|
February 01 2005: Welcome Back ! 2004 was definitely a roller coaster ride in regards to Computer Technology. We covered quite a lot of ground from the peaking of the existing technologies which led to the realization that the quest for performance in megahertz alone was no longer a viable endeavor, to the failed promises from both Intel and Apple to deliver over hyped milestones, to the slow uptake of emerging technologies such as PCIe, DDR2 and the expected 64 bit wave. Back in April I covered the “End of the Megahertz
Myth” when in an unprecedented move, Intel finally conceded that
performance could no longer be judged by megahertz alone, and in doing
so, shifted their focus away from marketing their CPU’s on clock
speed , and introduced new model numbering system that effectively eliminated
raw clock speeds from its product names. Apple also had their share of problems with the transition to .09 Micron, with the IBM built 970 also suffering major heat dissipation problems, resulting in the scalability basically stopping dead from its initial launch. They did manage to squeeze a new Dual 2.5 System onto the market, ( which is nothing more than an overclocked 2.0 chip ). To keep the thermals in check, they needed to take the extreme step of water cooling the chips, but in their usual brilliant marketing spin, they managed to use the Jedi Mind Trick to convince the punters that it was actually a cool new innovation, instead of an extreme measure to compensate for the overclock. :-) It wasn’t surprising that the much over hyped 3.0 GHz chip that Steve Jobs proudly boasted would eventuate, has mysteriously disappeared into the ethers. AMD had a successful but rather laid back approach to
the whole year. Their Athlon 64/Opteron chips made some major inroads
into numerous markets traditionally dominated by Intel, they also scored
major points by forcing Intel to adopt their AMD64 extensions, but didn’t
really consolidate the momentum that they had garnered probably due
to the unfortunate fact Microsoft failed to deliver XP/2003 64 bit O.S
that would have helped them cement their advantage of bringing X86-64
to the market first. Instead by the time the O.S arrives in Q2 2005,
Intel will have multiple 64 bit capable chips on board to combat the
possible exodus . Their transition to .09 Micron has apparently been
the smoothest of all, although actual product is in very short supply,
so perhaps the reported smoothness of the transition The big news that bubbled to the surface later in the
year was the transition to Dual Core Chips that was being fast tracked
by all of the majors. I covered that emerging landscape extensively
in June / September 2004. Just to recap. |
As it became increasingly obvious that
the megahertz madness of recent times had reached a dead end, all
of the chip giants had to find another way to continue ramping future
CPU performance, without relying on faster clock speeds. There are
many of ways of improving performance of central processing units
including larger Caches, FSB, adding media/vector processing extensions
such as Altivec and SSE, branch and memory pre-fetch, out of order
execution mechanisms, etc. All of these techniques have proven to
be of great value in improving performance, but they have always been
coupled with increased clock frequencies to deliver the end results.
Another way of improving performance is to enable the execution of
multiple threads at the same time. Intel’s first implementation
was Hyperthreading, which virtually splits the CPU into 2 cores. Although
not as powerful as true Multi processor systems, it did prove an advantage
in CPU intensive applications, and hinted at what was to come. The
next logical step is to implement dual CPU cores on a single die
|
