Clemson (SC) - Researchers at Clemson University's Center for Silicon Nanoelectronics have developed a new process technology which reduces power loss by six times, "to a level never seen before," according to the press release.  This new technology results in reduced heat, and will allow future multi-core chips to have less physical cores because each one can be clocked higher and do more work.  Future handheld devices could also benefit by requiring less power to do the same work seen today.


A process technology is the physical manufacturing processes involved in making a semiconductor device.  This typically involves creating things in the sizes we're familiar with, using 65nm, 45nm, etc.  Tiny copper lines are created which connect all of the on-die circuits necessary for carrying out work.  Transistors are the fundamental workhorses of the semiconductor.  It is typically these devices where the bulk of the heat is generated.  Each transistor generates an incredibly small amount of heat, but in today's semiconductor devices there are literally 100s of millions of transistors.  Their cumulative heat effects become significant.

Heat generation is also affected by clock speed.  If a modern 3.0 GHz processor dissipating 65 watts of heat at 3.0 GHz were taken and underclocked to 500 MHz, it might be able to operate without any active cooling at all.  The slower clock speed results in less heat being generated.  And, if that same 3.0 GHz processor were taken and overclocked to 4.0 GHz, it might require a high-end liquid or some type of higher-end active cooling system that can cool the chip down to minus 40C or so (due to the temperature differential inside the chip).

The Clemson scientists are claiming to have developed a new process technology which will allow for much less heat to be generated at a reduced manufacturing cost.  According to their press release, their patented method "has the potential to improve the performance and lower the cost of next-generation computer chips and a number of semiconductor devices, which include green devices such as solar cells."

No real details on the process technology, materials used, manufacturing methods, or anything further was given in the release.  It appears to be a notification about what they have, not how they got there.

Researchers at Clemson include Aarthi Venkateshan, Kelvin F. Poole, James Harriss, Herman Senter, Robert Teague of Clemson and J. Narayan (of North Carolina State University at Raleigh).  These results were initially published in Electronics Letters on October 11, 2007.  See Volume 43, Issue 21, and specifically pages 1130, 1131.  The patent mentioned here is covered by an existing broad-base patent issued to Singh and Poole of Clemson University back in 2003.


Author's opinion
It will be interesting to see if this one truly pans out.  Over the years, many scientists and researchers have come forward making huge proclamations about some new technology or discovery which will fundamentally change the way semiconductor devices are manufactured.  To my knowledge, unless they were silently incorporated into existing designs, microprocessors haven't suddenly moved forward significantly from something like a 90nm copper process to a 2nm gold process.

A decrease of 6x the power loss would result in a significant reduction in generated heat.  CPUs which are limited by their thermal envelopes today would not be so with such process technology at their hearts.  Speeds of 4.0 GHz, even 5.0 GHz, with the same thermals we see today, could be possible.
 
 

Part of the problem...

Dec 03, 2007 11:33     

I would just like to see a system running at 3ghz with passive cooling. Might be better used in video cards that are loud and hot, maybe hit SLI speeds with one card and save noise and heat to boot. The article says there's a cost saving, but somehow I don't see that translating to the consumer end, so maybe we can jump everything a year or so... Also, are we looking to cool chips to -40c or by 40c? Keeping a running processor anywhere a negative anything temp wise seems kinda tough, and kinda silly, they run fine at 40c, least, it seems they do.

Jossrik report abuse vote down vote up Votes: +0

I am sure they meant 40C. [-40C, indicating that i...

Dec 03, 2007 12:43     

I am sure they meant 40C. [-40C, indicating that it is super-cooled. There is a temperature differential between the heat sink and the CPU insides. The -40C will make sure that such high overclocking will not damage the chip. -Rick]
Luke report abuse vote down vote up Votes: +0

Could be interesting.

Dec 03, 2007 14:44     

Go Tigers ... sorry just had to get that out there since this is coming from my alma mater LOL. Anyways I think if this truly pans out then it'll be a great thing that could potentially result in faster processors and/or cost savings which I'm all for.
rodneyg report abuse vote down vote up Votes: +0

Not a big surprise..!

Dec 03, 2007 15:58     

It is possible that some of those so called advancements are being bought by big companies but they will introduce this new stuff into their next generation release. They will always give out advancements gradually. That's how big companies operate. They want to keep you interested just when you think it's not possible to get better. Just my crazy theory. (Conspiracies...Conspiracies...)
WRACK report abuse vote down vote up Votes: +0

-20C to -40C is about right

Dec 03, 2007 16:16     

They do in fact, mean what was stated. Please google "Vapochill Cooling Systems" if you're curious. Most of these achieve temperatures of -20C or better, and the author may have been referring to high end equipment, some of which is cooled via liquid nitrogen. Also, please see: http://www.tomshardware.com/2003/12/30/5_ghz_project/ from Tom's own site.
Quartz report abuse vote down vote up Votes: +0

Diamond in silicon substrate

Dec 03, 2007 16:27     

Singh and Poole: http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/7858/21636/01003152.pdf
Cerian report abuse vote down vote up Votes: +0

No, they meant -40c

Dec 03, 2007 16:38     

Chips overclocked significantly require liquid nitrogen cooling, keeping the chip cooled to temperatures in -40c and beyond (liquid nitrogen boils at −196.5 °C, so keeping it liquid it must be cooler than that). As mentioned in the article, the actual temperature inside the die is a lot hotter, so keeping the outside of the chip at -40 is what is required to keep the temperature inside the die at a usable level.

Chris report abuse vote down vote up Votes: +0

@Rick re: super-cooled CPU

Dec 03, 2007 17:06     

Didn't someone hit 8GHz on something like a Pentium D using LN2? (liquid nitrogen) [On a Pentium 4, not on their current Core architecture. The Netburst platform is more conducive to higher clock rates. However, a 5.5 to 6 GHz overclock on Core would like best the 8 GHz P4 in performance and throughput. Also, I don't remember reading how stable the machine was. It may have POSTed, booted and ran CPU-Z, but not much more. It's only when the machine is really processing that it would overheat and lock up. Personally, I never understood why people care about a machine able POST, but not boot at a given clock rate. What's the point? It's a boat anchor at that point. My personal choice for overclocking abilities is, at what speed will it POST, boot and run anything you throw at it without locking up, rebooting, computing incorrectly, etc.? That's the true test, in my book. -Rick]

MtYermom report abuse vote down vote up Votes: +0

6 fold vs. 6 times

Dec 03, 2007 18:18     

why is it that so many technology news articles, no one knows the difference between 6 fold and 6 times? They are a HUGE difference, but it so many articles, espcially on here, I see they use "fold" in the title and "times" in the article. "Fold" is the same thing as exponential. 2(6 fold)=2^6=64 2x6=12 [According to dictionary.com, "sixfold" is six times. -Rick]
Stephen report abuse vote down vote up Votes: +0

How about silver at 1nM?

Dec 03, 2007 18:34     

Interesting you mention the possiblilty of using gold instead of copper. I've always thought of gold as being the ultimate conductor but actually silver is a better conductor than gold! I remember when aluminium was the only means of making interconnects until it became more like a resistor as the semiconductors shrunk. Fortunately someone discovered that copper still conducts well at the nM sizes and figured out how to use it. Is copper now becoming the limiting factor as aluminium was? In any event it seems Moore's Law is going to continue...

big_davos report abuse vote down vote up Votes: +0

8GHz P4 vs 5.2GHz C2DX6800

Dec 03, 2007 19:14     

1M Pi decimals: 18,516s vs 9.84s What is not clear is if the C2D was using 2 cores to calculate pi. If that was the case 1 C2D core at 5.2GHz calculates Pi as fast as a P4 at 8GHz.
Peta report abuse vote down vote up Votes: +0

sounds like...

Dec 04, 2007 03:29     

another paper victory... "No real details on the process technology, materials used, manufacturing methods, or anything further was given in the release. It appears to be a notification about what they have, not how they got there." This paragraph is enough to prove my point. No offense.. but that's what I feel... I think what we need is adaptive control system for processor frequency based on the type of work. for an example.. typical user can do wither one of them plain web browsing office document editing video playback hardcore gaming (in my point of view) ROUGHLY, for plain browsing X00 MHz is enough. Office.. maybe around 1GHz...Video Playback 1-1.5 GHz (actually.. depends on the format.. High Def needs more processing power).. gaming.. top freq (2.5-3GHx). If the processor can change frequency based on the app.. its much worthwhile rather changing it based on the work load. Why I say so... when video play back... it doesn't need the 100% workload on the processor... so why increase the voltage on processor???.. just keep it low where sufficient freq. is provided. Afterall.. video playback not gonna shorten its time even all cored at top freq. My capitalisation.. make processor more smart.. so that based on app.. freq changes. I guess.. there should be a hardware-software integration to make this happen. Its a long thing... maybe.. better not do it.. (no point changing the whole micro architecture for one purpose)

Alien report abuse vote down vote up Votes: +0

Shrink

Dec 04, 2007 07:02     

One thing that has always bothered me is that Videocard companies never even care about shrinking their processes. Intel and AMD try to push lower processes to achieve better yields and lower temperatures. Isn't Nvidia's 8000 series based on a 90nm process?
Fabio report abuse vote down vote up Votes: +0

As an aside: pretty nice article

Dec 04, 2007 12:55     

I'd just like to say that I appreciate this article. It shares the press release, but has been made into a real article: it adds some background information (semiconductor power usage) for understanding, and some extra research (Electronics Letters cross reference) to help the readers quickly know the context of the announcement. I also appreciate the added opinion -- especially because it is marked as such. A lot of us readers post complaints when these details are not present, so I'm glad to get give some credit when they are here. Pretty nice work, Rick.

Dave report abuse vote down vote up Votes: +0

Re: Not a big surprise..!

Dec 05, 2007 19:51     

The simple answer to companies that release incremental improvements is simple. That's the only way to advance. The fact is, each incremental improvement costs lots of money on the technology side. Without selling something to make up that money, these big businesses would go out of business. It's not like these companies have 50Ghz processors and and all kinds of toys they are reluctant to release because they want to take the 'incremental money.' The truth is, the semiconductor industry is actually struggling (Yes, I have talked to people in the industry) to keep up with Moore's law. Quite frankly, they are doing a damn fine job!

Paul Wowk report abuse vote down vote up Votes: +0

Easily find electronic and semiconductor datasheet

Dec 11, 2007 01:09     

"DSP Innovations has recently launched a FREE data sheet search site. The site, www.datasheetpro.com, enables component and design engineers to easily find electronic and semiconductor datasheets necessary for their projects. Please try out the service and provide your feedback to support -at- datasheetpro [dot] com."
Catherine report abuse vote down vote up Votes: +0