I would like to lay out some of the facts about Intel-TSMS FinFETs. TSMC recently published two papers on FinFET CMOS in IEDM, December, 2010, one for high performance 22/20nm and the other for low power 32/28 SoC technology. TSMC reported in both cases a comprehensive data on process, device, SRAM and reliability including their statistical data meaning ready for volume production. TSMC, however, has not announced when it will be manufactured. Intel has not published its FinFET data except mostly news release for public consumption, and not for technical community. This is in contradiction to Intel’s past practices. Intel published new technology nodes such as 65nm, 45nm, and 32nm either in IEDM or in VLSI symposium before production or at the same time except its 22nm FinFET technology. Why? So, we can not compare technically or to see how TSMC chip matches that of Intel because no comparative technical data is provided by Intel. Intel has announced volume production around the end of 2011. It is very unfortunate to claim that TAITRA is not aware of TSMC’s FinFET technology progress, and even can not distinguish FinFETs from TSVs. My question to Dylan McGrath and those posted here is whether they have read TSMC papers on FinFETs published in IEDM as I referred above. They should, and see what technical information provided by TSMC compared with Intel. TSMC is a foundry company. As such it dose not compete directly with Intel. TSMC is ahead of its rivals such as IBM alliances and GlobalFoundries at 28nm that is in volume production this year. ARM has joined TSMC recently leaving IBM alliances for development of ARM processor at 28nm and FinFETs in near future. Therefore, TSMC dose not have to manufacture its FinFETs this year, but will be ready when in need. TSMC will be a formidable competitor for development of FinFETs beyond 22nm such as 14/12nm and 8/6nm in 2016 time period.
I don't believe it was sensationalism on purpose. I think it was an honest mistake. The author of the report, despite talking to experts, didn't understand that these were two completely different technologies. He emailed me earlier this week to say as much and also apologize. The report treated the technologies as identical, when in reality they are quite different. But, as I said, it was an honest mistake based on a lack of understanding.
When you say "flawed report that falsely equivocated" and later say "this was an honest mistake" these are countradictry... I think you mean 'conflated'..?? and not "equivocated"=past participle, past tense of e·quiv·o·cate
Verb: Use ambiguous language so as to conceal the truth or to avoid committing oneself. Was it sensationalism on purpose - just ignorant? or both?
Thanks Dylan for coming back and clarifying the story. Not many people would offer a Mean Culpa when they make a mistake. So Kudos to you and EETimes and shame on all the other journalists who still continue to carry that story.
Blog Doing Math in FPGAs Tom Burke 23 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...