Index Home About Blog
From: (MitchAlsup)
Newsgroups: comp.arch
Subject: Re: RAM vs ROM speeds.
Date: 1 Sep 1998 17:10:04 GMT

>I don't believe that I've ever seen a description of why
> ROM is slower than RAM.

ROM is NOT slower than RAM. ROM chips are slower
then SRAM chips (see below).

A ROM cell is a simgle transistor coupled to either
VDD or VSS on one side and a bit line on the other
side. An SRAM cell is (typically) 6 transistors and
3X as wide in the word line direction and  X as
large in the bit line direction. For the same number
of words and bits/word, a ROM is considerably smaller
than its SRAM counterpart. Since wire delay has become
a significant part of the overall logic delay of a circuit,
a ROM is both smaller and faster than an SRAM.

> I've always assumed that it was because RAM chips tend to
>be used in parallel, while ROM chips don't.

ROM chips tend to be used for boot processing and for
noncritical (timing wise) stuff. Therefore, the designer
is attempting to fit as much storage onto the chip while
simultaneously reducing power consumed by the chip.
In many systems, the ROM is seldom used, and must
consume as little power when not in use as is possible.

>Is it the case that ROM logic is simply slower than RAM,
>or is it that transfers from ROM are slower than transfers
>from RAM?

ROM chips can be made as fast as the fastest SRAM
chips if the consumer is willing to pay the power to
operate them at these speeds. Since there is little
demand for ultrafast ROM these days, you will find it
difficult to find ultrafast ROM.

The differences in speed between the ROM and the
DRAMs and the SRAMs are due to the markets these
differing designs are targeting, not an inherent problem
with the cell which maintains the stored data.

Mitch Alsup

Index Home About Blog