War over bragging rights to being first to 45nm broke out this month with Intel and IBM claiming to be first out of the starting gate. TSMC, usually vocal in these contests, stayed mum.
However the course may have a long way to run before we know who can actually make a 45nm chip with a commercially useful specification, and with commercially acceptable yields.
It is a byword in the industry that the introduction of a new material causes huge production headaches and, at 45nm, the industry is having to cope with two new material introductions: with metal gates instead of polysilicon gates with Intel even using different gate metals for p and n-channel devices, and with dielectrics using hafnium compounds, instead of silicon dioxide, to try and keep leakage manageable.
What effect these introductions will have on the yields, and the performance of devices, is probably still not apparent at this stage, though ARM’s CEO Warren East says he expects to see tape-outs on 45nm this year from his lead partners in the development of ARM’s 45nm physical IP, Samsung, IBM and Chartered.
Everyone will be looking back to the nightmare of the 0.13 micron generation where the introduction of copper and, in some cases, low-k dielectrics, put back the introduction of some companies’ products by a year while production engineers wrestled with the problems the new materials caused.
And why is TSMC so uncharacteristically silent? Is it still smarting that its high-performance 65nm process being developed with lead partner Altera won't be up and running until 2008?
While the bragging wars are always quite fun, let’s remember that, at 45nm, it may well involve a much wider than usual slip between cup and lip.
Leave a comment