just brew it! wrote:IMO the looming battle is teaching more developers how to write reliable multi-threaded code. Since clock speed isn't scaling any more, the only way to squeeze more performance out of our systems is more parallelism (more cores).
JBI wrote:[/color]]Sounds like we are overdue for a new metric to replace the traditional "process node" designation. Something that factors in number of transistors per square mm, clock speed, and power usage, perhaps?
Krogoth wrote:Moore's observation has already been invalidated almost a decade ago.
Moore wrote:The complexity for minimum component costs has increased at a rate of roughly a factor of two per year.
Krogoth wrote:Intel's PR department is trying to keep "Moore's Law" alive by shifting the focus to power consumption reduction, however this has its own limits. We are closing onto the hardware limits of what semiconductor-based computing can do. That's why the new focus is shifting gears onto a new computing model. Quantum computers are getting most of the buzz.
JBI wrote:Since clock speed isn't scaling any more, the only way to squeeze more performance out of our systems is more parallelism (more cores).
Glorious wrote:But that's not really the point of what he was saying, or how you (and most everyone) has misconstrued what he said.
Krogoth wrote:Power consumption is driven both by "clockspeed" and active transistor count.
krogoth wrote:Adding more cores/ICs into the design will still run into the problem of power consumption and thermal output if they are all being utilized.
krogoth wrote:The current power saving schemes on modern chips involved reducing voltage, clockspeed and putting idles parts of the silicon into a "sleep state".
Krogoth wrote:Moore's observation has been invalided by scale of economics on the production end.
Krogoth wrote:It is becoming more and more expensive to move to the next process node while the returns are diminishing.
Users browsing this forum: No registered users and 1 guest