Intel Needs More Time for Programmable Larrabee Graphics Project

Intel has decided to abandon plans for the 2010 release of a programmable graphics architecture code-named “Larrabee.” If released as Intel had originally planned, Larrabee would have served as a discrete graphics processing unit (GPU) that could be programmed like Intel’s x86 microprocessor.

Intel had previously promoted Larrabee as greatly increasing the flexibility and programmability offered by standard GPUs through the use of parallel processing. In particular, Intel said Larrabee held the promise of revolutionizing video game development.

However, last week, Intel released a statement saying Larrabee silicon and software development are “behind where we hoped to be at this point in the project.” As a result, Intel will launch Larrabee as a software development vehicle, including a throughput computing development platform.

Intel’s retreat from offering Larrabee as a commercial product is a big bonus for competing GPU providers AMD and NVIDIA. Neither AMD nor NVIDIA currently offers a GPU with the functionality Larrabee was supposed to possess, but they now have some breathing room to try to develop one.

Intel should benefit from the time, effort and money it has sunk into Larrabee by keeping it alive as a development platform that may aid advancements in parallel processing. The technology landscape is littered with the carcasses of “new and improved” technology products that were rushed to market despite known bugs, glitches and shortcomings in order to stay ahead of the competition.

The ruthlessly Darwinian IT market tends to discover and weed out new advances in technology that aren’t quite ready to be taken out of the oven. Intel is smart to avoid rushing Larrabee to the market as a commercial product if its quality and performance are not up to par. Sometimes not releasing the wrong product is as good as releasing the right product.