One of the Laws of ICs has been that processing capability improves by 1,000 times every 15 years.
When the first microprocessor-based calculators came out in the 1970s, they had a processing capability of about 1,000 instructions per second.
When the first personal computers came out they had a processing capability of a million instructions per second.
This was upped to around a billion instructions per second in the year 2000.
So it took 15 years to go from 1KIPS to 1MIPS.
And it took 15 years to go from 1MIPS to 1GIPS.
So, if progress is maintained, we should be at 1TIPS in 2015.
Is it going to happen?
Well four years ago Intel said that an 80 core processor could operate at 1TIPS.
But where are we now? Four cores?
Are we going to get to 1TIPS by 2015?
It doesn't look like it.
Hmm, what is the performance of the highest raw processing power chips, ie graphic accelerators?
Multicore processors? mmmmmmmmmm let me think. Ping! Transputers!
Absolutely, Dr Bob, right on
I'd say basically they're DSPs, Djonne, not what I mean which is CPUs.
Isn't Transputer a very nice idea, but a fail in term of applications?
I don't know the story as much as you do, but on wikipedia ( :D ) I read this:
"This was excellent performance for the early 1980s, but by the time the FPU-equipped T800 was shipping, other RISC designs had surpassed it".
My vision:
Transputer pioneered the architectural concepts of what is now the leading edge of raw processing power, but was never at this edge itself in consumer products, no?
Sure GPUs are DSPs, but when I think of processing power, I think of the global processing power of the system, and in this field such DSPs play an important and increasing role, no?
I thnk you're right, Djonne, but the Transputer showed a way forward to increase processing capability without entirely relying on Moore's Law.
Yes agreed, Djonne, DSPs play an important role for specific functions. And Yes, graphics performance is a key function of today's PCs. But the performance of the GPU does not define the benchmark performance of the system as a whole.
Interesting topic; since the mainstream software industry still doesn’t exploit a dual core to anywhere near its full potential, it seems somewhat futile to pursue the idea of a 80 core device?
The Transputer was revolutionary and deserved to be a great success as a product. The reason that it didn’t make it is very simple; the same reason why Philips’ TriMedia has also consistently failed to make any significant inroad in the market.
They were great ideas, and well ahead of their time. Unfortunately, the business management of both innovations was left in the hands of a bunch of arrogant condescending twats with egos so big they needed sizeable wheelbarrows to transport them about. I interviewed at Inmos in the 80s and was appalled at their attitude, which some years later I saw again in the TriMedia club. “Customers don’t matter; our technology will prevail and dominate the market, because it’s the best there is….”
‘nuff said?
My God, Big Softie, that's telling them. And you're right, that's exactly how Inmos thought, and that's why Inmos' marketing efforts to mainstream microprocessor customers were so pitiful. In a plea of mitigation - that's how many chip companies thought in 1979 - chuck the technology over the walls of the Ivory Tower and the peasants will scrabble to get hold of it.