引用:
作者weiter5494@2014-05-03, 09:03 PM #260
...28HPm - 提速30% vs 28LP
20SoC - 提速20% vs 28HPm
16FF - 提速40% vs 20SoC
16FF+ - 提速15%、節能30% vs 16FF
16FF+ - 提速40% vs 20SoC
|
比照TSMC最近兩季法說會訊息,所謂『16FF - 提速40% vs 20SoC』,應是16FF與16FF+的混淆。實際上可能是:
16FF -
提速27% vs 20SoC
16FF+ - 提速15%、節能30% vs 16FF
16FF+ - 提速40% vs 20SoC
目前28HPm打造的Snapdragon 801將可達2.5GHz(MSM8974ACv3),依照TSMC之效能指標,未來(以保守MSM8974VVv2的2.2GHz為推算基礎):
20SoC - 提速20% vs 28HPm → 2.2×1.2=2.64GHz
16FF - 提速27% vs 20SoC → 2.64×1.27=3.35GHz
16FF+ - 提速15%、節能30% vs 16FF → 3.35×1.15=3.86GHz
亦即,以16FF+打造之SoC,近4GHz速度是可期的。
PS:
不似28nm製程有28LP、28HPm、28HPC三種針對不同產品特性優化的細分,20SoC、16FF、16FF+均只有單一製程,目前仍不清楚其提速改善是否一體適用mobile SoC、CPU、GPU等不同訴求的處理器。缺乏foundry標準優化製程的便利,在遵循foundry設計準則的同時,預期fabless須倍加強化自身架構的優化與深化和foundry工程人員的互動。
類似TSMC的16FF、16FF+製程,
Samsung/GF目前也有14LPE(Low Power Early)、14LPP(Low Power Plus)兩種。
16FF - 2014年1月完成製程認證,4月收到first tape-out。
16FF+ - 2014年9月預計完成製程認證。
14LPE - 2014年2月完成製程認證,2014年底投產。
14LPP - 2015年導入。
鷸蚌相爭,孰得利?Intel嗎?No!我只想賺點貼補家用的錢。
Whether Intel could lure Apple or Qualcomm if the current suppliers of those companies continue to run up against problems with FinFet?Questions about what such would look like,
fears of 30%-40% gross margins play off against this year’s
better than $3bn mobile loss. One would think an Apple/Qualcomm arrangement would cease Intel’s present mobile course, and
eliminate this year’s $0.60 EPS hit. Intel’s wish is
to fill its fabs at profit rates sufficient to fund process R&D while focusing on PC legacy and fending off ARM in Data Center.
- Richard Whittington with Drexel Hamilton