Forecasting a dead end for chips

The 4004 microprocessor

(Credit: Intel)

(Editor's note, 10:25 p.m. PDT: The original headline on this blog was altered to remove the word "subatomic" because its usage may not have been appropriate.)

TEL AVIV, Israel--One of Intel's legendary chip developers says that the computer industry may no longer be able to infinitely shrink microprocessor die size.

And if anyone should know, it's Marcian E. "Ted" Hoff.

The designer behind the 4004 chip, Intel's first microprocessor, Hoff says that a technical end of the road may soon be within sight.

"We are approaching atomic dimensions," Hoff said, adding that "5 or 10 nanometers is as small as we can get. And I haven't seen much to change my mind."

Intel's most advanced chips are at 45 nanometers, while 32-nanometer technology should be out sometime next year. But Hoff, Intel's 12th employee, whose technology breakthrough helped create the microcomputer industry, cautioned that before long, "progress may slow or stop someday."

Ted Hoff

(Credit: Silicom Ventures)

This shouldn't come as a complete surprise. Hoff and other Intel officials have warned for years that they would eventually run up against a technical wall they might not be able to hurdle.

He said that development is getting much harder and that Intel is going to have to do some novel things to maintain the same technical momentum. Talking about previous generations of Intel microprocessors, Hoff said that "each one has been a variation of the previous one. So, essentially, we're running our computers on glorified 8088s."

Of course, Intel has been wringing its hands about Moore's Law for quite some time, and still the company manages to pull through.

[tacit]

Sloppy reporting--or at least, sloppy title. 5-10 nm is not "subatomic." The forecast dead end for components on an IC is decidedly above the atomic threshold.

[The_Decider]

IMO, the future of desktop chips is not larger and larger multicore systems but 2 or 4 chip processors that improve in other areas like cache, while the OS manages the chips in a more intelligent manner. The reasoning is that most desktop apps are not inherently parallel, you simply can't get them to run in parallel. However, each chip can be dedicated to certain tasks, 1 for the OS, and 1 for the apps in the case of a dual core setup. In a quad core system, 1 for the OS, 2 for apps and one for AV and AS(windows only). This will also relieve the not trivial problem of managing data in 4 L1's caches, L2 and main memory. There is a lot of complexity and overhead inherent in having to do this. Hardware companies are constantly trying to get faster chips with more cores without stopping to think 'why are we doing this?' and 'Is this of any real benefit'. There is no need for it on the desktop, and the software needs to dictate hardware, hardware can not dictate software. Progress for the sake of progress has given birth to craptastically horrible software design that uses the fact that hardware gets faster over time to hide its sloppy design and programming. Microsoft is only one smelly piece of rubbish in a huge pile of refuse that is the software landscape. It has also help spread this toxic and odd idea of spending resources parallelizing software that might only be 10-15% parallel at best. That is a massive waste of resources with no real benefit to the end user. Of course, once you leave the desktop there is a need for chips that can handle the hardware complexity associated with parallelism and that is where real benefit is seen.

[tech_crazy]

With all due respect, is this guy nuts or what? All the computer architecture features that have gone into these and other processors amount to nothing? Talk about blowing your own trumpet!

[3tire]

When you get into line widths of single digit nanometers, your structures can be influenced by "subatomic" forces. So, depends on how you read it.

[artistjoh]

The interview with Steve Jobs following his announcement of the iPhone 2.0 was very illuminating on the Apple view of this. He discussed a new approach he calls Grand Central as an integral part of the next OS version with true parallel computing to make full use of multiple cores. The central (and overlooked in the iPhone excitement) plank of this is Apple providing the tools to software developers to make it easier for them to make fuller use of the various cores. If Apple's recent easy to use SDK tools for the iPhone are any indication, we may be seeing the opening up of the multi-core dilemma in a big way.

It is interesting to speculate that the very close relationship Apple has developed with Intel in recent years (example: Intel developing a special small size processor for the Air at Apple's request) possibly means a very deep sharing of intellectual resources between the company's as they both come to terms with an eventual nano-meter limit and how to overcome it. The speculation that the solution will be shared by software developer in partnership with the chipmaker was confirmed by Jobs confident words.

Steve Jobs talked about the theoretical possibilities of this new approach (potential dramatic increases in RAM as example) that indicate that while improvements may not come from chip shrinkage at some stage, the processing performance improvements due to lateral developments are likely to escalate either as fast or even faster than we have seen to date.

[The_Decider]

All the tools to help facilitate parallel programming doesn't help on algorithms that are almost totally sequential.

[cintegrations]

http://technology.newscientist.com/channel/tech/dn14291-tiny-chip-could-keep-us-on-track-with-moores-law-.html

[osxxp]

"Hoff, Intel's 12th employee, whose technology breakthrough helped create the microcomputer industry, cautioned that before long, "progress may slow or stop someday."
Forecasting a dead end for chips may not be effecting Intel now; however Intel (if not others) may still stay in the game by custom tailoring chips and processors for target populations.

For example the gamers might want micro processing that is totally marketed for gaming if it only gains them a nano-thumper increase in game play Medical technology for analysis and even implants may need to have their chips or the whole processor custom made.

Heart impairments, even cardiac arrhythmia, are served by miniature "paddles" that are placed into the heart itself as implants [clear!] My father has one. Would he be better served by multiple implants that are "integrated" into the heart?

The heart and various organs might have several other multiple implants and it could be the fulfillment of hardware solutions to disease.

Who knows what other markets will want "designer processing" If that doesn't work Intel may want to take a page from Apples iPods and start selling Intel processors in a dozen fashion colors.

[Seaspray0]

Artistjoh, multiprcessor support has been around for years, i.e. microsoft NT workstation supported 2 processors and NT server supported 4, both developed in 1992. Apple has some catching up to do in this respect, since until recently they were using the powerpc chip. I don't forsee any problems since the underlying core code for multiprocessor support has been available for years and should be no problem integrating it. What normally lags are the applications being able to support muliple processors and that's dependent on 3rd party software companies. Another big improvement we all should consider is the number of bits the processor is capable of handling. In the 1980's, the 8 bit processor was the standard. We're now seeing 64 bit processors in personal computers. Everytime we double the bits, we can cut out all the code that requires handling data in excess of what the processor handles. That's extremely useful for things like floating point and graphical computations. It's not just the size that matters.

[Penguinisto]

Bzzt! PowerPC-based Apple machines have been able to do multiprocessing ever since OSX came out. Also, UNIX has been doing multi-processing since long before Windows NT was ever written. ;)

[Penguinisto]

One aspect that hasn't been considered is moving into three-dimensional CPU cores. While heating problems would naturally be an obstacle, there are working nano-tubing/coolant structure processes that can facilitate thermal dissipation, so it wouldn't be an impossibility. Plus, it is a far shorter trip between vertical layers than stretching traces outward towards the edges of an ever-increasing die size.