TSMC's 40-nm yield problems surfaced earlier this year, but the company claimed it largely resolved the problem. However, during a conference call on Thursday, graphics chip maker Nvidia Corp. discussed 40-nm capacity and yield constraints at its foundry partner--TSMC.

Advanced Micro Devices Inc.'s graphics chip group is also seeing similar problems at TSMC, but not all are suffering with lackluster yields. John Daane, president, chief executive, and chairman of the board of Altera Corp., said that the FPGA house has been shipping parts, based on a 40-nm process from TSMC. ''The yields are good,'' Daane told EE Times.

Nvidia, in contrast, has been vocal about the yield problems. ''Overnight, Nvidia discussed 40-nm capacity constraints at its foundry supplier TSMC,'' said Barclays Capital analyst C.J. Muse.

''Management discussed yields improving but that allocations still remain inadequate. Though yields were improving, they were important enough for TSMC to mention a chamber matching problem on its conference call earlier this earnings cycle,'' he said in a report, which is based on Nvidia's conference call to discuss its results.

Reports surfaced that TSMC is having issues with its ion implanter supplier, causing a shortfall of 40-nm parts. ''Demand (at Nvidia) far exceeded supply, particularly in the 40-nm product area; the company is in a 'sold out' situation and this is likely to continue for the next several months. Virtually all products are on allocation with very lean inventories in the channel,'' said Hans Mosesmann, an analyst with Raymond James & Associates, in a report.

''Both AMD and Nvidia are supply constrained, and Nvidia is likely getting most of the allocation,'' he said. AMD's graphics chip unit is also using TSMC as a foundry.

''With both AMD and Nvidia being supply constrained, the supposed 2-month AMD market advantage in new DX-11 GPU's is irrelevant in our view. AMD just missed an important window of opportunity this season in our opinion,'' he said.