AMD has just announced a 64-bit CPU, code named "Hammer." That shouldn't be startling. Intel announced its 64-bit Itanium long ago, SunSystems has its UltraSparc, and HP/Compaq has an Alpha. We are awash in a virtual ocean of 64-bits. AMD needs to do better than just matching the numbers and it thinks it has.
"The Hammer architecture innovations include a high-bandwidth, high-performance integrated memory, an input/output and multiprocessing controller, and a highly scalable system bus that uses HyperTransport technology with support for single- and multi-processor configurations," said Fred Weber, vice president and chief technical officer of AMD's Computation Products Group. He also announced that the Hammer runs not only 64-bit code but also 32-bit code as well--and does so without a performance penalty.
According to Weber, "AMD's approach to 64-bit computing puts the IT customer first. It enables IT managers to take advantage of existing support, allowing them to upgrade to 64-bit software at the appropriate time and preserve their investment in 32-bit applications." Make no mistake; AMD is playing directly to the heartstrings of IT TCO concerns, and in a period when IT is seeing a downturn in growth because of a slowing economy, it's a song everyone will be listening to. While AMD's up-and-running posture suggests that Hammer will be a chip for all purposes, the majority of Itanium-equipped systems--as the Gartner Group proposes--may be specialized workstations, proposes that "because the first Itanium is a better basis for a development as opposed to a deployment platform." Basically, there are no applications for Itanium. They will need to be specifically developed for that CPU's instruction set.
Is the Hammer a sure-shot for AMD? The real answer is not in the least. Getting the hardware side up and running will probably be the easiest of the things it needs to do. VIA Technologies recently announced the creation of its VIA Platform Solutions Division (VPSD) to produce motherboards using the VIA chipset solutions. AMD and VIA have long been conjoined at the chipset, and building a motherboard that will give the Hammer the 4- and 8-processor penetration AMD claims as its goal is probably already a fait accompli on the drawing board. The first hurdle it faces is to show that HyperTransport is a better option than Intel's 3GIO bus architecture and that it has legs enough to either compete or coexist with 3GIO when 3GIO is finally released.
A second and more important speed bump in AMD's fast track for the Hammer will be to prove its 64-bit competence. Running 32-bit software should be a given as much as the Hammer is, to a certain extent, a derivative of the company's current Athlon processor. Running 64-bit code as well as, if not better than, Intel's Itanium with applications written specifically for the Itanium's instruction set is another matter entirely.
Assuming AMD can accomplish both of these goals, and do so reliably, the payoff to IT is a huge postponement of what will certainly otherwise become a forced march to 64-bit applications littered with the unamortized cost of 32-bit software in its wake. It will also give IT managers the ability to consider competitive 64-bit applications in parallel environments with their 32-bit counterparts without needing to maintain two different hardware platforms to do so. Perhaps most importantly, it will allow for incremental hardware upgrades rather than an abrupt changeover that's guaranteed to spew red ink during a time when money is tight.
What do you think of AMD's Hammer chip? E-mail Bill or post your thoughts in our Talkback forum below.