A couple of notes.
Firstly, I doubt the man was an above-par engineer (edit: I'm not saying he is terrible, just that he is not god's answer to microchip engineering). His current profession is intellectual property lawyer (you know, they type of person that makes money from the patent and copyright laws/lawsuits that are so bad for the general public). Ad hominem aside, I seriously wonder if someone can be both a good lawyer AND a good engineer (note: I'm an EE student looking at the possibility of going into medicine. I realise that I can't be a great EE and a great doctor. the professions are just too different with each being extremely complex and the study of each being the work of many lifetimes).
The next mention is that he quit in 2006-2007. AMD has had major changes since then (also drafts of bulldozer were probably undergoing major revisions at that point).
Next up is computer design. The man is exaggerating when he claims that previous designs were designed exclusively by hand (edit: think about how much time would it take to place even 250 million transistors by hand WITHOUT trying to be efficient. It would take orders of magnitude longer to try making an efficient design. A good analogy is programming. Writing a large program in assembly is inefficient. Instead the program is coded in a higher level language such as C++. Only the mission-critical parts (such as those little loops which execute billions of times) are designed by hand assembly. The rest of the circuit design (a program is a method to rearrange the transistors in an integrated circuit) that is mostly computer designed with humans giving more general directions for most of the program and giving very specific instructions when needed.) and would also be exaggerating when he claims that current designs are made exclusively by computer. Let's assume that his 20% deficiencies and 20% larger size claims are true. It is still faster to have the computer crunch an initial design and then tweak the design for efficiency than to have an engineer spend months making a design that must still be tweaked (probably to a similar extent) to achieve maximum efficiency.
The only chips that could have used this alleged computer only design would be still undesigned chips. He claims that AMD has just been iterating the same architecture since A64. If this were true, the only computer designs that could have possibly been made would be scrapped designs (aside from 45nm bulldozer, all other scrapped designs were K8 or K10), Zacate, and Bulldozer. Bulldozer has design problems. Nothing more to say except that these issues are probably fixable given a revision and new fab. If Zacate is an example of computer design, sign me up for more. The chip is bulk (ie cheap to make) and has 146 fewer pins than atom making motherboards cheaper, smaller (fewer circuits), and faster to make while design of the same mother boards is faster and less error (less to account for). While being cheaper and easier to design for, the power consumption is still better than the latest atom due to the atom's need for a dedicated GPU (the atom IGP sucks). All this aside, saying that the changes from K8 to K10 were "minor revisions" is like saying that the changes from core to Sandybridge were "minor revisions." True, the revisions were incremental, but the changes were major (true quad-core, DDR2, DDR3, 20-30% IPC increase, power management, etc).
I believe that the only accurate statement made by this former lawyer/engineer employee is that the old design team (like old employees everywhere) fought to keep the system "as it has always been" and that the team either disliked or feared changes to "the way we do things."
Edited by hajile - 10/18/11 at 6:42am