[ANNOUNCEMENT] Wiki for discussing P35/IHC9(R)/SATA issues set up

韓家標 Bill Hacker askbill at conducive.net
Tue Nov 6 06:21:45 PST 2007


Aryeh M. Friedman wrote:
>>

>>From being around the industry since the late early 80's (I know you
> have been around longer) I have come to the following conclusions:
> 
> 1. Every 10 years or so there is a whole sell revolution in hw interfaces
> 2. The last one was Pentium, AGP, UDMA and PCI in the mid early 90's
> 3. The current one is PCI-E, 64bit, Multicore and SATA
> 
> Due to 3 I think if we get stuff right we are safe for a good 8 to 10 years.
> 

Not so. And 'safe' is a misnomer.

I've tracked VME, Multibus, PISA, and a slew of other industrial stuff as well 
as consumer.

PCI-e has pushed PCI-X off the shelves (mostly) in record time, even for 
server-grade boards. Simple economics. And size.

The rate of change accelerates as size/power/heat needs drop, and speed goes up.

Optical connectivity - in use for donkey's years inside Class 5 telco switches 
and 'big iron' - will be here in consumer MB 'Real Soon Now' from CPU outward 
[1] - simply because wire-bonding is taking up too much of the chip maker's 
budget, and sockets and connectors too much of the board maker's margins.

The bad news is we are not going to get 8 - 10 years. We'll be lucky to get 8 - 
10 *months* between silicon changes, and 2 to 3 years for bus & HAL changes.

The good news - looking at smart ID cards, mobile phones, NAND memory, etc. is 
that all this stuff will continue to get cheaper, smaller, less power-hungry.

Bill

[1] Already appearing for storage media and - more recently - audio connectivity.

And Intel have demoed - and patented - a process to grow the LED's right on the 
same die as the CPU. Goodby lead-bonding, BGA, and flip-chips.



More information about the freebsd-current mailing list