Intel showed off a new chip technology on Tuesday which, if successful, will allow one chip to act like two.
Called "hyper-threading," the new technology essentially takes advantage of formerly unused circuitry on the Pentium 4 that lets the chip operate far more efficiently -- and almost as well as a dual-processor computer. With it, a desktop can run two different applications simultaneously or run a single application much faster than it would on a standard one-processor box.
"It makes a single processor look like two processors to the operating system," said Shannon Poulin, enterprise launch and disclosure manager at Intel. "It effectively looks like two processors on a chip."
Paul Otellini, general manager of the Intel Architecture Group, demonstrated the hyper-threading technology at the Intel Developer's Forum in San Jose on Tuesday.
In the same speech, he showed off a 3.5GHz Pentium 4 running the computer game "Quake III" and managing four different video streams simultaneously. The Pentium 4 demonstration didn't depend on hyper-threading; instead, it came out as part of Intel's effort to show how consumers and software developers will continue to need faster PCs.
"There are a lot of tremendous applications on the horizon that will consume the MIPS (millions of instructions per second)," Otellini said. "Gigahertz are necessary for the evolution and improvement of computing."
Technically, hyper-threading takes advantage of additional registers -- circuits that help manage data inside a chip -- that come on existing Pentium 4's but aren't used. Through these registers, the processor can handle more tasks at once by taking better advantage of its own resources. The chip can direct instructions from one application on its floating-point unit, which is where the heavy math is done, and run parts of another application through its integer unit.
A chip with hyper-threading won't equal the computing power of two Pentium 4's, but the performance boost is substantial, Poulin said. A workstation with hyper-threaded Xeon chips running Alias-Wavefront, a graphics application, has achieved a 30 percent improvement in tests, he said. Servers with hyper-threaded chips can manage 30 percent more users.
The open question is whether software developers will latch onto the idea. Software applications will need to be rewritten to take advantage of hyperthreading, and getting developers to tweak their products can take an enormous amount of time.
Intel, for instance, has been working for well over a year to get developers to rewrite their programs to take full advantage of the features of the Pentium 4, which has been out for approximately nine months. The company even changed the migration program to speed the process of optimizing Pentium III applications for the Pentium 4.
Still, to date, only 30 applications have been enhanced to take full advantage of the Pentium 4, according to Louis Burns, vice president and general manager of the Desktop Platforms Group at Intel. But more are on the way, he said.
Otellini acknowledged that recruiting developers will take time.
"The real key is going to be to get the applications threaded, and that takes a lot of work," he said.
Nonetheless, adopting the technology to server and workstations applications should be fairly easy if the application already runs on dual-processor systems, other Intel officials said. "Thread your applications and drivers and OSes to take advantage of this relatively free performance," Otellini asked developers during his speech.
Hyper-threading, which will appear in servers and workstations in 2002 and desktops in 2003, is part of an overall Intel strategy to find new ways to squeeze more performance out of silicon. For years, the company has largely relied on boosting the clock speed and tweaking parts of the chip's architecture to eke out gains.
The performance gains to be achieved from boosting the clock speed, however, are limited. In all practicality, most users won't experience that much realistic difference between a 1GHz computer and one that contains a 2GHz chip, according to, among others, Dean McCarron, an analyst at Mercury Research.
Ideally, hyper-threading, which has been under development for four and a half years, will show meatier benefits. An individual could play games while simultaneously downloading multimedia files from the Internet with a computer containing the technology, Poulin predicted.
Hyper-threaded chips would also be cheaper than dual-processor computers. "You only need one heat sink, one fan, one cooling solution," he said, along with, of course, one chip.
Chips running hyper-threading have been produced, and both Microsoft's Windows XP and Linux can take advantage of the technology, according to Poulin.
Computers containing a single hyper-threaded chip differ from dual-processor computers in that two applications can't take advantage of the same processor substructure at the same time.
"Only one gets to use the floating point at a single time," Poulin said.
On other fronts, Intel on Tuesday also unveiled Machine Check Architecture, which allows servers to catch data errors more efficiently. The company will also demonstrate McKinley for the first time. McKinley is the code name for the next version of Itanium, Intel's 64-bit chip that competes against Sun's UltraSparc. McKinley is due in demonstration systems by the end of this year.
See Chips Central for the latest headlines on processors and semiconductors.
See the Hardware News Section for full coverage.
Have your say instantly, and see what others have said. Click on the TalkBack button and go to the Chips Central forum