Ch05 coa9e

Editor's Notes

• #3: We begin this chapter with a survey of semiconductor main memory subsystems,including ROM, DRAM, and SRAM memories. Then we look at error controltechniques used to enhance memory reliability. Following this, we look at moreadvanced DRAM architectures.
• #4: In earlier computers, the most common form of random-access storage for computermain memory employed an array of doughnut-shaped ferromagnetic loopsreferred to as cores. Hence, main memory was often referred to as core, a term thatpersists to this day. The advent of, and advantages of, microelectronics has longsince vanquished the magnetic core memory. Today, the use of semiconductor chipsfor main memory is almost universal. Key aspects of this technology are exploredin this section.The basic element of a semiconductor memory is the memory cell. Although a varietyof electronic technologies are used, all semiconductor memory cells share certainproperties:• They exhibit two stable (or semistable) states, which can be used to representbinary 1 and 0.• They are capable of being written into (at least once), to set the state.• They are capable of being read to sense the state.Figure 5.1 depicts the operation of a memory cell. Most commonly, the cellhas three functional terminals capable of carrying an electrical signal. The selectterminal, as the name suggests, selects a memory cell for a read or write operation.The control terminal indicates read or write. For writing, the other terminalprovides an electrical signal that sets the state of the cell to 1 or 0. For reading, thatterminal is used for output of the cell’s state. The details of the internal organization,functioning, and timing of the memory cell depend on the specific integratedcircuit technology used and are beyond the scope of this book, except for a briefsummary. For our purposes, we will take it as given that individual cells can beselected for reading and writing operations.
• #5: All of the memory types that we will explore in this chapter are random access. That is,individual words of memory are directly accessed through wired-in addressing logic.Table 5.1 lists the major types of semiconductor memory. The most commonis referred to as random-access memory (RAM). This is, in fact, a misuse of theterm, because all of the types listed in the table are random access. One distinguishingcharacteristic of memory that is designated as RAM is that it is possibleboth to read data from the memory and to write new data into the memory easilyand rapidly. Both the reading and writing are accomplished through the use ofelectrical signals.The other distinguishing characteristic of RAM is that it is volatile. A RAMmust be provided with a constant power supply. If the power is interrupted, thenthe data are lost. Thus, RAM can be used only as temporary storage. The two traditionalforms of RAM used in computers are DRAM and SRAM.
• #7: Figure 5.2a is a typical DRAM structure for an individual cell that stores 1 bit.The address line is activated when the bit value from this cell is to be read or written.The transistor acts as a switch that is closed (allowing current to flow) if a voltage isapplied to the address line and open (no current flows) if no voltage is present onthe address line.For the write operation, a voltage signal is applied to the bit line; a high voltagerepresents 1, and a low voltage represents 0. A signal is then applied to theaddress line, allowing a charge to be transferred to the capacitor.For the read operation, when the address line is selected, the transistor turnson and the charge stored on the capacitor is fed out onto a bit line and to a senseamplifier. The sense amplifier compares the capacitor voltage to a reference valueand determines if the cell contains a logic 1 or a logic 0. The readout from the celldischarges the capacitor, which must be restored to complete the operation.Although the DRAM cell is used to store a single bit (0 or 1), it is essentiallyan analog device. The capacitor can store any charge value within a range; a thresholdvalue determines whether the charge is interpreted as 1 or 0.
• #9: Figure 5.2b is a typical SRAM structure for an individual cell. Four transistors(T1, T2, T3, T4) are cross connected in an arrangement that produces a stable logicstate. In logic state 1, point C1 is high and point C2 is low; in this state, T1 and T4 are offand T2 and T3 are on. In logic state 0, point C1 is low and point C2 is high; in this state,T1 and T4 are on and T2 and T3 are off. Both states are stable as long as the directcurrent (dc) voltage is applied. Unlike the DRAM, no refresh is needed to retain data.As in the DRAM, the SRAM address line is used to open or close a switch.The address line controls two transistors (T5 and T6). When a signal is applied tothis line, the two transistors are switched on, allowing a read or write operation. Fora write operation, the desired bit value is applied to line B, while its complementis applied to line B. This forces the four transistors (T1, T2, T3, T4) into the properstate. For a read operation, the bit value is read from line B.
• #10: Both static and dynamic RAMs are volatile; that is,power must be continuously supplied to the memory to preserve the bit values.A dynamic memory cell is simpler and smaller than a static memory cell. Thus, aDRAM is more dense (smaller cells = more cells per unit area) and less expensivethan a corresponding SRAM. On the other hand, a DRAM requires the supportingrefresh circuitry. For larger memories, the fixed cost of the refresh circuitry is morethan compensated for by the smaller variable cost of DRAM cells. Thus, DRAMstend to be favored for large memory requirements. A final point is that SRAMs aresomewhat faster than DRAMs. Because of these relative characteristics, SRAM isused for cache memory (both on and off chip), and DRAM is used for main memory.
• #12: When only a small number of ROMs with a particular memory content isneeded, a less expensive alternative is the programmable ROM (PROM). Like theROM, the PROM is nonvolatile and may be written into only once. For the PROM,the writing process is performed electrically and may be performed by a supplieror customer at a time later than the original chip fabrication. Special equipment isrequired for the writing or “programming” process. PROMs provide flexibility andconvenience. The ROM remains attractive for high-volume production runs.
• #13: Another variation on read-only memory is the read-mostly memory, which isuseful for applications in which read operations are far more frequent than writeoperations but for which nonvolatile storage is required. There are three commonforms of read-mostly memory: EPROM, EEPROM, and flash memory.The optically erasable programmable read-only memory (EPROM) is readand written electrically, as with PROM. However, before a write operation, all thestorage cells must be erased to the same initial state by exposure of the packagedchip to ultraviolet radiation. Erasure is performed by shining an intense ultravioletlight through a window that is designed into the memory chip. This erasure processcan be performed repeatedly; each erasure can take as much as 20 minutes toperform. Thus, the EPROM can be altered multiple times and, like the ROM andPROM, holds its data virtually indefinitely. For comparable amounts of storage, theEPROM is more expensive than PROM, but it has the advantage of the multipleupdate capability.A more attractive form of read-mostly memory is electrically erasable programmableread-only memory (EEPROM). This is a read-mostly memory that canbe written into at any time without erasing prior contents; only the byte or bytesaddressed are updated. The write operation takes considerably longer than the readoperation, on the order of several hundred microseconds per byte. The EEPROMcombines the advantage of non-volatility with the flexibility of being updatable inplace, using ordinary bus control, address, and data lines. EEPROM is more expensivethan EPROM and also is less dense, supporting fewer bits per chip.Another form of semiconductor memory is flash memory (so named becauseof the speed with which it can be reprogrammed). First introduced in the mid-1980s,flash memory is intermediate between EPROM and EEPROM in both cost andfunctionality. Like EEPROM, flash memory uses an electrical erasing technology.An entire flash memory can be erased in one or a few seconds, which is much fasterthan EPROM. In addition, it is possible to erase just blocks of memory rather thanan entire chip. Flash memory gets its name because the microchip is organized sothat a section of memory cells are erased in a single action or “flash.” However,flash memory does not provide byte-level erasure. Like EPROM, flash memoryuses only one transistor per bit, and so achieves the high density (compared withEEPROM) of EPROM.
• #18: Main memory is composed of a collection of DRAM memory chips. A number ofchips can be grouped together to form a memory bank. It is possible to organizethe memory banks in a way known as interleaved memory. Each bank is independentlyable to service a memory read or write request, so that a system withK banks can service K requests simultaneously, increasing memory read or writerates by a factor of K. If consecutive words of memory are stored in differentbanks, then the transfer of a block of memory is speeded up. Appendix E exploresthe topic of interleaved memory.
• #21: The simplest of the error-correcting codes is the Hamming code devised byRichard Hamming at Bell Laboratories. Figure 5.8 uses Venn diagrams to illustratethe use of this code on 4-bit words (M = 4). With three intersecting circles,there are seven compartments. We assign the 4 data bits to the inner compartments(Figure5.8a). The remaining compartments are filled with what are called paritybits. Each parity bit is chosen so that the total number of 1s in its circle is even(Figure5.8b). Thus, because circle A includes three data 1s, the parity bit in thatcircle is set to 1. Now, if an error changes one of the data bits (Figure 5.8c), it is easilyfound. By checking the parity bits, discrepancies are found in circle A and circleC but not in circle B. Only one of the seven compartments is in A and C but not B.The error can therefore be corrected by changing that bit.
• #22: The first three columns of Table 5.2lists the number of check bits required for various data word lengths.For convenience, we would like to generate a 4-bit syndrome for an 8-bit dataword with the following characteristics:• If the syndrome contains all 0s, no error has been detected.• If the syndrome contains one and only one bit set to 1, then an error hasoccurred in one of the 4 check bits. No correction is needed.• If the syndrome contains more than one bit set to 1, then the numerical valueof the syndrome indicates the position of the data bit in error. This data bit isinverted for correction.
• #23: To achieve these characteristics, the data and check bits are arranged into a12-bit word as depicted in Figure 5.9. The bit positions are numbered from 1 to 12.Those bit positions whose position numbers are powers of 2 are designated as checkbits.
• #24: Figure 5.10 illustrates the calculation. The data and check bits arepositioned properly in the 12-bit word. Four of the data bits have a value 1 (shadedin the table), and their bit position values are XORed to produce the Hammingcode 0111, which forms the four check digits. The entire block that is stored is001101001111. Suppose now that data bit 3, in bit position 6, sustains an error and ischanged from 0 to 1. The resulting block is 001101101111, with a Hamming code of0111. An XOR of the Hamming code and all of the bit position values for nonzerodata bits results in 0110. The nonzero result detects an error and indicates that theerror is in bit position 6.
• #25: The code just described is known as a single-error-correcting (SEC) code.More commonly, semiconductor memory is equipped with a single-error-correcting,double-error-detecting (SEC-DED) code. As Table 5.2 shows, such codes requireone additional bit compared with SEC codes.Figure 5.11 illustrates how such a code works, again with a 4-bit data word.The sequence shows that if two errors occur (Figure 5.11c), the checking proceduregoes astray (d) and worsens the problem by creating a third error (e). To overcomethe problem, an eighth bit is added that is set so that the total number of 1s in thediagram is even. The extra parity bit catches the error (f).An error-correcting code enhances the reliability of the memory at the cost ofadded complexity. With a 1-bit-per-chip organization, an SEC-DED code is generallyconsidered adequate. For example, the IBM 30xx implementations used an 8-bit SECDEDcode for each 64 bits of data in main memory. Thus, the size of main memory isactually about 12% larger than is apparent to the user. The VAX computers used a 7-bitSEC-DED for each 32 bits of memory, for a 22% overhead. A number of contemporaryDRAMs use 9 check bits for each 128 bits of data, for a 7% overhead [SHAR97].
• #28: Figure 5.12 shows the internal logic of IBM’s 64-Mb SDRAM [IBM01], whichis typical of SDRAM organization.
• #29: Table 5.4 defines the various pin assignments.The SDRAM employs a burst mode to eliminate the address setup time androw and column line pre-charge time after the first access. In burst mode, a series ofdata bits can be clocked out rapidly after the first bit has been accessed. This modeis useful when all the bits to be accessed are in sequence and in the same row of thearray as the initial access. In addition, the SDRAM has a multiple-bank internalarchitecture that improves opportunities for on-chip parallelism.The mode register and associated control logic is another key feature differentiatingSDRAMs from conventional DRAMs. It provides a mechanism tocustomize the SDRAM to suit specific system needs. The mode register specifiesthe burst length, which is the number of separate units of data synchronously fedonto the bus. The register also allows the programmer to adjust the latency betweenreceipt of a read request and the beginning of data transfer.The SDRAM performs best when it is transferring large blocks of data serially,such as for applications like word processing, spreadsheets, and multimedia.
• #30: Figure 5.13 shows an example of SDRAM operation.
• #32: Figure 5.14 illustrates the RDRAM layout. The configuration consists ofa controller and a number of RDRAM modules connected via a common bus.The controller is at one end of the configuration, and the far end of the bus isa parallel termination of the bus lines. The bus includes 18 data lines (16 actualdata, two parity) cycling at twice the clock rate; that is, 1 bit is sent at the leadingand following edge of each clock signal. This results in a signal rate on eachdata line of 800 Mbps. There is a separate set of 8 lines (RC) used for addressand control signals. There is also a clock signal that starts at the far end fromthe controller propagates to the controller end and then loops back. A RDRAMmodule sends data to the controller synchronously to the clock to master, and thecontroller sends data to an RDRAM synchronously with the clock signal in theopposite direction. The remaining bus lines include a reference voltage, ground,and power source.
• #33: SDRAM is limited by the fact that it can only send data to the processor once perbus clock cycle. A new version of SDRAM, referred to as double-data-rate SDRAMcan send data twice per clock cycle, once on the rising edge of the clock pulse andonce on the falling edge.DDR DRAM was developed by the JEDEC Solid State TechnologyAssociation, the Electronic Industries Alliance’s semiconductor-engineering-standardizationbody. Numerous companies make DDR chips, which are widely used indesktop computers and servers.
• #34: Figure 5.15 shows the basic timing for a DDR read. The data transfer is synchronizedto both the rising and falling edge of the clock. It is also synchronized toa bidirectional data strobe (DQS) signal that is provided by the memory controllerduring a read and by the DRAM during a write. In typical implementations theDQS is ignored during the read. An explanation of the use of DQS on writes isbeyond our scope; see [JACO08] for details.There have been two generations of improvement to the DDR technology.DDR2 increases the data transfer rate by increasing the operational frequencyof the RAM chip and by increasing the prefetch buffer from 2 bits to 4 bitsper chip. The prefetch buffer is a memory cache located on the RAM chip. Thebuffer enables theRAM chip to preposition bits to be placed on the data bus asrapidly as possible. DDR3, introduced in 2007, increases the prefetch buffer sizeto 8 bits.Theoretically, a DDR module can transfer data at a clock rate in the range of200 to 600 MHz; a DDR2 module transfers at a clock rate of 400 to 1066 MHz; anda DDR3 module transfers at a clock rate of 800 to 1600 MHz. In practice, somewhatsmaller rates are achieved.Appendix K provides more detail on DDR technology.
• #35: Cache DRAM (CDRAM), developed by Mitsubishi [HIDA90, ZHAN01], integratesa small SRAM cache (16 Kb) onto a generic DRAM chip.The SRAM on the CDRAM can be used in two ways. First, it can be used as atrue cache, consisting of a number of 64-bit lines. The cache mode of the CDRAMis effective for ordinary random access to memory.The SRAM on the CDRAM can also be used as a buffer to support the serialaccess of a block of data. For example, to refresh a bit-mapped screen, the CDRAMcan prefetch the data from the DRAM into the SRAM buffer. Subsequent accessesto the chip result in accesses solely to the SRAM.