TCS - Computer Memory -- Part 2

Computer Memory -- Part 2

by Microprocessors Unlimited
(918) 267-4961
From the February 1997 issue of the I/O Port Newsletter

Cache Memory is a small, special, extremely high speed memory, local to the CPU, designed to accelerate processing of program instructions by the CPU. It is functionally sandwiched between the CPU and main memory. In effect, it is a high speed buffer, intended to have program instructions available for the CPU to execute quickly, without delay.

Cache memory is analogous to an efficient worker who has all of his most frequently used tools in a tool belt, ready to be used quickly. Main memory is analogous to a large portable tool box. A hard disk is analogous to a workshop.

Cache memory is often 3x to 5x faster than main memory. Because of this high speed requirement, cache memory is almost always static RAM or SRAM. Although static RAM is faster than dynamic RAM, it is more expensive. SRAM employs six transistors per bit of memory, whereas DRAM requires only one transistor per bit of memory. A static RAM bit is in essence a flip-flop circuit made with six transistors, very fast and consumes lots of power. A dynamic RAM bit is only one transistor with a small capacitor, slower, but consuming less power. Main memory, which is larger, slower, and less expensive is normally composed of various types of dynamic RAM or DRAM.

There are two types of cache memory: Level I Cache and Secondary Cache. Level I Cache or Primary Cache is internal to the CPU. Secondary Cache is external to the CPU, on the motherboard. Some earlier model PCs have CPUs which do not contain internal cache, instead they have Level I Cache installed on the motherboard very close to the CPU.

TAG RAM is the part of Cache which keeps track of memory locations being used in Cache memory, to help manage operation of the Cache.

PIPELINE BURST CACHE is a special Cache in a module which cuts the processing time of memory operations, thus maximizing the system execution rates. PIPELINE BURST CACHE works by fetching the next instruction while the current instruction is being executed, to minimize delays. Compare this to an oil pipeline being filled at one end while oil is being removed at the other end of the pipeline.

EDO RAM is a form of DRAM technology, which shortens the read cycle between main memory and the CPU. It allows a CPU to access 10% to 15% faster than comparable standard DRAM memory, or FAST PAGE MODE memory. EDO means Extended Data Out. Typical access times for EDO RAM is 60 ns.

SDRAM or Synchronous DRAM is faster than EDO RAM! Typical access times for SDRAMs are 7 ns to 10 ns! They are fast! Perhaps more importantly they are called synchronous because the data is clocked synchronously with the CPU, so there are no delays when data is handed to the CPU. This is the fastest RAM technology commonly available today. Yes, there are faster DRAMs available, but they are not readily available.

ECC RAM stands for Error Correcting Code RAM. ECC is a means to electronically confirm that data being read from memory has not been changed from when it was stored or written into the memory. It is based on some clever math. Perhaps the first use of this was the Hamming Code developed by Mr. Hamming of Bell Telephone Labs. By adding extra memory bits to a memory module, and adding fairly simple logic circuitry, it is possible to detect and correct single bit errors and detect multiple bit errors. By adding more bits of memory, even multiple bits of error can be detected and corrected.

As desirable as ECC is, it can be used only when your motherboard has been designed to use it, with a special memory controller. ECC is primarily used in high end PCs and file servers.

EOS stands for ECC on a SIMM. Thus EOS is an ECC built completely and entirely onto a SIMM, and can be installed in any PC! What is more remarkable this product is only about $50 more than other memory. We stock some EOS, but because it is new and not in high demand we stock very little of it. We have only in 72 pin SIMMs, in almost any size, 4 MB, 8 MB, 16 MB, etc. Because it is new, there are problems of motherboard compatibility. I can only wonder how many systems would run without ever halting if EOS were being used. I believe computers would be more reliable and service free with EOS. Please consider it for any serious computer application, and shop for a motherboard which will accept EOS.

The physical size of transistors has become so small that cosmic rays, no joke, and naturally occurring radiation present in many materials can upset the stored charge in the bit cell of a dynamic ram, in effect altering a bit of data. We believe that memory with parity should be used in all business computers. ECC Memory, if affordable, is even better. It is reassuring to know that high end file servers use ECC.

The simplest form of ECC would be memory with one extra bit, called parity. This allows for the detection that a single bit has changed or become corrupted. Detection of parity error then signals the PC to stop. Apple computers have traditionally not used parity, while PCs have. Now, the newer PCs do not use parity either. This concerns all of us, because integrity of data is critical. Who wants to trust a computer which will corrupt your data and not detect it?



For more information on the Tulsa Computer Society click here



This page has been accessed times.
Tulsa Computer Society 02/07/97
Don Singleton, President
tcs@galstar.com