Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by robocat on Wednesday February 10, 2021 @07:15PM
(#26096633)
> [Intel] cannot move from "hand polished" proprietary processes they've been developing for decades to anything else.
Later this year TSMC will be producing Intel CPUs on 5nm. Clearly contradicts your assertion.
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by monocasa on Wednesday February 10, 2021 @09:59PM
(#26097763)
That's the rumor, but it's still unsubstantiated.
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by robocat on Thursday February 11, 2021 @09:21PM
(#26110110)
True, not confirmed by Intel, but backed by fairly substantial sources: http://www.trendforce.com/presscenter/news/20210113-10651.ht...
And: Intel CEO Bob Swan said the company had identified a "defect mode" in its 7nm process that caused yield degradation issues. As a result, Intel has invested in "contingency plans," which Swan later defined as including using third-party foundries.
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by madengr on Wednesday February 10, 2021 @06:18PM
(#26096184)
What process is Xilinx using; not CMOS?
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by baybal2 on Wednesday February 10, 2021 @06:41PM
(#26096367)
CMOS, but likelly with much more differences from CMOS for mainstream logic.
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by stjohnswarts on Wednesday February 10, 2021 @08:56PM
(#26097342)
I doubt that. FPGA's are just digital circuits. They don't use any onboard "ROM" the vast majority load their programming from an external chip. Where are you getting they don't use standard CMOS tech?
Re:Samsung Foundry: New $17B Fab in the USA by Late 2023
by kevin_thibedeau on Wednesday February 10, 2021 @11:22PM
(#26098228)
Typical SRAM FPGAs use a conventional process but there are architectures that use floating gates and OTP antifuses that need specialized processing.
What US can't source for F-35 locally are Xilinx FPGAs at either 65nm, or 40nm nodes.
The question is not in node size as such as having an FPGA process capability.
FPGAs on generic CMOS don't work great. The ability to tweak, and optimize the process specifically for individual design is huge for FPGAs.
Intel with their CPUs are very much in the same situation, but in reverse. They cannot move from "hand polished" proprietary processes they've been developing for decades to anything else.