Engineers propose method to eliminate wasted energy in computer processors

Mar 08, 2012

In today's computer processors, much of the power put into running the processor is being wasted.

A research team at Case Western Reserve University came up with a novel idea called fine-grained gating, which saves power and money in a couple of ways: less energy would be used, and less heat produced.

"Using less power produces less heat. Less heat means less cooling is needed," said Swarup Bhunia, professor of electrical engineering and science and an author of the research. "That can avoid the need for a big fan to cool off the processor, which saves a lot of money."

Processors are used in a variety of products, from computers to cell phones. Operational costs could be cut by more than one-third, the researchers say.

Bhunia, PhD student Lei Wang and PhD alumni Somnath Paul, whose work was funded by the Intel Corporation; presented their idea at the 25th International Conference on VLSI (Very-Large-Scale Integration) Design.

They received the award for best paper at the conference, held in Hyderabad, India Jan 7-11.

Bhunia explained that two parts of a processor consume power: the datapath and memory. The datapath performs computations and takes control decisions, while memory stores data.

The waste is built-in. Computing rarely requires everything that a processor is capable of all the time, but all of the processor is fully powered just the same.

For example, while the processor might not always be doing addition, the component that performs addition is still being powered.

One attempt to improve in processors is through something called coarse gating. It switches off an entire block of the processor that is not being used.

In the previous example, the coarse gating solution would be to just simply turn off the addition block when it is not doing addition.

The problem with this method is that most of the time, some part of every component is being used in a processor. Finding an entire block that is not being used at a given time is tough.

The Case Western Reserve team's fine-grained gating idea is to shut off only the parts of a component that are not being used at the time. While the addition component needs to be capable of adding extremely large numbers, it rarely needs to actually add large numbers. The processor might be using the addition block constantly, but the parts needed to add large numbers can be turned off most of the time.

Memory works the same way. A processor needs to be capable of storing large numbers, but seldom actually stores them.

This may not seem like much, but add everything up and it makes a big difference. The team calculated that the total power savings for a typical in a high-performance system, such as a desktop computer, would be about 40%.

Bhunia explained that fine gating can't be applied to current processors, but could be used by companies to build next generation processors.

This new method does not only help corporations though. With fine-grained gating, a smart phone battery that lasted eight hours could now more than 11.

That's three more hours of Angry Birds and Words with Friends, which is a win for everyone.

Explore further: Research project on accident-avoiding vehicle concluded

Related Stories

DonanimHaber leaks data on new AMD processors

May 17, 2011

(PhysOrg.com) -- When it comes to the world of computer processors it is all about what it going to happen next. Even if the current generation of processors can do everything that you want them to that tantalizing ...

Intel updates Atom processor roadmap

May 18, 2011

(PhysOrg.com) -- Intel is one the biggest names in PC processors, if not the single biggest, but as is the way with all markets as new things come in the landscape can change in the blink of an eye. When the ...

Intel Celeron D Processor Now Available

Jun 24, 2004

Intel Corporation today introduced the Intel® Celeron® D processor for desktop PCs. This processor line represents a new generation of Intel technology for desktop value market segments. Intel also unveiled ...

Intel® Pentium® 4 Processor on 90nm Technology

Apr 12, 2004

Intel has recently published an article called The Microarchitecture of the Intel® Pentium® 4 Processor on 90nm Technology. This paper describes the first Intel® Pentium® 4 processor manufactured on the 90nm process. It rev ...

Recommended for you

A spray-on light show on four wheels: Darkside Scientific

6 hours ago

Darkside Scientific recently drew a lot of gazes its way in its video release of a car treated to the company's electroluminescent paint called LumiLor. Electroluminescence (EL) is a characteristic of a material ...

Research project on accident-avoiding vehicle concluded

Sep 12, 2014

PRORETA 3 is completed after three and a half years of research work: The comprehensive driver assistance and automated maneuver concept supports the driver in keeping the vehicle in a safe driving corridor- ...

Making drones more customizable

Sep 12, 2014

A first-ever standard "operating system" for drones, developed by a startup with MIT roots, could soon help manufacturers easily design and customize unmanned aerial vehicles (UAVs) for multiple applications.

User comments : 9

Adjust slider to filter visible comments by rank

Display comments: newest first

jimbo92107
5 / 5 (2) Mar 08, 2012
Having just heard this idea, how long it would take Intel to bring it to market?
dschlink
not rated yet Mar 08, 2012
2-3 years
hyongx
5 / 5 (3) Mar 08, 2012
2-3 years

"Bhunia explained that fine gating can't be applied to current processors, but could be used by companies to build next generation processors."
The next generation of processors, and associated architecture, still need to be developed. I say 3-5 years.
Sepp
5 / 5 (2) Mar 08, 2012
Great idea but here is a simpler one that doesn't require processor re-design: Slap a heat-to-electricity conversion module on top of every processor and use it to trickle charge the battery. Simple, and it would do the same thing... recover some of that wasted heat and ultimately save on energy use.
Husky
5 / 5 (2) Mar 08, 2012
saves power... ...or creates more room for overclocking
MIBO
5 / 5 (2) Mar 08, 2012
I don't understand what's new here, we've been using fine grained clock gating got many years.
Clock gating alone is indequate since significant power is consumed both in the gates and interconnect, and gating unused parts of the data-path is also common practice.
there are 2 main problems that have to be overcome and this doesn't address either of them.
first clock / signal gating requires additional gates which increase delays and consume power, so the power savings have to be worth inserting the gates.
secondly inserting gates into clock trees complicates clock tree balancing and these also require additional control logic to enable clocks during test.
The real issue is the complexity of design with fine grained clock gating, particularly for clock tree balancing and for test, these simply require better tools.
The real savings come from architectural design to minimize toggling without additional complex clock / signal gating.
Caret
5 / 5 (1) Mar 09, 2012
Very interesting approach. It seems they are not after clock gating. Instead they apply power gating to the unused part of the computing units such as the ALU and the memory. They seem to exploit data width (large or small) which is smart.
MIBO
5 / 5 (5) Mar 09, 2012
I did this several years ago in low power GPS chipset design.
http://www.faqs.o...90167594
It's not a new idea, actually we went a lot futher by architecting the data processing to minimize toggling and using this to specifically create an architecture that allows more efficient gating for part of the datapath that is not required in real time.
AXB
5 / 5 (3) Mar 10, 2012
Based on my experience power gating for a general-purpose processor is different ball game. The nature of data is very different from that in specific applications. Looks like the work has addressed some important challenges. Well done!
KBBa
Mar 14, 2012
This comment has been removed by a moderator.