0mz = FAST (but how do you overclock?)

[cjolley]

http://www.techreview.com/magazine/oct01/tristram.asp

Too cool.
chuck

[Bohrn]

That is very cool! I like the idea that it is a better, smarter way of doing it, not just brite force. Too bad it will take so long to implement.

[Jon P. Inghram]

Hmm... it might have a clock to govern the speed at which the individual elements operate, thus overclocking may still work.

[Wombat]

I found the article a little lightweight, and sometimes questionable in both facts and purpose. But asynchronous design <I>is</I> very cool. But it's a royal bitch to try to pull off in any kind of large-scale design. I'm surprised there wasn't more information from Sun, since last I heard they were the "closest" to having an asynch GP CPU. But please keep in mind that's like currently being the "closest" to having a power-efficient fusion reactor right now.

[cjolley]

Originally posted by Wombat
I found the article a little lightweight, and sometimes questionable in both facts and purpose. But asynchronous design <I>is</I> very cool. But it's a royal bitch to try to pull off in any kind of large-scale design. I'm surprised there wasn't more information from Sun, since last I heard they were the "closest" to having an asynch GP CPU. But please keep in mind that's like currently being the "closest" to having a power-efficient fusion reactor right now.

Oh, whadu you know Whipper Snapper.
Happy belated Birthday.
chuck

[TdB]

TdB

but if there is no clock, then how can they make one async chip faster than another?? smaller die-sizes???

seems like we NEED pr-rating after all

it is a shame that pentium chip wasn´t released, it would probably be faster than the pentium2, and maybe even the slower pentium3´s

btw have you heard anything about async fsb´s my computer-architechture-teacher said that it would would be easier to implement than an async cpu, and that it would be a really good idea. is this being researched aswell?? would this require a new pci-standard?

[Wombat]

I think async FSBs might actually be <I>less</I> likely. The current prevailing approach in asynch is to have a handshaking signal that couples along with the data, and determines when calculations are complete.
On a PCB, that would mean both more traces (bad), and having to protect these new signals from noise in more effective ways.

There's nothing that said the async Pentium was verified. 5 years ago, the tools were even less competent then they are now.
They probably still aren't good enough for anything over a few million transistors or so. That asynch Pentium would still be highly useful (and profitable) today in embedded or portable devices. There's got to be other reasons it wasn't released.

[Bohrn]

I'd say it was squashed as it would be hard to market, and at that time, Intel OWNED the PC processor market. They had no reason to rock the boat. Marketing a processor that is not easily quantitatively faster and better would be hard. People like to be easily able to one-up each other.

just my 2 clock cycles of thought on the matter.

[cjolley]

Well, it also might have cost $10,000 per chip to make.
Fine for research, useless as a product.
chuck