News

Letter: can Intel even afford to upgrade chip factories?

How can Intel extend Moore's Law? They can't upgrade their own wafer fab plants! - reader Sam Sethi responds to wireless news from the chip giant ...

Hi Guy

Well in a recent Intel report Pat Gelsinger mentioned that although Moore's Law can be extended for the Pentium chip the economic cost of doing so was preclusive. The last fab plant cost Intel $10bn. So what we are seeing is the law of diminishing returns being applied to the high tech industry.

This economic barrier is fuelling the distributed collaboration efforts of Intel (netbatch), Groove, Staffware, MatrixOne, eRooms and IBM with their GRID initiatives.

So clearly Intel is looking to do a Microsoft and bundle in more functionality into the chip. Adding in the radio components into the chip is logical, as the radio is something everyone can listen too whilst doing something else (working on their computer or on the move) unlike TV.

Equally important is the agnostic stance Intel is taking over Bluetooth, WiFi etc as it should not matter which connectivity is used.

One question I have is will they integrate a DAB (Digital Audio Broadcasting) receiver into the chip as your article fails to mention this. This "radio" initiative will be good if it helps accelerate the uptake in DAB receivers. The www.dab.org site shows that coverage in the UK is at 70% and that the radio stations BBC and Digital One are beginning to increase the number of stations.

In the future chip based radio's sit inside of mobile devices such as the ZooPad will enable . a DAB receiver to receive text and have some download functionality. This will make radio an two-way interactive medium and will change the type of programming we get in the future.

As for ultra-wide band (UWB) there are already political issues slowing down its potential. Recently AT&T, Sprint, Verizon etc lobbied the US FCC (Federal Communications Commission) because it would eat into their Wifi, GPS, BlueTooth and Cable businesses. They thankfully lost their first battle but they will be back. UWB is the future but not for several years yet.

Finally as for the "sensor network" this is great and is working along the lines of IBM's automnic initiatives. These initiatives are based on two driving forces. Firstly RELIABILTY in the future (even now) there are not enough engineers to build manage or maintain the systems. Computers are going to have to take over some of this work and become more intelligent.

And secondly AVAILABILITY, if the potential for GRID and Web Services is to be meet then availability of the network 24/7 will be key. As Sun kept saying "the network is the computer".

Intel are having to move away from pure PC chip sales. Chips for multiple new devices with multiple functions is their future. Will it be called Gelsinger's law in the future. i.e the number of functions available on a chip?