No end in sight for ramping processor speeds

When will bigger-better-faster end? Maybe never, if processor vendors have their way.

Most agree that Moore’s Law, which states processor performance will double about every 18 months, cannot continue indefinitely. However, Doug Cooper, Canadian marketing manager for Intel of Canada in Toronto, says there is no technical reason why Moore’s law will not continue until at least 2005.

“Why more performance? The way people use computers has changed. People don’t run one application or one piece of software now, exit it and start a new one. They typically have three or four things on their desktop at one time,” he said.

“In addition to that there are things like virus scanning and disk compression and decompression that just run in the background. These all affect performance.”

The need for faster systems continues to grow as the Web evolves and new technologies like voice recognition, 3-D interfaces and e-commerce take hold, he said. He also pointed out that processor speed is not the only thing that affects overall performance.

“PCI bus eliminated a barrier that had existed with the 486. The graphics performance was really limited to how fast you could push bits onto the screen. And PCI multiplied that by a factor of at least 10. Now, because we are introducing chipsets and processor technology in synch with one another, some of the larger roadblocks and bottlenecks have gone away,” Cooper said.

Byran Longmire, K-6 product marketing manager at Advanced Micro Devices Inc. (AMD) in Sunnyvale, Calif., agreed the processor market will move at an aggressive pace for the next few years and stated that AMD will most likely market 1GHz processors in the year 2000.

According to Longmire, when the company’s K-7 chip is released, it will feature an ED-6 alpha bus that will scale up from 200MHz because “bus speeds can actually hold back the performance of the processor getting out and actually being realized by the rest of the system.”

Throughput is affected by many other factors, including SCSI controllers, memory and caching capabilities, said Mike Oreskovic, product manager for extended desktop business units at Hewlett-Packard (Canada) Ltd., in Mississauga, Ont.

Oreskovic said the computers of the future will be more design specific, with a wider range of appliances for specific tasks. “Things will continue to get faster, but I think there will be more of a segmentation of the market.”

So although processor speed will continue to be a factor driving the future market, it will probably not be the most compelling reason to buy a new computer appliance, he said.

Tony Befi, vice-president for RISC/6000 product management at IBM Corp., in Austin, Tex., said there has been several points in the past where it looked like system performance was finally going to level off, but so far this hasn’t happened.

“A few years ago, you might have said that the technology to connect the transistors together on the chip was the thing that was going to cause Moore’s Law to turn over.”

If IBM had not started using copper interconnections, that probably would have been true, he said. “But by using copper, we have broken that barrier by lowering the resistance of the interconnect, allowing for much denser chips and much higher speeds.”

So the world doesn’t revolve around processor speed alone, Befi continued. “But if you don’t have a capable high-speed processor, then you don’t have a fundamental building block. I don’t see a point where people say ‘I just don’t need to go any faster.'”

He said application needs drive increases in processor speed. “But on the other hand, faster speeds open up new areas for technologies that were impossible before.”

When the Internet became prolific, it suddenly opened up a huge opportunity for application development, he said. And once that opportunity was there, then those applications drove the advancements that were made in computing.

Despite this, not everyone feels that faster is better.

Tormi J