3 ms·
Samsung unveils 7nm technology with EUV
- dbcooper 8y agoDavid Kanter's comments: https://www.realworldtech.com/forum/?threadid=177667&curpostid=177667 https://www.realworldtech.com/forum/?threadid=177667&curpost... https://twitter.com/TheKanter/status/1009617987341434880?s=19 https://twitter.com/TheKanter/status/1009617987341434880?s=1... Initial thoughts on Samsung’s EUV paper at #VLSI2018: SRAM yields are very low, 50% for 64x256Mb die on 1 wafer. Using EUV on metal layers without a pellicle is an odd choice. Significantly lower variability for EUV vs. LE^n as expected, unclear vs. SAQP. >It's not an odd choice it's essential. EUV requires more sensitive resists or higher source powers period. You can't use a pellicle because then you drop your wafers per hour another 10-15% (cus pellicles like everything, absorb EUV) and suddenly EUV isn't so profitable. Well it’s an odd choice if you want high yield! BEOL is fairly dense and no pellicle is like juggling knives. >Oh I'm well aware of the danger. But it's ok EUV is HVM ready right? This is everyone desperate to recoup what is now probably trillion+ dollars of EUV RND. The throughput is just going to get worse at smaller feature sizes as shot noise dominates requiring higher doses. EUV is definitely not HVM ready. Also, if it we have to wait too long, we will need DP EUV, which will be tremendously expensive.