Researchers have built the prototype to study how best to make that many cores communicate with each other. They're also studying new designs for cores and new architectural techniques, according to Manny Vara, a technology strategist with Intel's R&D labs. The chip is just for research purposes and lacks some necessary functionality at this point, but Vara says Intel will be able to produce a chip with 80 cores in five to eight years.
The chip is being called the Tera-Scale Teraflop Prototype. Intel is planning on releasing specifics about the research project at the 2007 International Solid State Circuits Conference in early February.
Since dual-core and quad-core chips were just introduced to the market in the past year, looking forward to an 80-core chip is a major departure from the expected natural progression in microprocessors.
"We're doing this as a test," Vara told InformationWeek. "Our scientists came up with all these different ideas and we needed to test them on a piece of silicon. We don't know if it works 'til we build one so we built it and then tested the heck out of it."
Vara says the 80-core chip uses less than 100 watts of energy; a dual-core chip uses 60 to 70 watts and a quad-core uses 105 to 130 watts. Of course the numbers for the 80-core chip could be affected by the fact that it's lacking some functionality, but Rob Enderle, president and principal analyst of the Enderle Group, says it's still a significant accomplishment.
Enderle calls Intel's research project "revolutionary."
"It showcases a focus on power efficiency and keeping things within a power and thermal envelope," he adds. "They recognize that they can't generate any more heat or pull more power or they'll just become power hogs. With that large a number of cores, it's significant they can get it down below 100 watts. So this is pretty impressive. Real impressive."
They're different kinds of cores, explains Vara, who adds that energy efficiency is a major part of the research project. "If you look at it, by the time you put dozens of cores on a chip, they won't be the same kind that you can put three or four on a chip today. The new ones will be much simpler. You break the core's tasks into pieces and each task can be assigned to a core. Even if the cores are simpler and slower, you have a lot more of them so you have more performance."
Vara notes that the power efficiency lies in the new, simpler cores.
"Think more-complex four cores compared to simpler 80 cores. Each of those four cores can do more individually than one of the 80," he explains. But with an 8-core chip "you will get a lot more performance and lower power because you have a lot of them running at lower speed. You're only using the cores you need. It's performance on demand. If you need more performance, it wakes up more cores, and when you're done, they go back to sleep."
With that many cores, Intel is able to design what Vara calls "core hopping." If one part of the chip gets hot, the work that those particular cores are doing is moved to other cores on another part of the chip. That, he explains, will lower the heat being generated.
"That's definitely exciting," says Dan Olds, a principal analyst of the Gabriel Consulting Group. "They're pushing the boundaries out in terms of how many cores they can squeeze onto a chip. It's about the horizontal scalability of cores on a single chip. They're relatively dumb and unsophisticated cores, but the experiment is how can they do it. How many can they pack in and still work together and communicate with each other?"
Part of the challenge with building a chip with so many cores has been to design a communication network so that they all can communicate with each other.
Handling that much network traffic has been quite a task, says Vara, who adds that the chip will be just slightly larger than the average chip today. "What we're doing is designing a network inside the chip. Today, you hear about high-performance computing and they have these big, fat super-powerful servers and they're all networked together. We're trying to basically do that, but on a chip. How do you bring a real network inside a chip so all the cores can talk to each other?"
"We're quite literally creating a network mesh to let each little core communicate with the other cores and the rest of the system," says Vara. "The cores will want to know what the other cores will doing so they don't fight."
While it may take five to eight years to come out with a working 80-core chip, Vara says IT managers might start watching for what he calls "different flavors" of quad-core chips. "Maybe you'll have interim chips where they have more complex cores along with simpler cores, too."
