Home
| Calendar
| Mail Lists
| List Archives
| Desktop SIG
| Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU |
On Sun, Jul 11, 2010 at 12:24 PM, Rich Braun <richb-RBmg6HWzfGThzJAekONQAQ at public.gmane.org> wrote: > There are two other key factor: ?cost and competition in the computer > manufacturing business. ?With each new generation of microchips, the doubling > of performance comes with it a doubling of cost--of the factories required to > make them. ?Intel now has to pay roughly a billion dollars to build a fab > plant. ?(This compares to $15 to $50 million for factories at the time of > Kurzweil's predictions.) That's an excellent point -- the reasons I think Moore's law will slow down a bit in the near future are two fold: 1. It is getting much harder to shrink transistor sizes as we go to 22nm and beyond. It will take extra R&D to reliably manufacture chips at those feature sizes 2. The cost involved in upgrading all the fab equipment to the next generation is > $1bn as Rich points out. So companies will want to amortize that cost over a larger number of chips, which means they may keep around that equipment longer. > What about corporate buyers? ?I personally believe that industry innovations > since the first TRS-80 came out 30+ years ago have been led primarily by > consumers. ?That trend continues today: ?graphics adapters and subsystems are > front-and-center where the action is, not just today but for the past few > years. ?VMware and its cousins can only go so far to drive demand on the > corporate side. ?At my own workplace, demand is coming not from our internal > requirements but from--*consumers* who want to download more and more video. Spot on again! AMD's next gen "Fusion" chips (and Intel's future chips too) will have graphics on the same die: so we'll now have multiple CPU cores, DRAM controller, I/O subsystem (PCIe links) and graphics all on a single piece of silicon. By "graphics" I mean not only the 3D engine, but also dedicated units for media decode and encode too -- all meant to improve media performance at much lower power levels (you want to be able to watch 2 full HD movies without plugging in, don't you?). Shankar (Disclosure: I work for AMD, but views above are my own)
BLU is a member of BostonUserGroups | |
We also thank MIT for the use of their facilities. |