This is the one year anniversary of my column. Iíd like to thank all
those who supported me this last year. Several have said youíve enjoyed
one article or another. I appreciate that. Hopefully, Iíll continue to
provide enjoyable articles over the coming New Year. Merry Christmas!
I suppose it was inevitable. Since outlining the differences between
CISC and RISC, several asked about the ongoing battle for supremacy between
the big two of mainstream x86-based processors: AMD (Advanced
Devices) and Intel.
To gain perspective, I think we should take a brief look at history and
then discuss how these two companies differentiate their products today.
In the early days, Intel and AMD maintained a technology licensing agreement.
Basically, this meant that when Intel developed new processor technology,
AMD was entitled to use it. To this end, both companies produced extremely
similar products, with AMD focusing on producing cheaper offerings than
Intel. Feeling the pinch of competition, Intel ended the licensing agreement
with the Socket 7 line of processors (the original Intel Pentium and the
AMD K6). With these CPUs, the two companies diverged.
Early differentiation was in the core architecture of each brand, with
each developing distinct internal command sets. It began with MMX
Intel and AMDís 3D-Now! Each continued to extend these commands further,
basing much upon SIMD (Single
commands which accelerate system performance in graphic intensive environments.
As mentioned in October, these technologies have gradually produced large
differences in the processors, much in the way core differences make RISC
and CISC architectures dissimilar. AMDís current focus is to process more
commands per clock cycle, while Intelís is on producing chips with the
highest internal clock speeds.
Closely tied to the core architecture is the physical construction of
the processors. After the initial divergence, Intel quickly moved from
socket to slot based processors (Slot 1), claiming the larger package provided
for expanded features (like room for more on-chip memory), while being
easier to cool and upgrade. AMD followed suit, releasing their own version
(Slot A), the first product designed after the end of the licensing agreement.
||Itís rumored that Intel initiated Slot 1 merely
as a means to change processor technology from the co-licensed Socket 7,
not for design reasons. This conspiracy theory may be supported by the
fact that both companies subsequently returned to socket designs.
Each company is now refining the production of their chips, moving from
the 0.18 to the 0.13 micron process. The smaller number correlates with
manufacturing smaller chips, which run more efficiently and can simultaneously
contain more transistors ó translating directly into more processor power.
Meanwhile, decreasing the size allows more space for increased amounts
of L2-Cache on the processors. When combined with new motherboard chipsets,
these processors support higher bandwidth FSB (Front
Bus) speeds with
faster external memory support through RDRAM (Rambus) and DDR (Double-Data
At the beginning of the year, Intel introduced its Northwood CPU core,
at 0.13 micron with 512KB of L2-Cache on a 100MHz FSB supporting 400MHz
RDRAM. The Pentium 4 2.2GHz, based on the Northwood core, outperformed
AMDís Athlon 2000+, which was actually a 1.6GHz processor based on AMDís
Palomino core, at 0.18 micron with 256KB of L2-Cache on a 133MHz FSB supporting
266MHz DDR RAM. By March, AMD released the 2100+, which proved to outshine
the P4 2.2GHz. Intel answered in May, taking another leap by moving to
the 133MHz FSB and 533MHz RDRAM, with the P4 2.533GHz. This processor beat
the 2100+ in all tests. In June, AMD made its move from 0.18 to 0.13 with
its Thoroughbred core and was able to reach the same performance marks
as higher-clocked P4ís, demonstrating that clock speed was not the only
measure of performance. However, these changes failed to attain the overall
performance of Intel. So, in August, AMD made adjustments to their CPU
core again with the Athlon XP 2600+, allowing them to surpass Intel. But
Intel quickly countered with its P4 2.8GHz, which bested AMD.
So, whatís my opinion of the two? That competition is healthy and the
battle is far from over. Projections suggest the Athlon XP 2800+ can regain
the performance lead. The FSB will increase from 133 MHz to 166 MHz, fully
supporting DDR 333MHz memory. This brings performance up to the level of
a top P4 model with 533MHz RDRAM (named PC1066). AMD will also be increasing
their level of L2-Cache to 512KB. Unfortunately, AMD won't be able to deliver
these CPUs until next year ó so Intel should retain the performance title
a little longer.
Based on the maturity of each processor, and increasingly stable motherboards,
I never hesitate to recommend either to customers. With performance so
close, itís more often a matter of brand loyalty than anything else. Though
I sometimes have a hard time listening to my own advice, I generally find
the best value when buying two levels below the latest offering (whichever
brand you choose). This means purchasing a 1.7GHz after the release of
the 2.0GHz model may give you the most performance for your dollar. Raw
performance comes at a premium.
Also, with designs changing every day, purchasing top-of-the line is
a quixotic endeavor. I suggest setting a budget and buying the beefiest
overall system within that budget. Put the motherboard, memory, and other
desired components together first. Then check the compatibility of these
components, purchasing the processor that fits your system. If itís Intelógreat!
Otherwise, AMD is always a viable alternative.
||To combat obsolescence, I occasionally recommend
dual-processor capable motherboards, but only purchasing a single processor.
Windows 2000 and XP support multiple processors, and you can give
your system a boost in the future when processor prices drop.