Rambus DRAM
is a dynamic RAM chip technology from Rambus of Mountain View, CA, USA (
href="http://www.rambus.com">www. rambus.com). It transfers data up to 10 times faster
than conventional DRAMs–at 600 MB/sec. It requires modified motherboards, but
eliminates the need for static-RAM cache. Rambus licenses its memory designs to other
semiconductor companies, which manufacture the chips. Intel and Rambus are expected to
come up with a faster version of Rambus, to be named nDRAM (Next generation DRAM), for use
with PIII/500 or faster chips toward end-1999 or next year.
DRAM is Dynamic RAM, and it remains the only
type used as main memory in a PC. It has, however, engendered many sub-species: FPM, EDO,
SDRAM, et al. All of them continue to be drastically slower than static RAM (SRAM).
But SRAM is far more expensive. Rs 2,000 would buy you 32 MB of DRAM, but only about 256
kB of SRAM.
Yet DRAM is far too slow for CPUs today. Forget the 500 MHz Pentium
III; it’s too slow even for a 66 MHz Pentium from 5 years ago.
That’s why CPU makers use SRAM "cache"–16 to 64
kB of SRAM usually embedded inside the CPU chip, which interacts with the processor much
faster than the DRAM would. It "caches" data and instructions, freeing up the
CPU to get on with its work, while it talks to the much slower DRAM in the main memory.
This is level 1 cache. Some CPU or PC board designs go further and add a further 128 to
512 kB of cache outside the chip, either on the same daughterboard (as in Slot 1 chips
like the Pentium II and III) or on the motherboard, in the case of Socket 7 chips like the
Pentium MMX or the
Cyrix MII.
alt="SRAM is fast and expensive, and DRAM is slow and cheap, so some manufacturers have included a small amount of SRAM directly into the DRAM chip, creating an on-chip cache"
align="right">But the RAM itself remains slow–a problem when, for
instance, a fast video card has to interact with main memory. In some systems such as the
Cyrix MediaGX, a part of main memory itself is allocated for video, making its speed even
more critical.
To get around the slowness of standard DRAM memory modules, some
manufacturers are beginning to include a small amount of SRAM directly into the DRAM chip,
effectively creating an on-chip cache. One such design is Ramtron International’s
ESDRAM.
ESDRAM is essentially SDRAM, plus a small amount of SRAM cache. This
allows for lower latency times and burst operations up to 200 MHz. Just as with external
cache memory, the DRAM’s cache is designed to hold the most frequently used data in
the SRAM cache, to minimize accesses to the slower DRAM.
One advantage to the on-chip SRAM is that a wider bus can be used
between the SRAM and DRAM. This effectively increases the bandwidth and speed of the DRAM
even when there is a cache miss (when some data being sought isn’t found in the
cache, and has to be sought in the slower DRAM).
There is at least one Socket 7 chipset around with support for
ESDRAM. The deciding factor in deciding if this will succeed will probably be the initial
cost of the modules. At present, ESDRAM costs over three times more than existing DRAM
modules, which is way too expensive to get truly mainstream.
COLOR="#ff0000" size="2">
DDR SDRAM size="2">
In the SDRAM used today (JEDEC
SDRAM),
there’s a design limitation of 125 to 133 MHz bus speeds. Clearly, bus speeds will
need to increase well beyond that so that memory bandwidth can keep up with future
processors. Emerging standards that look promising require other changes: special
pin-outs, smaller bus widths, and so on. For now, "Double Data Rate" SDRAM looks
very appealing. This design allows the activation of output operations on the chip to
occur on both the rising and falling edge of the clock; in conventional
SDRAM, only the
rising edge signals an event to occur. Thus DDR can effectively
double the speed of operation–up to 200 MHz.
There is already one Socket 7 chipset that has support for DDR
SDRAM, and more will follow if manufacturers decide to make this memory available.
COLOR="#ff0000">
Goodbye, EDO
EDO (Extended DataOut) RAM is fading out, but it’s still
supported by the largest number of PCs around, so it’s worth mentioning here.
Also called Hyperpage mode, EDO added 40 percent or greater
improvement in access time to the older DRAM, with the same amount of silicon and the same
package size. EDO works well with memory bus speeds up to 83 MHz. If the chips are 55 ns
or faster, EDO can be used even with a 100 MHz memory bus. One of the best reasons to use
EDO is that all of the current motherboard chipsets support it with no compatibility
problems, unlike some of the SDRAM modules now being used.
However, EDO is difficult to get (it’s now more expensive than
SDRAM) because of limited production.
Your PC’s motherboard probably has room for both EDO and SDRAM
(check out the boot-up information, which should then list both types). If you already own
EDO memory, there is no real reason to upgrade to SDRAM unless you require bus speeds
above 83 MHz. With a typical EDO timing of 5-2-2-2 at 66 MHz, there is almost no
noticeable improvement with SDRAM over EDO, and at 83 MHz it’s still negligible. If
you require 100 MHz bus operation, EDO will lag far behind current SDRAM in performance.
With information from
href="http://intel.com">intel.com, www.tomshardware.com,
and other sources on the Internet