"It takes about six learning cycle, roughly, to complete one generation."
Oral history of Shang-Yi chiang
interviewed by: Douglas Fairbairn
recorded on March 15, 2022
computer history museum (CHM)
"When we develop one node, basically you have some learning cycles. First, you do some simulation. And you have some idea, then you run wafers to prove that. So, you run a group of wafers according to simulation and you have some splits. The wafer runs through the fab, they come out and you measure them, you analyze them, and you try to improve and you run this again. This again, you run. So, this is learning cycle."
"It takes about six learning cycle, roughly, to complete one generation."
"My R&D wafer in the fab run much faster than yours, because my R&D engineer works three shifts and you only work one shift. So, your R&D wafer move eight hours a day, my work/move 24-hours a day. So, my wafers go three times faster, even if you are twice smarter than me, I still beat you up."
https://archive.computerhistory.org/resources/access/text/2022/07/102792671-05-01-acc.pdf
____________________________________
https://archive.computerhistory.org/resources/access/text/2022/07/102792671-05-01-acc.pdf
Chiang: Intel, Motorola, National, HP. And they're all my customers.
Fairbairn: Right
Chiang: And the board member they sent to Sematech for many companies happen to be the person also in charge of the supply chain. Well, they are my major customer, but on the Sematech board we kind of sit together, we can discuss things on equal base. Not like my customer. <laughs> And so, in that platform we are able to-- some time we are able to more freely exchange some information. So, one time at a dinner, they asked me, they said that, "We all take two years to develop one generation, how come you guys can do it in one or one-and-a-half year?" And they asked if some of your customer transfer technology to you or what not? And I told him, "No," I told him that, "That's not true." I think he probably implied we steal technology from customer, the way he talk.
And I say, "I'll tell you why." I said that, "When we develop one node, basically you have some learning cycles. First, you do some simulation. And you have some idea, then you run wafers to prove that. So, you run a group of wafers according to simulation and you have some splits. The wafer runs through the fab, they come out and you measure them, you analyze them, and you try to improve and you run this again. This again, you run. So, this is learning cycle." At that time, "It takes about six learning cycle , roughly, to complete one generation." Of course, you had some short loops and not just one. I said that, "My R&D wafer in the fab run much faster than yours, because my R&D engineer works three shifts and you only work one shift. So, your R&D wafer move eight hours a day, my work/move 24-hours a day. So,
my wafers go three times faster, even if you are twice smarter than me, I still beat you up." <laughter>
Fairbairn: That's what everybody says. Faster learning cycle, right?
Chiang: Faster learning cycle. And three to one is kind of a little bit exaggerated, because it's usually night shift it's not very effective. Just the idea. But because I knew at HP, TSMC R&D wafer did move much, much faster than HP. But HP is not a good benchmark
Fairbairn: Not a good bench-- yeah.
Chiang: And then they ask me, "How can you make your R&D engineer work night shift?" And I kind of joke with them -- and I can share with you the real reason what I think. But at that time, I told them, I said, "In Taiwan, we all have to serve the military." I said, "I did. When you're in service, you-- especially in the basic training-- you take a duty for the security guard."
Fairbairn: Mm hm, stand watch.
Chiang: Stand watch, right. "It may be my turn from 2 a.m. to 3 a.m. Then the guy would wake me up at 1:45. Then I got up, I change my clothes, I got my helmet, got my rifle, then I went over at 2 o'clock, and 2:45 I wake up another guy. And so, all my engineers have been through that. So, I tell him to, you know,
it's your turn to do that! <laughs> Don’t complain!" <laughter>
And what interesting at that time, the board member from Motorola, I just remember, his name was Bill Walker-- I don't know if you know him or not?
Fairbairn: No, I don't think I know him.
Chiang: Bill Walker. He's a big, big guy. Later on, I found he used to be a Marine. I knew that later.
Fairbairn: Yeah.
Chiang: Because he was one of my very large customers, I usually visit him once a year.
Fairbairn: Mm hm.
Chiang: I went over with our sales manager and the two of us usually went to his office. And he had one of his Supply Chain Manager - four of us sit-down. Usually Bill would give us a lecture telling us what we did wrong and how bad we behaved. I took notes. The meeting lasts an hour. Next time I visit him, they took me to a different room. I found it a little bit different. When they opened the door, there were about 20 people around the table. He was in charge of R&D and the manufacturing for Motorola Semiconductor Worldwide. He said, "These are my R&D and the fab managers in the entire world. I got them together. I want you to tell the same story to them." <laughter> So, that was what I always told them but they didn’t listen.
Fairbairn: So, what was the real answer about <laughter>?
Chiang: The real answer is I, honestly, I just share with you, I think the culture. Asians are more hungry, because we had a tougher life. So, to make money is more important to us. People are willing to sacrifice their own privacy, their private life in order to have financial security.
Fairbairn: That's what you did. You moved to Taiwan without your wife or your family, right?
Chiang: Right. Just to make a living.
Fairbairn: And work 22 hours a day.
Chiang: <laughs> But not later. So, I firmly believe this is one of the really important reasons why TSMC succeeded. It's culture. If equipment went down, because equipment depreciation cost was so high, you really want to run your equipment 24 hours a day. In United States, if equipment went down, wait until next morning. The people come in at eight o'clock and probably go to fix it, nine o'clock. Yeah. But if at two o'clock in the morning, we just called the equipment engineer, "You come right away," he won't complain. And his wife won't complain. And that's the way it is.
Fairbairn: Right.
Chiang: And that help a lot.
https://archive.computerhistory.org/resources/access/text/2022/07/102792671-05-01-acc.pdf
____________________________________
No comments:
Post a Comment