Central processor (cpu) — the electronic block or a microcircuit — the executor of machine instructions (a code of programs), a body of hardware maintenance of the computer or the programmed logic controler. Sometimes name micro processor (cpu) or it is simple processor (cpu).
Initially term central processor (cpu) the device described a specialized class of the logic cars intended for performance of difficult computer programs. Owing to enough exact conformity of this appointment to functions existing at that time computer processor (cpu), it has been naturally transferred on computers. The beginning of application of the term and its abbreviation in relation to computer systems was necessary in 1960th years. The device, architecture and realization processor (cpu) repeatedly changed since then, however their basic executed functions remained the same, as before.
The main characteristics ЦПУ are: clock frequency, productivity, power consumption, norms of lithographic process used by manufacture (for processor (cpu)) and architecture.
Early CENTRAL PROCESSING UNITS were created in the form of unique components for unique, and even unique, computer systems. Later from an expensive way of working out processor (cpu), intended for performance unique or several highly specialized programs, manufacturers of computers have passed to serial manufacturing of typical classes multi-purpose processor (cpu) devices. The tendency to standardization of computer accessories has arisen during an epoch of rapid development of semi-conductor elements, main frames and minicomputers, and with the advent of integrated schemes it became even more popular. Creation of microcircuits has allowed to increase even more complexity of the CENTRAL PROCESSING UNIT with simultaneous reduction of their physical sizes. Standardization and miniaturization processor (cpu) have led to deep penetration of the digital devices based on them into an everyday life of the person. Modern processors (cpu) it is possible to find not only in such hi-tech devices, as computers, but also in cars, calculators, mobile phones and even in children's toys. More often they are presented by microcontrollers where besides the computer on a crystal additional components (memory of programs and the data, interfaces, input/conclusion ports, timers are located, etc.). Modern computing possibilities of the microcontroller are comparable with processors (cpu) personal COMPUTERS of ten years' prescription, and more often even considerably surpass their indicators.
Prospects
The next 10-20 years the equipment processors (cpu), most likely, will change in view of that technological process will reach physical limits of manufacture. Probably, it will be:
Optical computers — in which instead of electric signals are exposed to processing light streams (photons, instead of электроны).
The quantum computers, which work entirely is based on quantum effects. Now works on creation of working versions quantum processors (cpu) are conducted.
Molecular computers — the computing systems using computing possibilities of molecules (mainly, organic). Molecular computers use idea of computing possibilities of an arrangement of atoms in space.
Architecture Neumann's background
The majority modern processors (cpu) for personal computers in general are based on this or that version of cyclic process of serial processing of the data, invented [a source?] John a background Neumann.
J. A background Neumann has thought up [a source?] the scheme of construction of the computer in 1946. Distinctive feature of architecture Neumann's background is that instructions and the data is stored in the same memory.
In various architecture and for various commands additional stages can be demanded. For example, for arithmetic commands additional references to memory during which time reading of operands and record of results is made can be demanded.
Stages of a cycle of performance:
- Processor (cpu) exposes the number stored in the register of the counter of commands, on the tire of the address and gives memories a reading command.
- The exposed number is the address for memory; memory, having received the address and a reading command, exposes the contents stored to this address, on the tire of the data and informs on readiness.
- Processor (cpu) receives number from the tire of the data, interprets it as a command (the machine instruction) from the system of commands and executes it.
- If last command isn't a transition command, processor (cpu) increases by unit (in the assumption that the length of each command is equal to unit) the number stored in the counter of commands; as a result there the address of the following team is made.
The given cycle is carried out invariably, and it is called as process (whence there was a device name).
During process processor (cpu) reads out sequence of the commands containing in memory, and executes them. Such sequence of commands is called as the program and represents algorithm of work processor (cpu). The sequence of reading of commands changes in case processor (cpu) reads out a transition command, — then the address of the following command can appear another. As other example of change of process the case of reception of a command a stop or switching in a mode of processing of interruption can serve.
Commands central processor (cpu) are the lowermost level of management of the computer, therefore performance of each command inevitably and certainly. It is not made any check on an admissibility of carried out actions possible loss of the valuable data, in particular, isn't checked. That the computer carried out only admissible actions, commands should be organized in appropriate way in the form of the necessary program.
Speed of transition from one stage of a cycle to another is defined by the clock generator. The clock generator develops the impulses serving by a rhythm for central processor (cpu). Frequency of clock impulses is called as clock frequency.
Conveyor architecture
Ability of performance of several machine instructions for one step processor (cpu) by increase in number of actuation mechanisms. Occurrence of this technology has led to essential increase in productivity, at the same time there is a certain limit of growth of number of actuation mechanisms at which excess productivity practically ceases to grow, and actuation mechanisms stand idle. The partial decision of this problem are, for example, technology Hyper Threading.
CISC-processors (cpu)
Complex instruction set computer — calculations with a difficult set of commands. Processor (cpu) ная the architecture based on the complicated set of commands. Typical representatives CISC are микро processors (cpu) families x86 (though many years these processors (cpu) are CISC only on external system of commands: in the beginning of process of execution difficult commands break into more simple microoperations (МОП’ы), executed by a RISC-kernel).
RISC-processors (cpu)
Reduced instruction set computer — calculations with the simplified set of commands (in the literature a word «reduced» quite often wrongly translate as "reduced"). The architecture processors (cpu), constructed on the basis of the simplified set of commands, is characterized by presence of commands of the fixed length, a considerable quantity of registers, operations of type the register-register, and also absence of indirect addressing. Concept RISC is developed by John Kokom (John Cocke) from IBM Research, the name is thought up by David Pattersonom (David Patterson).
Simplification of a set of commands urged to reduce the conveyor that allows to avoid delays on operations of conditional and unconditional transitions. The homogeneous set of registers simplifies compiler work by optimization of an executed program code. Besides, RISC-processors (cpu) differ smaller power consumption and a thermal emission.
Among the first realizations of this architecture were processors (cpu) MIPS, PowerPC, SPARC, Alpha, PA-RISC. In mobile devices are widely used ARM-processors (cpu).
MISC-processors (cpu)
Minimum instruction set computer — calculations with the minimum set of commands. The further development of ideas of a command of Chaka Moore which believes that the simplicity principle, primary for RISC-processors (cpu), has too quickly receded into the background. In the heat of struggle for the maximum speed, RISC has caught up and has overtaken many CISC processors (cpu) on complexity. Architecture MISC is under construction on push-down computing model with the limited number of commands (approximately 20-30 commands).
VLIW-processors (cpu)
Very long instruction word — a superlong command word. Architecture processors (cpu) with obviously expressed parallelism of calculations put in system of commands processor (cpu). Are a basis for architecture EPIC. Key difference from superscalar CISC-processors (cpu) is that for them the part processor (cpu) (scheduler) is engaged in loading of actuation mechanisms, on what small enough time while for VLIW-processor (cpu) the compiler is engaged in loading of computers is taken away, on what is taken away essentially more time (quality of loading and, accordingly, productivity theoretically should be above). Example VLIW-processor (cpu) is Intel Itanium.
Multinuclear processors (cpu)
Contain a little processor (cpu) kernels in one case (on one or several crystals).
Processors (cpu), intended for work of one copy of an operating system on several kernels, represent высокоинтегрированную realization мультиprocessor (cpu) ности.
To the first multinuclear micro processor (cpu) became POWER4 from IBM, appeared in 2001 and having two kernels.
In October 2004 Sun Microsystems has let out binuclear processor (cpu) UltraSPARC IV which consisted of two modified kernels UltraSPARC III. In the beginning 2005 has been created binuclear UltraSPARC IV +.
On November, 14th, 2005 Sun has let out eight-nuclear UltraSPARC T1 which each kernel carried out 4 streams.
On January, 5th, 2006 Intel has presented the first binuclear processor (cpu) on one cristal Core Duo, for a mobile platform.
In November, 2006 there was the first tetracyclic processor (cpu) Intel Core 2 Quad on kernel Kentsfield, representing assemblage two crystals Conroe
On September, 10th, 2007 have been put on the market нативные (in the form of one crystal) tetracyclic processors (cpu) for servers AMD Opteron, having in the course of working out code name AMD Opteron Barcelona. [2] on November, 19th, 2007 left in sale tetracyclic processor (cpu) for house computers AMD Phenom [3]. These processors (cpu) realize new microarchitecture K8L (K10).
Company AMD has gone on own way, producing tetracyclic processors (cpu) a uniform crystal (unlike Intel, the first tetracyclic processors (cpu) which represent actually склейку two binuclear crystals). Despite all progressiveness of the similar approach the first «четырёхъядерник» the firms, received name AMD Phenom X4, has turned out not too successful. Its backlog from modern to it processors (cpu) the competitor made from 5 to 30 and more percent depending on model and specific targets [a source isn't specified 678 days].
By 1-2 quarter 2009 both companies have updated the rulers tetracyclic processors (cpu). Intel has presented family Core i7, consisting of three models working on different frequencies. The basic highlights given processor (cpu) is use of the three-channel controler of memory (type DDR-3) and technology эмулирования eight kernels (it is useful for some specific problems). Besides, thanks to the general optimization of architecture it was possible to raise considerably productivity processor (cpu) in many types of problems. Weakness of the platform using Core i7, its excessive cost as expensive motherboard is necessary for installation given processor (cpu) on chipset Intel X58 and a three-channel set of memory of type DDR3, also having at present high cost is.
Company AMD has in turn presented a ruler processors (cpu) Phenom II X4. By its working out the company has considered the errors: the cache volume has been increased (obviously insufficient at first "Fenoma"), is made on 45-nanometer to technical process (it, accordingly, has allowed to lower a thermal emission and considerably to raise working frequencies). As a whole, AMD Phenom II X4 on productivity costs level with processor (cpu) ми Intel the previous generation (kernel Yorkfield) and rather considerably the source lags behind from Intel Core i7 [isn't specified 687 days]. However, with an exit processor (cpu) AMD Phenom II X6 Black Thuban 1090T a situation has considerably changed in favor of AMD. This processor (cpu) at the price costs at level Core i7 930, however can contend with a ruler processors (cpu) Core i7 in respect of productivity. Its high-grade 6 kernels perfectly approach for difficult multiline problems.
At present are in large quantities accessible two - four - and six - nuclear processors (cpu) (Intel Core 2 Duo on 65-nanometer kernel Conroe
Caching
The caching is use of additional high-speed memory (a cache, English cache) for storage of copies of blocks of the information from the basic (operative) memory, probability of the reference to which in the near future a bycicle.
Distinguish caches 1, 2 and 3rd levels (are designated L1, L2 and L3 — from Level 1, Level 2 and Level 3). The cache of 1st level has the least латентность (access time), but the small size, except that, caches of the first level often become the multiport. So, processors (cpu) AMD K8 were able to make simultaneously 64-bit record and reading, or two 64-bit readings for a step, AMD K8L can make two 128-bit readings or records in any combination. Processors (cpu) Intel Core 2 can make 128-bit record and reading for a step. The cache of 2nd level usually has considerably big латентность access, but it can be made much more on the size. A cache of 3rd level the biggest on volume and slow enough, but nevertheless it much faster, than operative memory.
The Harward architecture
The Harward architecture differs from architecture Neumann's background that the program code and the data are stored in different memory. In such architecture many methods of programming are impossible (for example, the program can't change the code during performance; it is impossible to redistribute dynamically memory between a program code and the data); but the Harward architecture allows to perform more effectively work in case of the limited resources, therefore it is often applied in built in systems.
Parallel architecture
The architecture Neumann's background possesses that lack that it consecutive. Whatever huge data file was required to be processed, its each byte should pass through central processor (cpu) even if over all bytes it is required to perform the same operation. This effect is called as a narrow neck Neumann's background.
For overcoming of this lack architecture processors (cpu) which are called as parallel were offered and offered. Parallel processors (cpu) are used in supercomputers.
As possible variants of parallel architecture can serve (on Flynn's classification):
- SISD — one stream of the commands, one data flow;
- SIMD — one stream of commands, is a lot of data flows;
- MISD — it is a lot of streams of the commands, one data flow;
- MIMD — it is a lot of streams of commands, it is a lot of data flows.
Digital alarm processors (cpu)
For digital processing of signals, especially at limited time of processing, apply specialized high-efficiency alarm микро processors (cpu) (DSP) with parallel architecture.
Manufacturing process
Originally before developers the technical project proceeding from which the decision on what architecture of the future processor (cpu), its internal device, a manufacturing techniques will be is made is put. Before various groups the problem of working out of corresponding functional blocks processor (cpu), maintenance of their interaction, electromagnetic compatibility is put. Because processor (cpu) actually is the digital automatic machine completely answering to principles булевой of algebra, by means of the specialized software working on other computer, the virtual model of the future processor (cpu) is under construction. On it testing processor (cpu) is held, execution of elementary commands, considerable volumes of a code, is fulfilled interaction of various blocks of the device, optimization is conducted, are searched inevitable at the project of such level of an error.
After that from digital base matrix crystals and the microcircuits containing elementary functional blocks of digital electronics, the physical model processor (cpu) on which electric and time characteristics processor (cpu) are checked is under construction, the architecture processor (cpu) is tested, correction of the found errors proceeds, questions of electromagnetic compatibility (for example, at almost ordinary clock frequency in 10 GHz pieces of a conductor of 7 mm already work as radiating or accepting aerials) are specified.
Then the stage of teamwork of engineers-shemotehnikov and process engineers which by means of the specialized software will transform the electric scheme containing architecture processor (cpu), to crystal topology begins. Modern systems of automatic designing allow, generally, from the electric scheme directly to receive a package of cliches for creation of masks. At this stage technologists try to realize the technical decisions put схемотехниками, taking into account available technology. This stage is one of the long and difficult in working out and sometimes demands compromises from outside схемотехников on refusal of some architectural decisions. It is necessary to notice that a number of manufacturers of custom-made microcircuits (foundry) offers developers (to the design center or fabless) the conciliatory proposal at which at a designing stage processor (cpu) the standardized libraries of elements presented by them according to available technology and blocks (Standard cell are used). It enters a number of restrictions on architectural decisions, but the stage of technological adjustment is actually reduced to game in the designer of "Lego". Generally, made under individual projects микро processors (cpu) are faster in comparison with processor (cpu) ми, created on the basis of available libraries.
The following, after a design stage, creation of a prototype of a crystal micro processor (cpu) is. At manufacturing of modern superbig integrated schemes the lithograph method is used. Thus, on a future substrate micro processor (cpu) (a thin circle from single-crystal silicon, or sapphire) through the special masks containing cuts, layers of conductors, insulators and semiconductors are serially put. Corresponding substances evaporate in vacuum and are besieged through mask apertures on a crystal processor (cpu). Etching when the aggressive liquid corrodes the sites of a crystal not protected by a mask is sometimes used. Simultaneously on a substrate it is formed a hundred order processor (cpu) crystals. As a result there is the difficult multilayered structure containing from hundreds thousand to billions of transistors. Depending on connection the transistor works in a microcircuit as the transistor, the resistor, the diode or the condenser. Creation of these elements on a microcircuit separately, generally, not favourably. After the termination of procedure of lithograph the substrate is sawn on elementary crystals. The thin gold conductors which are adapters to contact platforms of the case of a microcircuit are soldered to the contact platforms generated on them (from gold). Further the heat-conducting path of a crystal and a microcircuit cover, generally, fastens.
Then the stage of testing of a prototype processor (cpu) when its conformity to the set characteristics is checked begins, errors are searched remained not noticed. Only after that micro processor (cpu) it is started in manufacture. But even during manufacture there is a constant optimization processor (cpu), connected with the technology perfection, new design decisions, detection of errors.
It is necessary to notice that in parallel with working out universal микро processors (cpu), sets of peripheral schemes of the COMPUTER which will be used with micro processor (cpu) and on which basis motherboards are created are developed. Working out микро processors (cpu) a set (a chipset, English chipset) represents a problem, not less difficult, than creation actually microcircuits micro processor (cpu).
In the last some years were outlined a tendency of carrying over of a part of components of a chipset (the memory controler, tire PCI Express controler) in structure processor (cpu).
Power consumption processors (cpu)
Its power consumption is closely connected with manufacturing techniques processor (cpu) also. The first processors (cpu) architecture x86 consumed scanty (to modern measures) the quantity of energy making shares of watt. The increase in quantity of transistors and increase of clock frequency processors (cpu) has led to essential growth of the given parameter. The most productive models demand to 130 and more watt. Insignificant at the beginning the power consumption factor, now makes serious impact on evolution processors (cpu):
- Perfection of the production technology for consumption reduction, search of new materials for decrease in currents of leak, fall of pressure of a food of a kernel processor (cpu);
- Occurrence of sockets (sockets for processors (cpu)) with a great number of contacts (more than 1000) which majority is intended for a food processor (cpu). So at processors (cpu) for popular socket LGA775 the number of contacts of the basic food makes 464 pieces (about 60 % from total);
- Configuration change processors (cpu). The crystal processor (cpu) has moved with internal on outer side, for the best heat removal to a radiator of system of cooling;
- Integration into a crystal of temperature gages and system of protection against an overheat, reducing frequency processor (cpu) or in general stopping it at inadmissible increase in temperature;
- Occurrence in the newest processor (cpu) х the intellectual systems dynamically changing pressure of a food, frequency of separate blocks and kernels processor (cpu), both disconnecting not used blocks and kernels;
- Occurrence of power saving up modes for «sleeping» processor (cpu), at low loading.
Manufacturers
The most popular processors (cpu) make today firms Intel, AMD and IBM. The majority processors (cpu), now in use, are Intel-compatible, that is have a set of instructions and the interfaces of programming used in processor (cpu) х of company Intel.
Among processors (cpu) from Intel: 8086, i286, i386, i486, Pentium, Pentium II, Pentium III, Celeron (simplified variant Pentium), Pentium 4, Core 2 Quad, Core i7, Xeon (a series processors (cpu) for servers), Itanium, Atom (a series processors (cpu) for the built in technics), etc. AMD has in the ruler processors (cpu) architecture x86 (analogs 80386 and 80486, family K6 and family K7 — Athlon, Duron, Sempron) and x86-64 (Athlon 64, Athlon 64 X2, Phenom, Opteron, etc.). Processors (cpu) IBM (POWER6, POWER7, Xenon, PowerPC) are used in supercomputers, in video prefixes of 7th generation, the built in technics; earlier were used in firm Apple computers. According to company IDC, following the results of 2009 in the market микро processors (cpu) for the personal computer, laptops and servers corporation Intel share has made 79,7 %, share AMD — 20,1 %.
No comments:
Post a Comment