Max Chaikin and Ian King, Bloomberg, How Intel Makes a Chip, here. Nice basic presentation. Odd high tech presentation when you click through accompanied by something that sounds like 70’s porn soundtrack. Interesting about the Super 7. I want some super 7 juice to sell to the Banks – that is killer good news. Is Feldman still at Google NYC? Note there are no Banks in the Super 7!! They are going to finish full scale global banking automation with no control over the super 7 silicon? Guys at Morgan Stanley used to have enough juice to get silicon in the IBM Power designs but they used it for floating point execution in APL, that is why we cannot have nice things anymore. Blythe Masters must have a path into the super 7, find it.
Another way to make a chip faster is to add special circuits that only do one thing, but do it extremely quickly. Roughly 25 percent of the E5’s circuits are specialized for, among other tasks, compressing video and encrypting data. There are other special circuits on the E5, but Intel can’t talk about those because they’re created for its largest customers, the so-called Super 7: Google, Amazon, Facebook, Microsoft, Baidu, Alibaba, and Tencent. Those companies buy—and often assemble for themselves—Xeon-powered servers by the hundreds of thousands. If you buy an off-the-shelf Xeon server from Dell or HP, the Xeon inside will contain technology that’s off-limits to you. “We’ll integrate [a cloud customer’s] unique feature into the product, as long as it doesn’t make the die so much bigger that it becomes a cost burden for everyone else,” says Bryant. “When we ship it to Customer A, he’ll see it. Customer B has no idea that feature is there.”
Bob Colwell, Our Computer Systems are not Good Enough, U Washington 2015, here. Numerical Analysis comes back around 5 to 7 nm feature size with renewed interest in approximation algorithms and convergence proofs. Bullish for Pink I. Bad for Moore’s Law.
Thoughts on Comp Arch Futures, Dec. 2014, here.