by Lynn True
Most computer users are aware that the heart of their computer is the Central Processing Unit (CPU) and the system’s main memory, the Random Access Memory (RAM). This is where all of the program instructions are carried out and also where calculation res ults and data are stored and processed. The other components, such as the hard disks, video and sound cards, other controllers, and various parts related to either the input or output of data, are considered peripheral to this core, hence the name periph erals.
The main system memory, one of many kinds of modern RAM, must handle the spiraling demands of faster CPUs, more complex programs (and operating systems), and all of the other components we put in our machines. Some tasks can now be off-loaded to memory t hat is on your video card, sound card, or SCSI and other controllers.
Remember This !
Memory chips, which are made up of a transistor and a capacitor, store their data as electronic charges; the transistor turns that charge, which is actually stored in the capacitor, on or off. RAM memory allows the CPU to change the on or off state of th e electronic charges, whereas a Read Only Memory (ROM) charge state is either permanently on or off.
The CPU (for most PCs and clones it’s an Intel, AMD or Cyrix) assigns and remembers what information goes where via the data bus, the connection that allows the actual data communication. When the CPU is ready to collect the stored instructions or data w hich it needs to do whatever it’s doing, it searches in a particular order.
First, it looks in the very fast but very small amount of memory inside itself, called the L1 cache. Next, it tries the larger, slower (but still relatively fast) L2, or second level cache memory. If it still can’t find what it wants, it searches in the even larger but definitely much slower main system memory. The last place the CPU will look is in a physical storage area such as a hard disk, which is considered extremely slow by CPU standards.
The Windows operating system also uses hard disk space to store data and instructions that overflow the amount of system RAM available (the “swap file”), but, of course, hard drives are much slower than RAM chips. The best way to make sure your Windows s ystem runs as fast as it can is to make sure it has enough system RAM. The type of RAM you need depends on your hardware configuration.
How Fast Is Fast?
One of the most important measurements of memory performance is access time, the amount of time it takes for the CPU to request and receive a specific piece of data from the memory. The challenge the computer industry faces is creating faster and faster memory to keep pace with the ever increasing speed of CPUs—and how to keep the cost of memory competitive at the same time. That’s good for consumers—with RAM prices at an all -time low, there’s no excuse for crippling a computer because of lack of RAM.
Most modern system RAM chips have a 50 to 60ns (nanosecond, one billionth of a second) access time; older RAM from just a few years ago had access times of 100 to 120ns. Ideally, memory access time would be zero, but the only way that could happen would be if the CPU itself contained enough memory space to store all of its data.
Speed It Up!
One of the ways a CPU can do its duties faster is by utilizing cache memory. Cache chips usually have a 20ns access time or better, but they are more expensive than the main system RAM, so your system has a smaller amount of cache memory. Because of this size limitation, only specific data is put there.
The goal of a memory caching system is to process memory requests as close to the CPU’s speed as possible by predicting—and having ready to send—the information your CPU is expected to use next. A separate, quite sophisticated cache controller directs th is data prediction/selection process.
The function of the L2 cache is to be between the CPU and the RAM and to provide data to meet the CPU’s requests faster than the main memory. When the correct data is found waiting in the L2 cache, it is called a cache hit.
So, What Kind?
As you’ve probably noticed recently, ads for computers have changed as computer configuration choices continue to increase exponentially. Instead of just listing the amount of RAM, nowadays the ads state exactly what kind of RAM it is—with initials. The types of memory available have become a veritable alphabet soup of names.
Confused by all those acronyms? The most common types of RAM are:
RAM
The initials stand for Random Access Memory (RAM), which is a general term that was originally applied to any memory that can be read/written in a random order (usually called nonlinear). Lately the term RAM has referred specifically to chip -based memory because all chip-based memory is random-access. Read Only Memory (ROM), like the name indicates, can be read from but not written to; it is different from RAM but not its opposite.
SIMM & DIMM
Both Single Inline Memory Modules (SIMMs) and Dual Inline Memory Modules (DIMMs) are common types of modules in which RAM is packaged. SIMMs are older technology with a 32 -bit data path; they can be installed one at a time on most 486 or older machines but must be in pairs for Pentium computers because of the Pentium’s wider data path.
DIMMs are newer and have a 64-bit data path—a better match for Pentium and other more recent CPUs—so they are equivalent to two SIMMs. They can be installed one at a time in most Pentium machines, thereby making them more economical in the long run.
DRAM
Dynamic RAM is the standard type of main memory in modern computers. Its name comes from the way it works: information is stored in a capacitor as a series of electrical charges which build and discharge within one millisecond of each other. To retain th e data the chip must be constantly refreshed—the reason that it is called dynamic.
FPM RAM
Originally, all PC main memory was Fast Page-Mode RAM, but since there wasn’t any other kind there was no need to state the type. As memory technology progressed, its speed increased from 120ns to 60ns. But 60ns became too slow when Pentium computers were able to run the bus at 66 MHz—it couldn’t k eep up! Actually, 60ns RAM does its data access at less than 30 MHz, which led to the invention of RAM caching.
EDO RAM
A newer, improved type of FPM RAM, Extended-Data-Out RAM (EDO RAM), speeds up the data read/write cycle by being a little smarter; it knows that most of the time when the CPU requests to use a particular memory space it will probably want some more of what’s nearby, so it “hangs around” instead of star ting each new request from scratch.
Unfortunately, however, even though EDO RAM can be up to 40% faster and can operate at a bus speed of 66 MHz, the latest processors can support even faster bus speeds.
BEDO RAM
Another forward step in EDO memory technology is Burst Extended-Data-Out RAM (BEDO RAM).The term burst means that four requests at a time are sent by the CPU, with the memory location information for all four requests at the beginning of that block of data. The first request is processed at the regular (50ns) speed becaus e of the normal instruction-related delays involved in the read/write process, but the next three requests already have their “overhead” information present by the time they are ready to be processed, so it only takes about 10ns to fill those requests.
BEDO RAM was developed to be a less expensive alternative to SDRAM (more about that below) and currently it operates at almost the same speed as SDRAM. But, again, even though BEDO RAM is substantially faster than plain ol’ FPM RAM, it still can’t functi on reliably at bus speeds faster than 66 MHz.
SDRAM
The new kid on the block is Synchronous Dynamic RAM (SDRAM), and you can only recently find this type of memory in PC ads. SDRAM allows two requests from the CPU to be processed at the same time and it can operate at bus speeds up to 100 MHz (and it won’ t be long until the rest of the computer catches up!) A major accomplishment, though, is that this type of memory is synchronized with the system clock. Intensive research and development is still going on.
SLDRAM
An improvement on SDRAM, SLDRAM can operate at even higher bus speeds and uses small packets of data (similar to networks) to synchronize with the bus and to communicate between the main memory and the CPU. This new technology is currently being proposed as a standard by the SCI Association at California’s Santa Clara University and various industry leaders.
SLDRAM’s packet system of communication will decrease the need for trying to squeak more performance out of existing DRAM chip design. Ideally, SLDRAM should be high -performance, lower cost memory.
We’ll just have to wait and see what happens...
Reprinted courtesy of Lynn True.
E-mail me at webmaster@hal-pc.org with any comments you have and tell me what you want to see here.