Stay on top of the latest tech news with our free IT News Digest newsletter, delivered each weekday.
Automatically sign up today!
Intel plans to release chips with two processing cores next year, but that’s just the start.
The Santa Clara, Calif.-based chip giant intends to exploit the concept of using multiple processor cores–the part of the microprocessor that does the math–as a way to increase performance of its chips during the next few years.
Chips with four cores and eight cores will eventually join dual-core chips. The company’s research department is also looking at the feasibility of creating chips with hundreds of cores to assist servers and supercomputers with large numbers of relatively repetitive calculations, said Steve Smith, vice president of the desktop platforms group at Intel.
The focus on multiple cores arises from Moore’s Law, which dictates that the number of transistors on a chip doubles every two years. In the past, the extra transistors have been used to increase the size of the cache, an integrated pool of memory for rapid data access, or to boost other performance-enhancing features, such as instruction-level parallelism, which lets a chip do more than one task per clock cycle.
However, Intel is now employing the extra transistors to create additional cores to boost performance, a measure it says it can take without drastically increasing chips’ power consumption. Power consumption was among the reasons that Intel shifted to the dual-core strategy and canceled Tejas, a high-speed, single-core successor to the Pentium 4. Continuing the trend of quickly upping its processors’ clock speeds would have eventually boosted their power consumption above practical levels, Intel has said.
“We’re doing (dual-core chips) because we believe it’s a more efficient way to use our transistor budget,” Smith said.
One of the key areas where consumers may notice the dual-core difference will likely come when they run two or more applications, Smith said. With two cores, a consumer can edit a movie from a digital video camera in the background or record a television show using a personal video recorder setup–while surfing the Web, sending e-mail and performing other tasks.
Intel isn’t the only chipmaker expecting twins. Nearly every processor company has multiple-core plans. IBM has been selling dual-core chips for servers for a few years, while ARM sells dual-core chips for cell phones. Intel rival Advanced Micro Devices has said it will create chips with two, four and eight cores, too. Its first dual-core chip will also come out in 2005.
Meanwhile, ClearSpeed has developed a 96-core chip for augmenting supercomputers, similar to chips from Japan’s RIKEN and others. Nonetheless, Intel is the world’s largest chipmaker, so the vast majority of the population will experience dual-core computing through its products.
“This will be looked back upon in 10 years as the beginning of the new processor architectures,” said Dean McCarron, principal analyst at Mercury Research. “The manufacturing challenge is not all that great. It is easier to do this than to do a complex single processor that’s twice the size.”
As some companies start making dual-core chips by merging two of their single cores into one piece of silicon, they will begin to differentiate their products by adding different types of input-output mechanisms and other features, McCarron added.
The first dual-core Intel chip for desktops, code-named Smithfield, will come out in desktops in 2005. Its first dual-core notebook chip, code-named Yonah, will begin shipping late in 2005 and likely won’t appear in notebooks in any great numbers until 2006. A dual-core version of Itanium, called Montecito, will also debut in 2005 while Tulsa, a two-headed Xeon for servers, is set to come out in the first quarter of 2006.
Smithfield will also be paired with a feature called Active Management Technology, which will allow an IT manager to control a PC remotely. If a PC is spitting out viruses, for example, an IT manager can shut it off with the management technology. “If you are going to curtail an outbreak, you want to contain it in a matter of minutes,” said Justin Rattner, an Intel fellow.
While volumes of dual-core chips will likely be small in 2005, they are expected to represent 70 percent of notebook and desktop of the chips produced by Intel by the end of 2006 and 85 percent of the server chips.
Because of the additional core, Smithfield will initially be larger than its single-core predecessors and hence more expensive to manufacture, Smith said, but the chip will decrease in size over time. This pattern is typical in processors.
Smithfield will be made on the current 90-nanometer manufacturing process, which means that the average size of features on the chip will measure 90 nanometers. Yonah will be made on the 65-nanometer process coming next year. A nanometer is a billionth of a meter.
Smith also acknowledged that Intel’s desktop and notebook chips, now quite distinct architecturally, will begin to converge.
“Over time, these will look increasingly similar from a feature and capability perceptive–more so than they do now,” Smith said. Among other things, “it makes for consistent software development.”
Software compatibility will not be a problem, Smith said. Microsoft’s Windows XP is already designed to run on two processors. Many desktop applications have also already been tweaked to take advantage of hyperthreading, a technology in current Pentium 4s that lets a single chip do two actions at once, almost like a dual-core chip.
Hyperthreading, in fact, will make dual-core chips seem like four-core chips. Servers will be able to take advantage of this feature, and hyperthreading will be a feature of dual-core Xeons and Itaniums. By contrast, hyperthreading may not be manifest in the early dual-core notebook and desktop chips, other Intel execs have said, because client applications haven’t been tweaked to run on four processors yet.
Still, while software development may be fairly straightforward, software licensing for dual-core and multicore processors has proven trickier, especially in servers. Despite Intel and AMD recommending that companies charge fees by the number of processors each server contains, regardless of how many cores each processor includes, some companies’ policies dictate charging by the number of cores each processor in a given server contains.
CNET News.com’s John Spooner contributed to this report.