Are TSMC’s Yield Issues Going to Cause GPU Costs to Rise?

TSMC might be the world’s largest independent semi-conductor company, but the way things are going lately, it’s hard to know if that’s going to be the case in a few years. In recent months, much has come to light regarding low yields on its 40nm processes, and according to industry-tracking site DigiTimes, the problem is unlikely to be solved anytime before May. If that’s true, it’s not only bad for the company, but other companies as well. Not to mention, us consumers.

As I mentioned in our news section last week, NVIDIA has been having a hard time reaching acceptable yield levels with its GF100 chips, and the success rate is rumored to be as low as 2%. That’s truly horrible, because in the end, it means that for NVIDIA to make any gain in revenue off of the chips, the resulting card will have to be priced high. But then there’s the added problem of performance. If the card doesn’t blow away ATI’s current HD 5000 series, then no gamer is going to want to pay the huge premium.

NVIDIA stands to be hit the worst with TSMC’s issues, because the company hasn’t even managed to get its first true next-gen 40nm chip out the door, so it’s going to be awfully hard to ramp up production and meet demands with such issues. We saw virtually the same thing with ATI this past fall. There were production problems, and that resulted in two things… higher priced cards and low supply.

Despite the issues ATI ran into, it seems to be very fortunate to have things play out the way they had. Because as it stands today, they have product, and NVIDIA doesn’t. Though NVIDIA will soon, it stands to be seen just how much they can keep up to demand. And plus, no one will pay $600 for a GPU that’s only 25% faster than a $400 one. Not many, anyway. And if TSMC doesn’t fix the issue soon, it could be our ATI cards that see a rise in cost as well.

Is anyone else out there really hoping that GlobalFoundries can get its game going? And soon?

Although TSMC recently said the defect density of its 40nm technology has already dropped from 0.3-0.4 per square inch to 0.1-0.3, the sources pointed out that the improvement in overall yield still needs more time before catching up with market demand.