For all the ballyhoo about supercomputers and their feats of CPU derring-do, the actual cabinet space necessary to put them together tends to be ginormous—row upon row of racked out silicon toiling in tandem like something out of the mainframe-like room at The Matrix‘s “Source.” Some of these setups make even WarGames‘ WOPR (remember that?) look like a toy box.
Intel may be about to change all that, with a processor built on an architecture it’s dubbed “Many Integrated Core” or MIC.
But first let’s talk FLOPS, or floating point operations per second, which signify a popular measure of supercomputer’s processing prowess. We know from the latest benchmarks that the world’s fastest computer, Fujtisu Co.’s K supercomputer, can do 10.51 petaflops, which amounts to over 10 quadrillion calculations per second.
(MORE: And Today’s ‘Fastest Computer in the World’ Award Goes to…)
In scientific notation, we’d call 1 petaflop equivalent to 1 x 10 to the 15th power floating point operations per second. So over 10 of those add up to what the K supercomputer can do. But it takes the K supercomputer 864 computer racks and 88,128 parallel-linked, eight-core CPUs to pull that off. That’s a whole lot of CPUs.
Enter Intel’s “Knights Corner,” a commercial co-processor based on the company’s MIC architecture and capable of 1 TFLOPS (teraflops) double precision performance. Granted, that’s running software that may have been optimized for Intel microprocessors, notes the company, but okay, 1 teraflops, equivalent to 1 x 10 to the power of 12, or 1 trillion calculations per second.
By comparison, a single Intel i7 processor with six cores can do about 158 GFLOPS or 158 billion calculations per second. Knights Corner—sounds like a chess move, right?—is thus over six times faster than some of the fastest consumer-grade processors available right now.
Intel says it represents “the first demonstration of a single processing chip capable of achieving such a performance level.” Contrast with Intel’s demonstration of a single TFLOPS computer back in 1997 using 9,860 Pentium Pro processors. Remember the Pentium Pro? I do. That’s about the time I started hand-building my home computers, shortly after my first Pentium (and last package-bought desktop), a Dell Dimension XPS 90 with a 90MHz second-generation Pentium CPU.
Knights Corner will be built on Intel’s three-dimensional tri-gate tech, which runs on microscopically small 22 nanometer silicon (or one-billionth of a meter). That’s an astonishing 10 nanometers smaller than the company’s current 32 nm “Sandy Bridge” chip family.
Intel describes the Knights Corner co-processor as “very unique” because “unlike traditional accelerators, it is fully accessible and programmable,” which means it’ll be visible to applications “as though it was a computer that runs its own Linux-based operating system independent of the host OS.”
Another upside of the MIC architecture, claims Intel, is that it can run existing apps without ports, which means—yep, it’ll work with existing x86 applications.
When? Intel’s not saying, nor is it hinting how much one might cost. Whatever the case, chances are it’ll be a heck of a lot cheaper than the $55 million Computerworld says the U.S. government paid for equivalent computing speed 15 years ago.
MORE: Intel Trades Over 50 Years of Chip Design for ’3D’ Processors
Matt Peckham is a reporter at TIME. Find him on Twitter at @mattpeckham or on Facebook. You can also continue the discussion on TIME’s Facebook page and on Twitter at @TIME.