AMD and Intel Won’t Let x86 Die Without a Fight

Intel's Knights Ferry development card

AMD at the Computex show in Taiwan today showed off a new processor dubbed the Fusion APU (accelerated processing unit), which combines traditional CPUs with graphics processors. While the critics were underwhelmed by the demo, AMD (s amd) is following its rival Intel (s intc) down a path to keep x86 processors both powerful and power-efficient as computing goes mobile on the consumer side and requires millions of processor cores running in a “cloud” on the server side. The design tweaks both chipmakers are implementing are key to ensuring that their core x86 business doesn’t become sidelined by new competition from Nvidia (s nvda) and ARM (s armh).

On the server side, Intel and AMD are attempting to make their chips more efficient while boosting the ability to do highly parallel processing, used for transactions, video encoding and crunching certain types of data. It’s an area where Nvidia has made huge inroads with its GPUs and as such, something Intel and AMD, which has its own GPU business, couldn’t ignore. Intel has tried to create a GPU-like chip using an x86 architecture dubbed Larabee, but pulled the plug on it.

Intel said earlier this week that it would deliver new products based on its Many Integrated Core (MIC) architecture it helped develop for Larabee that will scale to more than 50 cores on a single chip. The product, while aimed at the high-performance computing space (GigaOM Pro, sub req’d), will also relate to Intel’s cloud efforts. Certain clouds may require specific HPC-level performance such as those used for delivering and transcoding media.

The Aubrey Isle chip inside the Knight's Ferry development platform

The first product, codenamed “Knights Corner,” will be made using Intel’s 22-nanometer manufacturing process and will scale to more than 50 Intel processing cores (pictured) on a single chip. AMD plans to launch an APU for the server side as well. I’ll ask Rick Bergman, senior VP and general manager of AMD’s Products Group, about it later this month at our Structure 2010 conference in San Francisco.

For servers that don’t benefit from parallel processing, such as those running websites, David Perlmutter, executive VP and co-general manager of Intel’s architecture group, showed off low-power blade servers running Atom chips at Computex. We’ve outlined how a host of companies ranging from Marvell (S mrvl) to SmoothStone are experimenting with the ARM architecture used in cell phones to build lower-power servers, which poses a threat to Intel and AMD’s core business.

On the consumer side, Intel is also facing competition from ARM in smaller form factors such as netbooks, tablets and smartphones, a market the AMD has pretty much ceded while it focuses on laptops and PCs. In this market, AMD hopes its APU will wow users with awesome multimedia performance at lower power consumptions, while still using software designed for CPUs. Intel is relying on Atom on the low end and its fancy Corei5 and Corei7 processors for performance-hungry mobile and desktop machines.

As the computing pie has gotten bigger — with more variety in client devices and millions of servers delivering web-based services — Intel and AMD, which have long owned the largest slices, are now realizing that others are hungry, too. To keep their chips and the x86 architecture in demand, they’re trying to take on the characteristics of their rivals in the mobile and GPU sectors. We’ll see if they can adapt.