ED#76 : More Trouble For AMD In 2008?

[Dashken]

AMD posted a massive loss of over 1.7 billion US dollars recently, the most massive of five continuous quarters of losses. Of course, AMD is putting most of the blame on their expensive acquisition of ATI.

But just how does the year 2008 bode for AMD? Their Quad-Core Opteron (Barcelona) and Phenom processors are already on sale, and so are the new ATI Radeon HD 3870 and 3850 graphics cards. Will they save the day for AMD-ATI in 2008?

Here's a quote from the editorial :-

"With the long delays in getting newer and faster models of their new Quad-Core Opteron (Barcelona) workstation/server and Phenom desktop processors out into the market, AMD is faced with a daunting year ahead. "

Link : ED#76 : More Trouble For AMD In 2008?

[cpuscientist]

About Dirk Meyer's statement:
"last month, while we were in the final stages of system validation, we uncovered a design error which is sensitized under very obscure operating conditions".

I must disagree with him the same way I disagree with bad politicians.
It is not true that it's only one design bug that was needed to fix. Base on my R&D experience and my analysis about the long delay, I suspect that Barcelona faced a lot of issues from a range of analog, digital, logic and mixed-signal standpoints. For example: write/read timing, power circuit design bug in the sub-systems which must affect memory and cache coherency and power bus causing logic, reliability and aging issues. This analysis seems to align with Dirk's shy away from the "Prevailing wisdom". It seemed that his confidence is gone because the Barcelona must be run at slower speed to avoid the obvious problems. If the device cannot run at the projected speed, then it seems to face both circuit drive strength and logic issue due to write read timing. What about reliability and aging factor, this concern must be the major stumbling block against product launch. I also suspect that AMD process has great difficulty in production yield. Low yield should be a factor impeding product launch.
If all these are suspected to be factors that impact 65nm, the same truth will hold for AMD 45nm design and process. Desktop and Laptop CPU's must be questionable, too.
My recommendation for AMD is to go back to the drawing board and find better design solution such as better simulation to improve CPU physical development, concurrently.

Automation-simulation modeling and RTL must be used to speed up process, check and optimize all circuits including I/O. Hire more engineers (not high level management)and buy more equipments to fix the situation. That is the way to stick to "compelling performance and value" you said it, Dirk.

Continuing restructure to bare bone will not help the situation, it only gets rid of good people for two reasons: 1/the good workers will leave for a better running company and some got cut due to bad politics. 2/What left would be just the political groups and those who hang on to the job because they are uncapable to find jobs out there. Logically, this will not lead to the road that should bring AMD back to its feet.