two SIMMs were required for 16-bit
memory access (supported by the 386)
and four for 32-bit access (supported
by the 486).
INTRODUCTION OF
THE CPU MULTIPLIER
In 1992, INTEL released the 486DX2
processor which introduced the
concept of the CPU multiplier. Up until
then, the CPU ran at exactly the same
speed as the system bus, whereas
the 2x multiplier seen on the 486DX2
allowed the CPU to operate at twice
the speed.
The 486DX2 was also the first time
that INTEL started binning CPUs with
chips failing internal tests being sold
as the 486SX, which had the FPU
included but disabled. This practice
was later applied to CPU frequencies
as well, where a processor failing
testing using a 2x multiplier might
be sold as one using a 1x multiplier.
Worth noting is that there was a
24 The OverClocker Issue 24 | 2013
80487 co-processor which could be
bought separately for systems with
the SX chip so as to add the missing
FPU functionality. This was a fully
functioning 486DX chip which actually
disabled the 486SX CPU and took over
all CPU functions.
This really opened up the gates for
overclockers, as while some failed
chips could only manage a slight
increase over their stock speed,
others could get to within a few MHz of
the next speed grade or even more.
As CPU voltages could not be
selected by the system automatically,
the user would have to dial in the
voltage manually using either DIP
switches or jumpers. A higher than
stock voltage often meant that the
processor would be capable of
frequencies in excess of models
which were two or more speed
grades above. This also introduced
the concept of “clock ceilings”, where
between the manufacturer and
overclockers the speed limits of a
process and architecture could be
found.
By 1993, INTEL had released the
Pentium processor using the P5
architecture, but with an FPU bug,
these CPUs suffered from limited
widespread adoption. These original
Pentium processors had an FSB of
either 60MHz or 66MHz, and a 1x
multiplier.
With successively higher bus
speeds, higher frequency RAM
became a reality, while the new 64-bit
memory controller meant a wider bus
would be needed for the RAM. The old
30-pin, 8-bit SIMMs were replaced
by 72-pin, 32-bit EDO (Extended
Data Out) SIMMS. As the Pentium
supported 64-bit memory access, EDO
SIMMs required installation in pairs.
EDO RAM had another advantage,
however. When starting a new cycle,
it could keep the data path open from
the previous memory access, bringing
timings down to 5-2-2-2.
The following year, the company
released IntelDX4 (sometimes
referred to as the 486DX4/100) which
had a 3x multiplier. This allowed for
CPUs to operate at up to 100MHz by
using a 33MHz FSB and 3x multiplier,
although some versions used a 25MHz
FSB and 3x multiplier. We had finally
arrived at the era of CPU multipliers.
As useful as multipliers had proved,
there was one large downside (at
least from a CPU manufacturer’s
perspective). The 486 SX25 for
example could often operate with
perfect stability simply by moving
a jumper or flicking a DIP switch to
increase the system bus to 33MHz.
It was so easy and efficient that
several retailers caught on, leading
to fraudulent sales and the first
overclocking scam. INTEL's first
attempt to combat this was only
several years later with a locked 2x
multiplier on the Pentium 133 (S-Spec
SY022). The multiplier lock was seen
again on most Pentium MMX chips,
and with the introduction of the
Pentium II it became the norm; a lock
which would stay in place until the
launch of the Pentium Extreme Edition
in late 2003.