14 DAY TRIAL // Even though AMD has been selling DirectX 11-capable 40nm-based graphics cards for months, the general problems with those cards were low inventories. Even though Advanced Micro Devices did its best to cope with the demand for new GPUs, the company's efforts were largely impacted by the low yields of 40nm chips from the Taiwan Semiconductor Manufacturing Company. More recently, however, Digitmes reports that TSMC has finally resolved the issue.
Both NVIDIA and AMD rely on chips manufactured by TSMC and the part of the reason why AMD didn't grab as large a part of the GPU market as it could is due to the low supplies. While the CPU and GPU maker did what it could to sell as many cards as possible, NVIDIA delayed the launch of its Fermi without having to worry about AMD getting too big a head start.
Recently, information on the Fermi and upcoming AMD plans has been making its way to the web. NVIDIA will soon launch its next-generation GF100, and maybe even dual-GPU Fermi cards, and AMD has already introduced a DirectX 11 video card for the mainstream segment. Completing the triangle of fortune is TSMC itself, which is reported to have at long last gotten past the yield difficulties.
Digitimes reports that, according to Mark Liu, senior VP of Operations at TSMC, the semiconductor manufacturer has finally managed to fix the chamber-matching issues that were so severely impacting upon the 40nm yields. With this matter out of the way, rates are now at about the same level as its 65nm process. In addition, the company has completed the construction of Phase 5 of Fab 12. Located at the Hsinchu Science Park (HSP), Taiwan. Phase 5 will start volume production of 28nm products. Future plans of the company include constructing Phase 6 of Fab 12, which will handle 22nm production.
If the problem has truly been resolved, this means that NVIDIA will not suffer from the same inventory issues as AMD when it finally launches its next-generation CUDA.