that kind of heat on a daily basis. I mean sure, you can get a Core 2 with a 135°C TJmax up to 100°C every now and then, but you should always design to keep chips under 80°C.

Laptops and Slimline Desktops throw this out of the window, and are designed right on the breaking point. I actually saw one of these Slimlines in person about a year ago, and was shocked when the MCP heatsink singed my fingers a bit when I checked to see if it was hot. It was hot alright, probably pushing 80°C just at idle. No wonder the chips are dying!

So yes, NVIDIA is to blame for telling everyone their chips could handle 100°C, but c’mon guys. Any idiot that believed them is getting what they deserve right now. I’m glad to say that Nordic PC has had zero, I’ll repeat that, zero failures in NVIDIA GPUs in the 8000 and 9000 series. It’s called designing for longevity, and it’s something you won’t get from the big guys.