IBM eyes industrial use for high-performance Cell chip

IBM Corp.'s design services unit has begun promoting the new Cell microprocessor, developed together with Sony Corp. and Toshiba Corp., to potential users outside of the consumer electronics industry in which the chip is likely to first appear.

IBM Engineering & Technology Services (E&TS) will offer companies looking to use the processor help in designing Cell-based hardware and systems, it said Wednesday. Cell was developed by the three companies and Sony-unit Sony Computer Entertainment Inc. (SCEI) as a high-power processor particularly suited to graphics applications and has been described by the development partners as a "supercomputer on a chip."

Cell is perhaps currently best known as the chip that will form the heart of SCEI's successor to the PlayStation 2 game console and both Sony and Toshiba have also talked about using the chip in a range of video-related consumer electronics products, such as televisions. IBM E&TS said Cell will especially target applications in which graphics processing is important, including those in the aerospace, defense, industrial or medical segments.

Customers will be able to access a Cell software simulation environment through IBM's Deep Computing Capacity on Demand Center.

Despite its roots in the consumer gaming world, at least one industry insider thinks Cell could provide some competition for established processors in such industrial applications.

"Some of the architecture of the Cell came from IBM's networking group," said Kevin Krewell, editor of Instat/MDR's Microprocessor Report. "There's a good amount of similarities between high-processing network processors and Cell. With a few changes I could see it being used in networking applications. Also, if you look at 3D image processing there's a lot of parallels there."

The chip is likely to provide competition for general processor makers like Intel Corp. and also graphics processor specialists like NVidia Corp. and ATI Technologies Inc., he said.

One possible worry for would-be users is supply of the chips, said Krewell. After SCEI launches its next-generation PlayStation there could be a danger that the game console pulls all or most of the chips produced, leaving others to fight for remaining chips. He cited some problems IBM has had supplying processors to Apple for use in Apple's computers. However, he noted that Cell production would be handled by all of the development partners and not just IBM alone, so supply may be enough to meet all demand.

Moore's Law was not a law at all - (one example can disprove a Law) - it is not even a theory, it is not even a postulate. It is debunked. The Cell and all future processors will be multi-cored.

Huh? Moore's Law is an empirical observation and prediction, not a law per se, but those are semantics. Moore's observation is valid, and multi-cored chips are further proof of Moore's Law.

Moore said is the number of transistors that could be manufactured on a single silicon chip doubled every 12 months. He later revised this to 18 months, and later expanded it to 24 months.

There are two technology drivers of Moore's Law: silicon wafer manufacturing and lithography. Larger silicon die sizes and smaller line traces combine to increase to increase the number of transistors per chip.

Moore said nothing about chip frequency increases over time.

The reason we are seeing multi-core chips now is because the number of transistors on a single chip now greatly exceeds the number of transistors required to construct a single complex central processor unit core. In other words, the exponential growth of transistor density exceeds the growth in the ability of processor architects to use those transistors to create a single processor core. Therefore, replicating multiple CPU cores on a single chip is the most effective way to make use of the available transistors.

In the 1970s central processor units were constructed on circuit boards, and a computer required multiple circuit boards. In the 1980s, CPUs were on a single chip, but components like math coprocessors were on another chip. In other words, an entire single processor computer could be constructed on a single circuit board. In the 1990s the coprocessor and other things were brought on the chip, and a multi-processor computer could be constructed on a single circuit board. This decade, multiple CPUs are on a single chip, so the multi-processor computer no longer requires a circuit board, it can be done on a single chip.

All of these are examples serve as proofs of Moore's observation and prediction.

Moore's Law was not a law at all - (one example can disprove a Law) - it is not even a theory, it is not even a postulate. It is debunked. The Cell and all future processors will be multi-cored.

The law was really an observation that integrated circuits were doubling in complexity every year, and he guessed it would continue for 10 years. Ten years later he revised it to 18 months as things had slowed down.

That's all it's about -- transistor count goes up. That can be translated into computing power, or cost savings, or energy savings, but all those are extrapolated from the transistor count. The Cell can go multi-core because they can now pack enough transistors on the chip, due to Moore's Law. It hasn't failed yet, but may eventually due to the limitations of physics.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.