AMD will not likely be able to match Intel's scale of manufacturing & production anytime soon, so they're taking the other direction: make the chips "smarter." Expanding SSE will hopefully accomplish that goal.

"AMD designers examined existing instruction sets, such as the Apple/IBM/Freescale Altivec instruction set, then identified three software types to optimize SSE5 around: compute-intensive applications, such as financial simulation and life sciences; multimedia applications, like high-definition video encoding and image processing; and security applications, where data is encrypted across the Internet as well as an entire hard drive."

According to the Inq, Intel's Harpertown Penryns & Seaburg chipset will easily break the 100GFLOP barrier. What's even more interesting, though, is that the desktop chips will be unlocked versions of the same SKUs, and use the same chipset, but designed for lower latency, & toting the "Skulltrail" moniker. Frightening news for AMD....

"After all, an extra 20 - 30% speed boost on CPU, and additional 30+ memory throughput due to optimised FSB latency coupled with both higher bandwidth and lower latency by those upcoming FB-DIMMs, should give a very decent application performance increase, not to mention ability to feed, say, four Nvidia G92 GPUs simultaneously?"