The regular digital electronic chip era is coming to an end.

[Wayne morellini]

What is happening is the are now talking about magnet computing at 40 terahertz, and silicon transistors at 1 million times faster than silicon speeds. That is just the start. By using calculated result tables, you now get low energy high speed alternatives to FPGA. I've been investigating this sort of technique for some time in my own work. But, there is a faster technique in optical computing, which is too complex for me. And a new electronic.

There is a thinner slice of time left to develop existing chips before they are outdated. If they can get the price of the technology cheap enough in coming years, we might have very high end super computing out competed within years, and normal computing years after that. So, a two year time frame to develop a chip to have a 2+ year life cycle. Maybe it will take 20 years, but I expect its over.

So, what big opportunities are out there despite this, server applications. Presently, server farms are taking up a noticeable fraction of energy in their locations. This is very expensive before, and more so presently. Low energy technologies can save a lot of money. It doesn't matter if they have to rewrite server software to work on a low energy low cost hardware, if it saves the industry billions a year in electricity costs. It makes sense. Any server farm made within the short term, benefits from this technology.

In the past people didn't realise, that complex architecture could be done simply with low energy, making it suitable fur processor on memory chip technology. Dram memory process nodes, are too some to take normal processors, but I believe this won't be true fur misc, which is vastly simpler.

Thank you

[Jurgen Pitaske]

There are always new technologies around the corner.
I rather wait to see when there is a price and a delivery time and documentation.

Regarding server farms and replacing them with lower power options I have not seen the numbers.
The ones installed now are there and do the job well it seems.
How much does it cost to replace them and who pays for this.
And is the better silicon actually available?
As soon as the numbers are there it makes sense discussing this further.
It is not just the cost of electricity.

[Zbig]

> What is happening is the are now talking about magnet computing at 40 terahertz, and silicon transistors at 1 million times faster than silicon speeds. That is just the start.

...which will allow to software makers use of even more „modern” software development
technologies than OO, probably it'll be „point and click”. On the other hand: on the rule „the
gear is here, why not use it” we (the users) will have to equip that gear with terabytes of RAM
and petabyte-sized SSDs, to have enough space where all that bloat will be kept.

All this to perform (90% of the time) the tasks, for which 286/12 MHz/1 MB RAM would do,
was the software created from the ground up in assembler „as it used to be in the early days”.
Or -- in Forth... why not.

[Wayne morellini]

Well, that's not how it operates. Design discussion starts with what is possible, figures or not, wherever they have to generate the figure or not. Otherwise businesses fail. One cannot insist that everything is fine for them and remain competitive based on acknowledgement reactionary thinking.

It is talking about new server farms, so a replacement cost is gauranteed. I should have remembered to put in new server farms explicitly. It is standard practice to replace equipment and build new farms

It as about the accumulative cost of electricity. Replacing A with cheaper B that has less lifetime running costs.

Anyway, very tired, so I hope that covers everything?

[Wayne morellini]

:)

Yep. Except I suspect it will be difficult to make the code that much bigger per set featured routine. But, they will likely want to apply large AI to it, consuming memory. At the atomic level, the storage density is unreal.