On The Insider: Britney's Bikini-Clad Top 10
BNET Business Network:
BNET
TechRepublic
ZDNet
TalkBack 7 of 13:
Next »
« Previous
What is limiting duh scope of duhvelopment efforts
One of duh best ways to speed up CPUs is by removing the clock.

Synchronization is the single largest limiting factor in performance. All of the stages, pipelines, units and caches are spending large percentages of the total processing time being "busy" waiting for each other.

Asynchronous processing allows all parts of the CPU that _can_ do more work to proceed with it until there is some critical piece of data that they have to wait for. Removing the clock also reduces the transistor count (only with the simpler, less highly developed designs at the beginning of development of course, and only in comparison to synchronized circuitry, later they will get higher transistor counts too, I believe), thereby reducing the circuit "length" in terms of how many transistors are used to process the signals, as well as the die size and waste heat, which is the other current limiting factor keeping synchronous CPUs from expanding any faster after feature size.

However this is far from a no-brainer since it is completely discontinuous from current design process and theory at the silicon level and also requires entirely new compiler design as well in order to process the data in a way that keeps it flowing through the CPU without having massive caches to hold, and then locate again in the place it is stored, the data you seek to manipulate.

But we have very fast CPUs now that could be applied to solving the very different needs and methods that asynchronous processing would require. And we have now reached hte point that it would be affordable to turn some of these high-powered systems loose on a project like this. But who wants to be the first to start investing in a new paradigm while the current one has universal acceptance and a mature manufacturing and marketing infrastructure. Perhaps the Chinese, as they are not already in over their heads in infrastructure that can't be converted to the new use and they are looking for something that is both distinct and utilitarian.

The down-side for them would be who to sell it to, unless they can find a way interface it to systems that currently service synchronous hardware or produce motherboards and peripherals affordably to work with the clockless chips. Even then, what software will work with it?
Posted by: Still Lynn   Posted on: 11/23/03 You are currently: a Guest | Members login | Terms of Use

Alert moderator to an offensive message

Subscribe to this discussion via Email or RSS

Opinions  FirstNLastN | 11/20/03
I bet you hate Microsoft also  adecrignis@... | 11/21/03
boost P4 CLOCK!  michael-t | 11/20/03
Yes, but will 4Ghz be 4Ghz ?  BitTwiddler | 11/20/03
Speed increase  michael-t | 11/20/03
More of the Well Duh, department.  No_Ax_to_Grind | 11/20/03
What is limiting duh scope of duhvelopment efforts  Still Lynn | 11/23/03
mind you AMD is on its physical limits  adecrignis@... | 11/21/03
not sure which is cheating...  liberalenextrema@... | 11/21/03
Superior Chip  tcom5@... | 11/21/03
Athlon/Athlon 64 4GHz...  liberalenextrema@... | 11/21/03
AMD SHOUDN'T DECIEVE US WITH "3200" NOMENCLACURE.  THOMAS STEWART VONDRASHEK | 01/22/04
Intel will get it right! And Faster too!  bunnyman | 02/10/04

What do you think?

SponsoredWhite Papers, Webcasts, and Downloads

advertisement

Enterprise Applications

  • Check out some of the easiest and most powerful ways to boost productivity while saving money on your application infrastructure. See ZDNet's comprehensive Enterprise Application resource center, now!
  • New Online Dashboard
  • Read about top issues IT decision-makers face every day, plus get cost effective solutions to real life IT problems. Oracle Topline