4 Core – 99 cents

INTEL Got Crushed in the 70s, Redux

 

INTEL VP of Marketing, Bill Davidow(1) was the premier technology marketing wizard of the late 20th Century.  The first words of his masterwork, Marketing High Technology(2) capture the essence of high tech corporate combat,

 

            "Marketing is civilized warfare". 

 

The master marketer described having INTEL's pride and joy 8048 shot to pieces by Fairchild's 3870:

 

"The firm had never been a factor in the microcomputer market, and there was little reason to expect it ever would be.  ........ It was therefore all but inconceivable that a second-string squad from a last-place team could be doing that to us.  But they were, and we were losing orders all over the place."

 

The first big INTEL loss was a million pieces for the on-board computer at Chrysler Huntsville.(3)

 

INTEL lost because an obscure engineer discovered a way to make the 3870 for half the cost of the 8048.

 

Bill continued:

 

 "Faced with that competition, it became depressing even to make customer calls.  Big deals were being lost everywhere in the world. .........Before long the Fairchild 3870 was being quoted in the market at well below INTEL's manufacturing cost."

 

There is no worse feeling than discovering your enemy has a much bigger gun or a much lower cost of goods.

 

INTEL learned the lesson and a decade later went on to crush the second largest computer company in the world, Digital Equipment, by offering computing power at a much lower cost with the x86 architecture.

 

Fast Forward Four Decades

INTEL is no longer the scrappy startup that took on and beat the big boys.  The 20th century leaders have passed from the scene.  Noyce who invented the IC and won the Nobel has passed on to the great fab in the sky. 

 

From his yacht Gordon Moore is watching his "law" be repealed by Patterson's Walls.

 

Andy Grove literally wrote the book on MOS(4).  He solved the industry killing Vt drift.  "Only the paranoid survive," will be his epitaph.  He's been off the board for over a decade.

 

Now the company whose name is synonymous with microprocessors is facing an existential crisis on two fronts; one technical and one economic.

 

Law Hits Wall

The INTEL innovation miracle depended on "Moore's Law" first published in 1965(5).

 

The optimum number of transistors on a chip doubles every two years.

 

That two year cycle is still built into INTEL's "tick-tock" planning cycle of process upgrades and major design releases.  Moore's Law has run head on into Patterson's Three Walls(6), particularly "The Power Wall".  As legacy microprocessors get more transistors, they get hotter.  As they get hotter, they get slower.  There is no known technical solution.

 

INTEL's marketing solution is "multicore".  Customers get slower microprocessors but more of them on the same die. 

 

Multicore slams Patterson's "Memory Wall" even harder than single core.  Think of breathing underwater through a straw.  Now share that straw with three of your friends.  Like you and your friends, INTEL multicore is drowning.  Sandia Labs reports that the Memory Wall is so severe that 16-cores actually run the same speed as single core on data-intensive applications.(7)

 

Now TOMI Technology appears to traverse the Memory Wall, even without the latest processing:

 

"Venray's prototype can deliver 4.6 times the bandwidth of Intel's best

server processor despite using 10-year-old DRAM technology."

- Microprocessor Report

 

That Depressing Feeling Returns

INTEL's process technology has almost always made it the low-cost producer.  All who competed cut prices to gain market share while INTEL cruised along at 65% gross margin.  It has been over three decades since INTEL sales force felt that depressing feeling of being behind the cost curve. 

 

TOMI Technology(tm) redefines the eonomics of computing by combining CPU and memory and doing so using just about the cheapest transistors in production today.

 

The TOMI Aurora prototype is a 4-core CPU + 64Mbit DRAM that can sell for less than a dollar, even using a near obsolete 110nm process.

 

As described by Fish & Hong (8) using a two generation old 70nm process, a TOMI 8-core CPU + 1G DRAM clocks 1.66Ghz at 45mw/core.  It can sell for less than $3.00. 

 

As Bill Davidow wrote about from 1976, " "Faced with that competition, it became depressing even to make customer calls. "

 

(1) Now at Mohr Davidow Ventures. http://en.wikipedia.org/wiki/Bill_Davidow

(2) http://www.amazon.com/Marketing-High-Technology-William-Davidow/dp/002907990X/ref=sr_1_1?s=books&ie=UTF8&qid=1283552673&sr=1-1

(3) h/t - Ralph Laughlin

(4) http://www.amazon.com/Physics-Technology-Semiconductor-Devices-international/dp/0471329983

(5) http://en.wikipedia.org/wiki/Moore's_law

(6) http://www.slidefinder.net/f/future_computer_architecture_david_patterson/6912680

(7) http://spectrum.ieee.org/computing/hardware/multicore-is-bad-news-for-supercomputers

(8) https://www.venraytechnology.com/TOMIWhereDidthePowerGo.pdf