2018-02-23 18:58:03 +00:00
|
|
|
|
---
|
|
|
|
|
created_at: '2016-11-13T10:39:44.000Z'
|
|
|
|
|
title: Why CPU Frequency Stalled (2008)
|
|
|
|
|
url: http://spectrum.ieee.org/computing/hardware/why-cpu-frequency-stalled
|
|
|
|
|
author: sajid
|
|
|
|
|
points: 51
|
|
|
|
|
story_text:
|
|
|
|
|
comment_text:
|
|
|
|
|
num_comments: 48
|
|
|
|
|
story_id:
|
|
|
|
|
story_title:
|
|
|
|
|
story_url:
|
|
|
|
|
parent_id:
|
|
|
|
|
created_at_i: 1479033584
|
|
|
|
|
_tags:
|
|
|
|
|
- story
|
|
|
|
|
- author_sajid
|
|
|
|
|
- story_12942732
|
|
|
|
|
objectID: '12942732'
|
2018-06-08 12:05:27 +00:00
|
|
|
|
year: 2008
|
2018-02-23 18:58:03 +00:00
|
|
|
|
|
|
|
|
|
---
|
2018-03-03 09:35:28 +00:00
|
|
|
|
Advertisement
|
2018-02-23 18:19:40 +00:00
|
|
|
|
|
2018-03-03 09:35:28 +00:00
|
|
|
|
[![charts small
|
|
|
|
|
view](/img/31759-1372099381585.jpg)](/img/0408_data-xlrg-1372100315886.jpg)
|
2018-02-23 18:19:40 +00:00
|
|
|
|
|
2018-03-03 09:35:28 +00:00
|
|
|
|
|
2018-02-23 18:19:40 +00:00
|
|
|
|
|
2018-03-03 09:35:28 +00:00
|
|
|
|
Image: Intel; Charts: Michael Vella
|
|
|
|
|
|
|
|
|
|
Not so long ago, competitive sorts would boast of the cycle rate of
|
|
|
|
|
their PC’s central processing unit. But now it seems the only people who
|
|
|
|
|
talk it up are the overclockers—hobbyists who push their CPUs beyond
|
|
|
|
|
their specified limits. There are two reasons: CPU clock rates peaked a
|
|
|
|
|
few years ago \[see graph, top\], and they aren't a very useful key to
|
|
|
|
|
chip performance anyway.
|
|
|
|
|
|
|
|
|
|
The clock keeps a processor's parts working in unison, like rowers on a
|
|
|
|
|
galley ship. Other things being equal, the more ticks you have per
|
|
|
|
|
second, the more work will get done.
|
|
|
|
|
|
|
|
|
|
So why not push the clock faster? Because it's no longer worth the cost
|
|
|
|
|
in terms of power consumed and heat dissipated. Intel calls the
|
|
|
|
|
speed/power tradeoff a ”fundamental theorem of multicore
|
|
|
|
|
processors”—and that's the reason it makes sense to use two or more
|
|
|
|
|
processing areas, or cores, on a single chip.
|
|
|
|
|
|
|
|
|
|
Intel reports that underclocking a single core by 20 percent saves half
|
|
|
|
|
the power while sacrificing just 13 percent of the performance. That
|
|
|
|
|
means that if you divide the work between two cores running at an 80
|
|
|
|
|
percent clock rate, you get 73 percent better performance for the same
|
|
|
|
|
power. And the heat is dissipated at two points rather than one. So even
|
|
|
|
|
though the cutting-edge logic chip gulps ever more power \[see graph,
|
|
|
|
|
center\], it isn't about to melt its way through the floor.
|
|
|
|
|
|
|
|
|
|
That bodes well for Moore's Law, which predicts that about every two
|
|
|
|
|
years, manufacturers will double the number of transistors they cram
|
|
|
|
|
onto a given bit of silicon. The fundamental theorem says that we'll
|
|
|
|
|
still be able to make full use of those transistors for a good long
|
|
|
|
|
time. If once the whole choir of transistors had to sing to the beat of
|
|
|
|
|
a single metronome, now it can split up into sections—and harmonize.
|
|
|
|
|
|
|
|
|
|
**Count Paces? Or Measure The Distance Traveled?**
|
|
|
|
|
|
|
|
|
|
The rising power consumption of CPUs \[graph, center\] made it less
|
|
|
|
|
attractive to focus on cycles per second, so clock rates stalled
|
|
|
|
|
\[graph, top\]. A better gauge of performance, the number of
|
|
|
|
|
instructions performed per second \[graph, bottom\], continued to rise
|
|
|
|
|
without betraying any hint of the stall. That's because work once done
|
|
|
|
|
in a single processor is now divided among several processing
|
|
|
|
|
cores—four of them in the case of Intel's Quad-Core chip \[below\].
|
|
|
|
|
|
|
|
|
|
![data f1](/img/dataf1-1372099277050.jpg)
|
|
|
|
|
|
|
|
|
|
Image: Intel; Charts: Michael Vella
|
|
|
|
|
|
|
|
|
|
Advertisement
|