OTT metal but interesting

I wondered in a recent post to what precision the semiconductor Industry now worked and looked up on Wikipedia "Moores' Law" I remembered the humorous version which assumed semiconductor sales guys sold like car sales guys and went something like

"You remember that last year we sold you a car with the comfort of a Rolls Royce,200mph top speed and 100 miles to the gallon-this year 's model has basically the same spec. but does Mach1 as top speed and 1000 miles to the gallon"

" Yes thats all very well but last year's model cost us £50,000 . How much is this one going to cost us?"

" About ten quid"

The Wkipedia article is a lot more serious but almost equally staggering and a very interesting read. They are now apparently approaching atomic scale transistors on chips but in the main opinions are that Moore's Law will continue to hold until at least 2020.,and probably well beyond.

The sad part of the story is that practically all the big players around when I worked in the industry have sold their semiconductor business as no one except Intel (who made a quantum technology jump) could consistently make a profit. There was always just too much competition,and still is altho it is n't the same guys, although Texas Instruments and AMD and a few other originals are still going.

Regards

Edited By alan frost on 17/11/2014 09:06:51

Edited By alan frost on 17/11/2014 09:18:16

Yes, progress has been astounding.

In the late 1960s I was doing computer software research using an ICL mainframe computer. It had a cpu running at a 1.4 Mhz clock speed and a massive 192 k of memory. It cost about £3 M, or about £60 M in today's money and took up one entire floor of the maths building with massive amounts of air conditioning to keep it cool.

Today I have a smartphone in my pocket that has a quad core cpu running 1000 times faster with nearly 200,000 times the memory. It can do things we only dreamed of at that time and the cost? 1/500,000 of the cost of the mainframe. So something like 500,000,000 more processing power per pound!

So performance/price has doubled every 1.5 years over that time.

Russell

Russell , as you worked in the maths dept. I'll take your word that 2 to the power of 25 odd is 500 million,. Its not a hard sum but you mathematicians probably just know that. I tend to know only up to 2 to the tenth is a thousand (more or less).

Regards

I was in the electronics dept but was using the maths dept computer at about 3 a.m. when all the maths people were asleep

500 million is nearer to 2 to the power of 29. I was talking about computing about 45 years ago.

Russell.

Shades of the original IBM PC and Bill Gates' prophecy that 640K or ram was enough for anyone.

Its so easy to completely lose track of numbers that are bandied about, that people have no real concept of how big they are. Take the terabyte disk, for example. If you took one standard fag paper for each byte in a terabyte, and stacked them one above the other. How high would the stack be?

Don't calculate it, just take a gut guess.

It would be about 17,500 miles high!

Surprised?

Precision: The chips in our current product are obsolete after 3 years so we have just requalified with 20nm line width ones. Smaller so more chips per wafer so more profitable for them and cost us tens of thousands to do the work. Stuff does get cheaper but at the expense of being unmaintainable after only a few years.

Then there was O'Hanlon's law - " A PC full of memory will always cost £60". It's held good for 25 years...well roughly.