|
||
 |
||
|
| Memory: ROM & RAM |
 Email This
 Print ThisView My Personal Library |
|
Hardware February 2000 • Vol.8 Issue 2 |
Memory: ROM & RAM | ||
Jump to first occurrence of: [RAM]  Data storage is one of the most important and complicated functions of a computer. This is why memory is a necessity in terms of computers. Understanding all the complicated jargon and hardware involved with temporary and permanent storage is a necessary and complicated task, as well. We've compiled some of the most common questions about memory to help you understand some of the inner workings of your system. Q: What is memory? A: We can broadly classify memory as random-access memory (RAM) or read-only memory (ROM), with RAM being the most predominant. Technically, both RAM and ROM are random access. (Random access is the ability of a computer to locate and retrieve data immediately, without having to start at the beginning and read all the data until it finds the appropriate information.) This is why it might be more accurate to call RAM read/write memory. Because of the potential confusion, manufacturers are now using the terms volatile storage for RAM and non-volatile storage for ROM. (NOTE: See below for more information on memory.) Q: What is the difference between memory and storage? A: On some platforms, memory is known as main storage, and disk space is auxiliary storage. Multitasking operating systems will map virtual-storage areas on disk to real storage, so multiple programs can share the same physical memory addresses (data location). Any device that can store data is a storage device, but not all storage devices are memory devices. The term memory generally refers to small integrated circuits called chips. Q: How is data stored in a memory chip? A: A memory chip contains thousands or even millions of cells. Each cell may be in one of two states (usually represented as a 0 or a 1) and represents a single bit (a binary digit, either a 1 or 0) of data. It is important to note that memory chips do not store bytes (see below) of data; they store only bits. The memory controller keeps track of where the individual bits of the byte are stored, which may actually be in multiple chips. Q: What are a byte, kilobyte, and megabyte? A: Eight bits combined together make a byte and represent a single character. There are 1,024 bytes is a kilobyte (KB), and 1,024KB makes a megabyte (MB). Q: Are there different kinds of memory? A: All memory chips store data in cells that the system accesses by row and column coordinates (or addresses), but different kinds of memory vary in speed, density, and configuration. They also vary in their features. Q: What is chip density? A: A memory chip is comprised of one or more arrays of cells. The total number of cells in the chip is the density of the chip. For example, a four megabit (Mb) chip has 4 million cells; a 64Mb chip has 64 million cells. ROM chips typically have no more than 2 million cells; RAM chips can have 256Mb, and they are getting denser every year. Q: What is chip configuration? A: The number of arrays and the number of cells within each array is called the chip configuration. For example, a 128Mb chip that has eight arrays of 16 million cells would have a 16Mx8 configuration. Alternatively, another 128Mb chip might have only four arrays, but each array would have 32 million cells, making it a 32Mx4 configuration. Q: How is chip speed measured? A: Memory-chip speed is measured in nanoseconds (ns), which is one-billionth of a second. To illustrate how fast that is, a nanosecond is to one second what a second is to 31.8 years. Not all memory chips are measured in the same way, however, so you may not be able to directly compare speeds. Q: What is the difference between RAM and ROM? A: As the name implies, you cannot write to read-only memory; your system can only read it during normal operation. The computer manufacturer preprograms the data; the data will remain intact without any power. Random-access memory allows the CPU to read and write information at any time. This information is erased, however, when you turn the system's power off. We can further differentiate RAM chips between Static RAM (SRAM) and Dynamic RAM (DRAM), depending upon whether the data in the cells needs to be refreshed. (SRAM doesn't need to be refreshed.) Q: What does a computer use ROM for? A: When you power on a computer, it needs to load the system-initialization program from a known location. Early PCs required the operator to manually set the device address. However, this data is now stored in the complementary metal-oxide semiconductor (CMOS) or is at a fixed location. In the PC, the location is an address within a ROM chip called the Basic Input/Output System (BIOS, a special piece of software that controls the startup process of a computer and other basic functions). Q: What is RAM used for? A: The processor in a computer runs more than a million times faster than the typical auxiliary storage devices can operate (nanoseconds vs. milliseconds). For this reason, it is necessary to have a fast, temporary repository for the data that the system is most likely to request, which RAM is particularly suited for. Q: Which is faster, ROM or RAM? A: RAM is faster, which is why computers use it for main memory. PCs use ROM only for the initial boot program. Q: How does RAM operate? A: All memories have the same basic operation but variations exist. In general, signals sent on the Row Address Selector (RAS) and Column Address Selector (CAS) pins select a specific cell. For a read operation, the cell goes into an output buffer and then onto the associated output pin. A write operation would take the signal from the appropriate input pin and set the value of the cell accordingly. Q: What is DRAM used for? A: A computer usually implements main memory (sometimes called system memory) using DRAM because it can use DRAM in large quantities (tens or even hundreds of megabytes) for a modest cost. Even though DRAM is very fast, it is still as much as eight times slower than the processor, and many processor cycles can be lost waiting for the data to transfer from main memory. Q: What is SRAM used for? A: Most computer systems use a small amount of SRAM (no more than a few megabytes, generally) as a small, very fast cache. Even one megabyte of cache memory may add as much as $20 to the cost of the system. (Cache is a bank of high-speed memory set aside for frequently accessed data.) Q: How does cache work? A: Cache theory says there is a good chance that if a computer accesses a piece of data once, it will access it again soon (temporal locality) or that it will access the piece of data adjacent to it (spatial locality). When a system accesses a byte from memory, it stores it in the cache, along with the next several bytes, in the hope that the processor will find what it needs more quickly. Q: What is the difference between SRAM and DRAM? A: SRAM cells are a set of four to six transistors that makes up a sequential logic gate called a flip-flop. These flip-flops operate fast and retain their state (0 or 1) as long as there is power, giving rise to the term static RAM. DRAM cells are tiny capacitors, where the amount of charge in the capacitor determines the state of the cell. Whenever your system reads a cell, the capacitor is discharged and must be refreshed. This makes the operation of a DRAM device relatively slow compared to SRAM. In addition, the capacitors tend to lose their charge after awhile, requiring a regularly scheduled refresh operation, which also impairs performance. The main advantage of DRAM devices is the cost; they are usually five to 10 times cheaper than the same density SRAM. Q: Why is SRAM so much more expensive than DRAM? A: SRAM cells are larger than DRAM cells because of their circuitry; six transistors requires much more real estate than a single trench capacitor. This means fewer SRAM chips from the same-sized wafer. Because the cost of producing the wafer is the same, the per-unit cost of SRAM is much higher than DRAM. Q: What are SIMMs, DIMMS, and RIMMs? A: To facilitate easier memory upgrades, DRAM chips are mounted on small printed circuit boards (PCBs), which are in turn mounted into special sockets on the motherboard. These memory modules are single in-line memory modules (SIMMs) or dual in-line memory modules (DIMMs), depending upon the size. A newer type of memory, Rambus, is mounted on a special board called a Rambus in-line memory module (RIMM). These modules have a number of tin or silver leads, called pins, on the edge, which connect the control, address, and data lines of the chips and motherboard. Q: What is the difference between SIMMs, DIMMs, and RIMMs? A: SIMMs may be 30-pin or 72-pin. Older PCs used 30-pin SIMMs, but they are not longer in use. The 72-pin SIMM came into widespread use with the Pentium-based PCs. DIMMs have either 144 pins (for laptops) or 168 pins (for most other PC systems). RIMMs are 144-pin modules. Each type of module will plug into its corresponding socket. Q: Why do I need to install two SIMMs at a time in my system, but I only need to put in a single DIMM? A: SIMM modules read and write 32 bits of data at a time, but the Pentium data bus is 64 bits wide. This means two SIMM modules are required to fill up the bus width. On the other hand, DIMM modules are 64 bits wide. Q: Can I mix SIMMs, DIMMs, and RIMMs in my PC? A: Generally, the answer is no. There are circumstances, however, where this may work. Your motherboard manual or computer manufacturer will provide you with this information. Mixing modules on a motherboard that is not designed for this may cause damage to the motherboard, memory, or both. Q: Can I use any memory modules in my PC? A: Usually, a computer system can recognize chips with only certain speeds, sizes, configurations, and features, which the memory controller determines. For example, a 128MB module might be from 16 64Mb chips or from eight 128Mb chips. If the motherboard chipset does not support 128Mb chips, the system will not recognize the module, or it may show up as a 64MB module. Even if the chipset can handle the 128Mb chips, it may have problems with the configuration of the chips. Q: Who made my memory modules? A: The name of the module manufacturer will generally be on the PCB itself, not on the chips. You will see a sticker or a silk-screened name on the PCB. The name on the chips indicates the manufacturer of the chips only. If you can't find a name on the module, the manufacturer is probably a third-party, generic manufacturer. Q: What is the speed of my memory? A: Even though some manufacturers will include the speed of the chips in part of the name, some will not (most notably, Hyundai and LGS). The only reliable way to determine this is to locate the manufacturer's data sheet and look up the specs. Q: Are all DRAMs the same? A: The basic operation of all DRAM memories is the same, but there are many flavors, including FPM, EDO RAM, SDRAM, DDR SDRAM, and DRDRAM. The motherboard chipset will likely not be able to support all these types. In addition, modules may have other special features, such as error checking or buffering, which the chipset may not support. (See the "Species Of DRAM" sidebar for more information.) Q: What is FPM? A: Fast Page Mode (FPM) is a variation on the basic memory operation that allows a system to access multiple columns in the same row without having to respecify the row address each time. This saves time in accessing data. FPM is available only on SIMMs, and it is now almost obsolete because of advancements in DRAM technology. Q: What is EDO? A: Extended Data Output (EDO) RAM is an enhancement of the FPM operation. With FPM, the data output line was disabled as soon as the next column address was being input. By allowing the data to remain on the output line while a new column was being read in, the time between column output and data readiness could be shortened by as much as 40%, resulting in faster page operation. EDO is on both SIMM and DIMM modules, and many systems still use it. SDRAM, however, is the current memory of choice. Q: What is SDRAM? A: Synchronous DRAM (SDRAM) performs all operations under control of the system clock (the main clock in a computer that coordinates all timing applications). With asynchronous operation, the CPU has to wait after a memory request because it doesn't know exactly when the data will be ready. With synchronous operation, the CPU can spend the time between the request and the completion of the request doing other tasks because it knows exactly how many clock cycles the operation will take. In addition, SDRAM has the ability to internally generate the next column address and immediately send the data out (this is called burst mode). This means that after the first access, the system can transfer data every clock cycle. SDRAM is only available on DIMM modules. Q: What is Direct Rambus memory? A: Direct Rambus Dynamic RAM (DRDRAM) is essentially DRAM that can output data extremely fast. The data is broken up into packets and then reassembled at the other end. Instead of having a single memory controller handling all transfers, each Rambus device has its own controller. Just as the Synchronous operation of SDRAM allows the CPU to use the wait time more efficiently, Rambus operation allows the CPU to use the memory bus more efficiently. DRDRAM is available only on RIMM modules. Q: What is DDR? A: Although there are many variations of DRAM, the internal operations are all essentially the same and suffer the same, relatively slow operation. As a means of improving the output speed of DRAM, designers have devised a way for two transfers to occur during the same clock cycle. The clock signal is actually a very regular voltage change on a particular pin. By allowing data to be output on both up and down signals (rather than just the up signal), a system can transfer twice as much data using the same clock speed. This is called Double Data-Rate (DDR) operation. Q: What is the difference between an x32 and x36 (or x64 and x72) module? A: Manufacturers identify modules using the module width and the number of memory lines. For example, a 32MB SIMM would be an 8Mx32 module; a 32MB DIMM would be a 4Mx64 module. If the module includes parity or Error Checking and Correction support, it will require additional output pins, making it an 8Mx36 (or 4Mx72) module. Q: Why isn't my 10ns SDRAM 10 times faster than my 60ns EDO? A: The EDO speed indicates the time from the first row access until the data appears on the output line, whereas the SDRAM speed indicates the time between column addresses in burst mode, which is when data is transmitted all at once rather than sequentially. In actual operation, the 10ns SDRAM would be about 40% faster than the EDO (assuming no caching is involved). Q: What is Parity? A: Parity is a form of error detection. When a byte is stored in memory, the 1s or 0s are counted. If the result is odd, a 1 is stored in the parity bit; otherwise, a 0 is stored. Later, when the byte is read, the 1s (or 0s) are again counted and compared with the stored parity bit. If they match, the byte is assumed correct, but if they don't, a data error is signaled indicating data corruption. This technique is limited because it can detect only single-bit errors. Q: What is ECC? A: Error Checking and Correction (ECC) is a more advanced form of error checking. A 64-bit value is put through an algorithm, resulting in an 8-bit ECC word. Just as with parity, this value is stored when the data is written and then recalculated and compared when the data is read back in. The advantage of this method is that if a bit changes, this process, known as single-bit correction, can recalculate and restore the original bit value. Though ECC can detect 2-, 3-, and 4-bit errors, it cannot correct these. Q: What is PC100 (or PC133) SDRAM? A: When memory bus speeds increased from 66 megahertz (MHz) to 100MHz (and then to 133MHz), a set of guidelines was introduced to help ensure reliable system operation. SDRAM modules that conform to the PC100 (or PC133) specification are no different in operation than non-conforming SDRAM is; they are simply SDRAM modules that operate on the faster bus. Q: Can a PC100 (or PC133) SDRAM be used on a 66MHz bus? A: Yes, a memory chip can always operate at a slower speed than it was designed for, but it may not operate at a faster speed. Q: What is latency? A: The amount of time that passes from when an operation begins until the next operation can begin is the latency period. For example, the amount of time from when a column is accessed and the data is available is called the CAS Latency. Q: What do CAS2 and CAS3 mean? A: These terms refer to a CAS Latency value of two clock cycles and three clock cycles, respectively. Q: What is buffered and registered memory? A: If an EDO module has more address lines than the motherboard can handle because of the number of chips or the chip configuration, a special buffer chip must make sure the correct chips are being accessed. With SDRAM, registered memory accomplishes the same thing. The memory controller must support this type of module if it is to be used.  by Dean Kent
|
|
Home Copyright & Legal Information Privacy Policy Site Map Contact Us