What is the name for the part of the memory used to store data temporarily while it is waiting to be used?

Memory

RAM is often an area that offers significant performance improvements for most applications, including SQL Server and Windows 2000. RAM provides the working memory of your system; by having a great deal RAM, you avoid having to access the much slower disk arrays. More RAM is always better, but systems that run parallel queries and populate full-text indexes (and most are primarily DSS systems) require even more RAM.

If the more RAM, the better, how much can you get? Windows 2000 Server is a 32-bit operating system; as a consequence, it can address only 4GB of memory. Windows 2000 Advanced Server and Data Center are also 32-bit operating systems, but through the Address Windowing Extensions (AWE) API, it can address 8GB and 64GB, respectively. You need at least 64 MB of RAM for SQL Server, plus 32 MB of RAM for the operating system. The more RAM you have, the more data that can be stored in the data cache and the more procedures that can be stored in the procedure cache. If the data and procedures that you are querying are in RAM, the server will not have to visit its disk system, providing a substantial improvement in performance.

RAM and ROM

RAM is volatile memory used to hold instructions and data of currently running programs. It loses integrity after loss of power. RAM memory modules are installed into slots on the computer motherboard. Read-only memory (ROM) is nonvolatile: Data stored in ROM maintains integrity after loss of power. The basic input/output system (BIOS) firmware is stored in ROM. While ROM is “read only,” some types of ROM may be written to via flashing, as we will see shortly in the Flash Memory section.

Note

The volatility of RAM is a subject of ongoing research. Historically, it was believed that DRAM lost integrity after loss of power. The “cold boot” attack has shown that RAM has remanence; that is, it may maintain integrity seconds or even minutes after power loss. This has security ramifications, as encryption keys usually exist in plaintext in RAM; they may be recovered by “cold booting” a computer off a small OS installed on DVD or USB key and then quickly dumping the contents of memory. A video on the implications of cold boot, Lest We Remember: Cold Boot Attacks on Encryption Keys, is available at http://citp.princeton.edu/memory/. Remember that the exam sometimes simplifies complex matters. For the exam, simply remember that RAM is volatile (though not as volatile as we once believed).

6.16.2 Random Access Memory

RAM can be modeled in a number of ways in VHDL. In the example RAM model [4] in Figure 6.87, the address, data, and control signals are shown. Each of 16 addresses holds eight bits of data. Data is written to the memory when the CE (chip enable) and the WE (write enable) signals are active low, and data is read from the memory when the CE and the OE (output enable) signals are active low. In models of this type, care is needed to identify what will happen in the circuit if unexpected control signals are applied on CE, WE, and OE. Line 26 in the code sets the RAM output to high impedance under other conditions. An example VHDL code implementation for this design is shown in Figure 6.88.

In this example, the input address signal is an integer type, and the data is a bidirectional (inout) standard logic vector.

An example VHDL test bench for this design is shown in Figure 6.89. As data is written to and read from the RAM model, the applied stimulus is set to high impedance (z) when data is to be read from the memory, and set to the logic levels required to store when data is written to the memory. In this test bench, a value of 12910 is written to the memory address 0 and then read back.

Memory

Random access memory (RAM) also comes in a few flavors, but all of them appear to look the same. However, there are slight differences that need to be considered when you’re upgrading your computer, because they have different form factors and speeds.

Whether you’re using this computer for a DVS server or a workstation, it’s best to max out the memory on the motherboard, but keep in mind that the standard 32-bit Windows OS can only read up to 4 GB of RAM. If the software or hardware requirements call for anything more than 4 GB of memory, it’s recommended that you upgrade to a 64-bit system.

Memory Types

Memory arrays are specified by their size (depth × width) and the number and type of ports. All memory arrays store data as an array of bit cells, but they differ in how they store bits.

Memories are classified based on how they store bits in the bit cell. The broadest classification is random access memory (RAM) versus read only memory (ROM). RAM is volatile, meaning that it loses its data when the power is turned off. ROM is nonvolatile, meaning that it retains its data indefinitely, even without a power source.

RAM and ROM received their names for historical reasons that are no longer very meaningful. RAM is called random access memory because any data word is accessed with the same delay as any other. In contrast, a sequential access memory, such as a tape recorder, accesses nearby data more quickly than faraway data (e.g., at the other end of the tape). ROM is called read only memory because, historically, it could only be read but not written. These names are confusing, because ROMs are also randomly accessed. Worse yet, most modern ROMs can be written as well as read! The important distinction to remember is that RAMs are volatile and ROMs are nonvolatile.

Robert Dennard, 1932–

Invented DRAM in 1966 at IBM. Although many were skeptical that the idea would work, by the mid-1970’s DRAM was in virtually all computers. He claims to have done little creative work until, arriving at IBM, they handed him a patent notebook and said, “put all your ideas in there.” Since 1965, he has received 35 patents in semiconductors and microelectronics. (Photo courtesy of IBM.)

The two major types of RAMs are dynamic RAM (DRAM) and static RAM (SRAM). Dynamic RAM stores data as a charge on a capacitor, whereas static RAM stores data using a pair of cross-coupled inverters. There are many flavors of ROMs that vary by how they are written and erased. These various types of memories are discussed in the subsequent sections.

III.A RAM

Random access memory (RAM), is the most important type of memory in semiconductors because of high-speed access to a memory word at any address. Information can be written in and read out of a memory at any of its address locations with the same access time, unlike floppy disks. Depending on how memory cells are realized by an electronic circuit, RAMs are further classified into two types—static RAM (SRAM) and dynamic RAM (DRAM). Each cell of a static RAM consists of a transistor circuit (called flip-flop) realized in CMOS, as shown in Fig. 10. Thus as long as a power supply is connected, stored information is maintained. Each memory cell of a dynamic RAM keeps information by storing an electric charge on a very small capacitance, which is usually called storage capacitance. Information 0 and 1 is represented by electric charge and no charge, respectively, on the storage capacitance in a memory cell. Figure 11 shows an example of a dynamic RAM, where only one MOSFET is connected to the ground through a storage capacitance C. Since an electric charge stored on capacitance C gradually leaks (mostly from the diffusion region of the MOSFET to the ground), the same information must be written again before the charge completely leaks out. In other words, the memory cell must be refreshed. Because a single MOSFET is used instead of many MOSFETs used for SRAM in Fig. 10, a dynamic RAM occupies a much smaller area. Thus in the same chip area of a static RAM, a dynamic RAM can be packed about 4–10 times more, depending on the technology, but SRAM is usually much faster than DRAM in reading and writing.

Every four or four and a half years, RAM with memory capacity four times greater than the previous one has been introduced with a new processing technology. In other words, the number of transistors on a chip doubles every year and half and this empirical observation is called Moore's law. When the new RAM is introduced, the price of a memory package is initially high, and goes down due to manufacturing improvements.

The performance of DRAMs became a bottleneck for faster operation of a computer. Conventional DRAMs have been replaced by new types of DRAM, such as extended data out (EDO) DRAM, double-data rate (DDR), synchronous DRAM (SDRAM), and Rambus DRAM.

Primary storage

Primary storage, sometimes called main storage, can be random access memory, read only memory or a combination of both.

Random access memory – called RAM, this is where data and programs are transferred to (from secondary storage) when the CPU requires them. It is volatile in that the contents of RAM are only available when the computer is on. When required, the CPU transfers data or programs from secondary storage to RAM. It does this because the CPU can access data thousands of times faster from RAM than from secondary storage. An extremely fast type of RAM, called cache (pronounced kay-sh) is often found on modem computer systems. This is a section of computer memory that can be accessed at very high speeds and in which information is stored for fast retrieval. A cache size of 256Kb is common. Cache is a very expensive form of RAM, and consequently, only small amounts are used in computer systems.

Read only memory – called ROM, this contains data and instructions required by the computer that never change. Consequently, these instructions are permanently etched into the chip when manufactured and so are not lost when the computer is turned off (non-volatile). No data can be stored in ROM since it is read-only.

Combination of RAM and ROM – some ROM chips can be altered after manufacture. These are called PROMs (Programmable ROMs), EPROMs (Erasable PROMs) or EEPROMs (Electronic EPROMs). A special kind of EEPROM called Flash Memory can be erased and re-written in blocks, instead of one byte at a time (bytes are discussed later). They are used in hardware devices where change is constantly taking place. Flash memory enables the functionality of hardware to be upgraded without replacing the hardware.

8.3.3 RAM

Random access memory is an array of memory registers in which data can be stored and retrieved; it is short-term memory and is sometimes called read–write memory. It is memory that is external to the microprocessor, usually in the form of a bank of semiconductor chips on the motherboard (logic board) to which the user can add extra memory by purchasing additional chips. RAM is volatile, meaning that it is a storage medium in which information is a set of easily changed electrical patterns which are lost if power is turned off because the electricity to maintain the patterns is then lost.4 For this reason disk drives (hard drives, CDs, etc.) or flash memory sticks which have the advantage of retaining the information stored on them even when the computer is off are used for permanent storage. Disks, for example, can do this because they store information magnetically, not electrically, using audio and video tape technology which lays down the information as a sequence of tiny permanent magnets on magnetic tape. The downside of disk storage is that it is many orders of magnitude slower in transfer of information than RAM is (typically 1 ns for RAM and 10 ms for hard disks). Hence, if disk storage has to be used when working with an application program in which information and data are fetched from memory, processed, and then temporarily stored, and this cycle is repeated over and over during execution of a program, one can see that the program would run terribly slow. It is precisely for this reason that high-speed RAM is used during execution of a program and is therefore referred to as the main memory. The slower disk storage is referred to as secondary memory.

Virtual memory is a clever technique of using secondary memory such as disks to extend the apparent size of main memory (RAM). It is a technique for managing a limited amount of main memory and a generally much larger amount of lower-speed, secondary memory in such a way that the distinction is largely transparent to a computer user. Virtual memory is implemented by employing a memory management unit (MMU) which identifies what data are to be sent from disk to RAM and the means of swapping segments of the program and data from disk to RAM. Practically all modern operating systems use virtual memory, which does not appreciably slow the computer but allows it to run much larger programs with a limited amount of RAM.

A typical use of a computer is as follows: suppose a report is to be typed. Word-processing software which is permanently stored on the hard disk of a computer is located and invoked by clicking on its icon, which loads the program from hard disk into RAM. The word-processing program is executed from RAM, allowing the user to type and correct the report (while periodically saving the unfinished report to hard disk). When the computer is turned off, the contents of the RAM is lost—so if the report was not saved to permanent memory, it is lost forever. Since software resides in RAM during execution, the more memory, the more things one is able to do. Also—equivalently—since RAM is the temporary storage area where the computer “thinks,” it usually is advantageous to have as much RAM memory as possible. Too little RAM can cause the software to run frustratingly slow and the computer to freeze if not enough memory is available for temporary storage as the software program executes. Laptops nowadays require at least 4 gigabytes (GB) of RAM and for better performance 8 or even 16 gigabytes of RAM. Typical access times for RAM are under 1 ns. If a CPU specifies 1 ns memory, it can usually work with faster chips. If a slower memory chip is used without additional circuitry to make the processor wait, the processor will not receive proper instruction and data bytes and will therefore not work properly.

In the 1980s capacities of RAMs and ROMs were 1 M × 1 bit (1-megabit chip) and 16 K × 8-bit, respectively, and in the mid-1990s 64 M × 1-bit chips became available. Memory arrays are constructed out of such chips and are used to develop different word-width memories; for example, 64 MB of memory would use eight 64 M × 1-bit chips on a single plug-in board. A popular memory size is 16 MB, consisting of eight 16-megabit chips. (Composite RAM, which has too many chips on a memory board, tends to be less reliable. For example, a 16 MB of composite RAM might consist of 32, 4-megabit chips, while an arrangement with eight, 16-megabit chips would be preferable.) The size of memory word width has increased over the years from 8 to 16, 32, and now 64 bits in order to work with advanced CPUs which can process larger words at a time. The more bits a processor can handle at one time, the faster it can work; in other words, the inherent inefficiencies of the binary system can be overcome by raw processing power. That is why newer computers use at least 32-bit processors, not 16-bit processors. And by processing 32 bits at a time, the computer can handle more complex tasks than it can when processing 16 bits at a time. A 32-bit number can have a value between 0 and 4,294,967,295. Compare that to a 16-bit number's range of 0–65,535, and one sees why calculations that involve lots of data—everything from tabulating a national census count to modeling flow over an airplane wing or displaying the millions of color pixels (points of light) in a realistic image on a large screen—need 32-bit processors and are even more efficient with 64-bit processors. A simple 16 × 8-bit memory array is shown in Fig. 8.5.

10.1 Quantum Random Access Memory

A random access memory allows memory cells to be addressed in a classical computer: it is an array in which each cell of the array has a unique numerical address. A QRAM serves a similar purpose (Giovannetti et al., 2008).

A random access memory has an input register to address the cell in the array, and an output register to return the stored information. In a QRAM, the address and output registers are composed of qubits. The address register contains a superposition of addresses ∑jpj|j 〉a, and the output register will contain a superposition of information, correlated with the address register: ∑jpj|j〉a|Dj〉d.

Using a “bucket-brigade” architecture, a QRAM reduces the complexity of retrieving an item to O(log 2n) switches, where n is the number of qubits in the address register.

The core idea of the architecture is to have qutrits instead of qubits allocated in each node of a bifurcation graph (Figure 10.1). A qutrit is a three-level quantum system. Let us label the three levels |wait〉, |left〉, and |right〉. During each memory call, each qutrit is in the |wait〉 state.

The qubits of the address register are sent through the graph one by one. The |wait〉 state is transformed into |left〉 and |right〉, depending on the current qubit. If the state is not in |wait〉, it routes the current qubit. The result is a superposition of routes. Once the routes have thus been carved out, a bus qubit is sent through to interact with the memory cells at the end of the routes. Then, it is sent back to write the result to the output register. Finally, a reverse evolution on the states is performed to reset all of them to | wait〉.

The advantage of the bucket-brigade approach is the low number of qutrits involved in the retrieval: in each route of the final superposition, only log N qutrits are not in the | wait〉 state. The average fidelity of the final state if all qutrits are involved in the superposition is O(1 − ϵ log N) (Giovannetti et al., 2008).