Efficient Chip

[xoxoxoBruce]

From the Dallas Business Journal:

Researchers at Texas Instruments and Massachusetts Institute of Technology have drawn the curtain back on an energy efficient chip design that they believe will lead to self-charging cellphones and dramatic advances in implantable medical technology.
A team of researchers at Cambridge, Mass.-based MIT and Dallas-based TI worked together to create a microchip that requires significantly less electricity than the industry standard of 1 volt. Their creation, which will be shown to the International Solid State Circuits Conference in San Francisco, Calif. on Monday night, requires only 0.3 volts, a two-thirds reduction in required electricity.
In order for consumer technology to take advantage of the breakthrough, said MIT Professor Anatha Chandrakasan, memory and logic circuits will have to be redesigned due to the new chip's unique DC-to-DC power conversion, which takes place on the chip instead of through an external component. Products bearing the new chips could be available in five years.
TI and MIT researchers envision the breakthrough leading to cellphones that run off ambient energy; portable computing devices with vastly improved running times; and medical implants that are powered simply by body heat and movement.
"We are proud to be part of this revolutionary, world-class university research," said Dr. Dennis Buss, chief scientist at Texas Instruments.
The project was funded in part by a grant from the U.S. Defense Advanced Research Projects Agency.

[tw]

The TI chip consumes less power. Easily understood by any communication major who did not understand details in that simplified press release. Naive computer assemblers are always hyping more fans. Therefore a chip that uses less power is a major accomplishment? This TI technology will appear in their MSP430 series microprocessors. Power decreased by making transistors smaller to operate at lower voltages, and locating a DC to DC converter on the chip. Well, most all chips contain a DC to DC converter (so that a negative voltage supply is not required - the difference between Intel's 8085 and Zilog's Z-80). Smaller transistors mean operating at lower voltages. Nothing new. Earliest Intel Pentiums did same; but with the DC to DC converter on the motherboard. Pentiums are so much larger and run so much faster that the converter must be separate. TI simply made a slower chip with same small transistors so that another DC to DC converter could be moved off motherboard and onto IC.

A nice little improvement of same significance was announced by Intel. A 'longest to execute' instruction is division. Intel recently improved division to reduce execution time by one half. A significant accomplishment just like the TI announcement.

Meanwhile, that communication major would not appreciated major industry changes occurring at the same conference. Generations ago, non-volatile memory use principles discovered in the early 1800s. 40 years ago, Intel pioneered static memory (21xx series), non-volatile memory (27xx and 28xx series), and microprocessors (80xx series). Radical changes were announced in memory. Conventional memory was about storing electrical charges. STMicroelectronics and Intel have finally started shipping samples of phase change (ovonic unified) memory. Data stored by changing structure of materials. Freescale and NEC have started shipping magnetic memories. NEC has set a new speed record at 250 MHz for 1 Mbit memory. Texas Instrument and others are shipping ferromagnetic memories that do in semiconductors what was once done by Space Shuttle's iron core memories.

Researchers in Eindhoven Netherlands set a new 780 bits / second speed record for a 64 bit RFID chip at a distance of 10 cm.

MIMO (Multiple Input, Multiple Output) technology is the future in non-America digital TV, Wi-Fi (802.11N), WiMaxLTE, UWB radios, and 4G cell phones. Multiple transmitters and receivers on separate channels provide faster and more reliable data rates compared to using channels on sub-carriers. Another major change previewed this week. Not grasped by communication majors that only saw 'low power' in a TI product. Only saw what the reporter could understand.

Some familiar standards are 802.3 (ethernet networking) and 802.11 (WiFi). The latest standard is 802.15.6 - body area networking. IOW McCoy's 'sick bay' bed is obsoleted by electronics embedded in a body. Simply enter a room to be medically diagnosed? Networking identifies an epileptic seizure before the seizure starts.

What will be the new Moore's law? A controversy. Multicore processing is the future. "I would like to call it a corollary of Moore's Law that the number of cores will double every 18 month", according to Anant Agarwal of Tilera Corp that ships a 64 core processor. He also predicts in 2017, an embedded processor will have 4096 cores. That's 4096 computer processors in one chip. Rich Hetherington of Sun believes it will only be 32 to 128 processors by 2018 where each multicore will process 500 to 1000 threads. IBM's McCredie has a more traditional belief, "Today's data centers run a big pile of goofy apps, where many people don't even know where their source code is anymore. Google is an exception. We still have discussions about single threaded apps with customers who may run these applications forever."

Whereas Moore's Law was the benchmark for hardware innovation, software has no such benchmark. Transmeta founder Dave Ditzel discusses his experience in Sun Microsystems. He describes designing Sun's first 64 bit CPU that waited almost a decade before a 64 bit Operating System was finally developed.

Industry earthquake changes in memory and wireless communication make that low power TI chip just another innovation. Who would appreciate so many other and more significant announcements at the same ISSCC? A communication major would only understood "Less Power" (Tim Allen?). Multicore programming is a looming 800 pound industry gorilla that all major semiconductors manufacturers are grappling with because software programmers just don't yet grasp the technology - multicore processors.

[classicman]

OMG - could you put that in english for the rest of us - - - - ok well for me anyway?

[xoxoxoBruce]

I got, from reading the news release, they aren't trying to increase computation speed or memory capacity, but trying to reduce power consumption.

TI and MIT researchers envision the breakthrough leading to cellphones that run off ambient energy; portable computing devices with vastly improved running times; and medical implants that are powered simply by body heat and movement.

[lumberjim]

i bet tw's post is interesting.

[xoxoxoBruce]

Well, yes and no.

[tw]

Originally Posted by xoxoxoBruce

I got, from reading the news release, they aren't trying to increase computation speed or memory capacity, but trying to reduce power consumption.

That newest MSP430 would make some very interesting applications. TI was actively promoting contests for best application of their MPS430 processor. But then even 1960 technology processors (RCA's COSMOS) consumed so little power that the largest consumer of battery power was the battery itself (also called shelf life). Watch batteries typically are not run down by the watch. That battery would die just as fast whether in the watch or just sitting on the shelf - again shelf life. Processors in watches 30 years ago consumed so little power.

Other reports also cited that TI 'low power' as a significant development. Significant developments from ISSCC included new memory technologies, MIMO, and discussions about multicore processors - the new Moore's law?

Reporters did what many computer assemblers also do. When a computer does not work, then blame only what they understand - heat. Then install more fans or hype Arctic Silver myths.

If more journalists had enough grasp to appreciate what really was significant this past week. Three new memory technologies are now avidly in competition to replace disk drives.

Not even mentioned was old news of an entire radio created in a buckyball tube. So small that its connecting wires are larger than the entire radio. Ahh, but that working radio was only a research project - not a new product.

Originally Posted by tw

... If more journalists had enough grasp to appreciate what really was significant this past week. Three new memory technologies are now avidly in competition to replace disk drives. ...

I've only read about the solid state drive in which the chief advantage seems to be durability, with only a few applications benefiting from a somewhat lower power consumption, and the major drawbacks currently being half the memory capacity at a 50% increase in cost making it impractical for most consumers at this time.

Are the other new technologies any better poised to see widespread application?

[tw]

Originally Posted by NoBoxes

I've only read about the solid state drive in which the chief advantage seems to be durability, ...
Are the other new technologies any better poised to see widespread application?

"Conventional memory was about storing electrical charges." A "solid state drive" is just that. An earlier attempt to replace disk drives with solid state was called "bubble memory". A potential 'disruptive innovation' called bubble memory eventually failed to make it.

Rephrasing the question is, "Are these disruptive innovations?" The disk drive inside a computer (called a cell phone or mobile) uses a solid state disk. Market for such drives has long existed. A 'less than one inch disk drive' existed maybe a decade ago. Can these new memory technologies do better? How long does it take for a disruptive innovation to change the marketplace? How many just 'knew' something was a disruptive technology when first made available?

We know these three memories are fundamentally radical new technologies to replace "memory by storing electrical charges". Companies such as Intel sold off their 'electrical charge storage memory' production lines to move into new memory technology. (Memory is the first component that Intel designs for each new technology processor.) We know these memories are technically superior to non-volatile memory found even in curent solid state disk drives. But does it meet a ballpark criteria of tens times improvement? Is it so disruptive as to threaten conventional disk drives?

Only time and innovation can answer that question. If there was a simpler answer, then I would not even bother posting the obvious. No way to answer that question with anything but a 'maybe'. New memory technology would have been a reporter's topic if he understood simple computer hardware concepts.