|
Echo #02
30 ноября 1996 |
|
PENTIUM - Pentium Processor - Technical review.
A good idea would have been and PC-shkam something to print?
Look amizhniki: the Pentium also has its own beauty. Not for
nothing that it Now the whole world considers the best
processor. Question misery PC-nis is inferiority rather
programmers, and machine architecture does not so bad. There
are no bad cars, there are bad programmers ...
TM
P R O C E S S O R
At Echo is published with the reduction
(M) (C) I N T E L '9 2
Technical Overview
(C) 1993. Alexander Kolpakov.
Compiled, translated.
- Advanced microprocessor technology firms INTEL -
Achieve the company INTEL in the art of designing and
manufacturing of semiconductors makes it possible to produce
more powerful microprocessors in increasingly small packages.
Developers microprocessors are currently working with a
complementary technological processes of metal oxide
semiconductor (CMOS) with a resolution less than a micron.
The use of submicron technology allows developers INTEL
firms have more transistors on each substrate. This made it
possible to increase the number of transistors for the family
of X86 from 29.000 the 8086 processor to 1.2 million in CPU
Intel486 DX2, the highest achievement in the Pentium processor.
Performed on 0.8 micron BiCMOS technology, it contains 3.1
million transistors. BiCMOS technology combines advantages of
two technologies: bipolar (speed) and CMOS (low power). With
more than two times more transistors Pentium processor as
compared to Intel486, developers have placed on the substrate
components, previously placed outside the processor. The
presence of components within reduces the access time, which
significantly improves performance. 0.8 micron technology from
INTEL uses three-layer metal and has a level higher than
original 1.0 micron double layer metal technology used in the
processor Intel486.
INTEL firm used the most recent advances in design
technology of microprocessors to achieve the benefits that are
comparable with alternative architectures used in scientific
and engineering workstations, while ensuring compatibility with
software costing $ 50 billion accumulated for a family of
microprocessors Series X86.
And the software itself for the Pentium processor
developed the new technology. More at the design stage
hardware processor design to become involved
Experts from all the major companies developing operational
systems and compilers - Microsoft, IBM, NeXT, Borland, Watcom,
MetaWare, etc. This enabled in hardware to support new
programming technology based branding providers of standard
software. On the other hand, even before the birth of a new
processor used methods Classical and spetsilnoy optimization
reveal specific advantages architecture X86, for example, the
use of commands Download recording, powerful regimes
adressatsii, removal of invariant sections of code of cycles,
etc. Now, just by recompiling the legacy applications can
improve their performance on the new processor has doubled.
This is Time can not offer none of the competitors company
INTEL.
- PENTIUM Processor -
- Technical Overview -
The new processor "Pentium" INTEL firm combines the
advantages inherent in traditional minicomputers and
workstations with the flexibility and compatibility, which are
characterized by platform personlnyh computers.
Designed for the needs of an increasingly complex
association Contemporary and buduyuschego application software
Pentium processor extends the range of microprocessor-based
architecture firm INTEL to new heights, obscures the
differences between the previously powerful computing
platforms, and created for an entirely new field of
applications desktops and servers.
- A new generation of processors the company INTEL -
Combining more than 3.1 million transistors on a single
silicon substrate, 32-bit Pentium processor is characterized by
high productivity with a frequency of 60 and 66 MHz. Its
superscalar architecture uses advanced engineering methods,
which allow you to perform more than one instruction per clock
cycle, resulting in What Pentium able to perform a huge number
of PC-compatible software faster than any other microprocessor.
Furthermore suschestvuyushih developments of software,
high-performance arithmetic unit with a floating comma Pentium
proessora provides increased computing power up required to use
previously unavailable technical and scientific applications,
originally designed for workstation platforms. As well as local
and global networks continue to displace older hierarchical
network, managed large computers, the advantages and
flexibility of multiprocessor operating system Pentium
processor - ideal for a host computer for modern client /
server applications used in the industry.
Because the Pentium processor can reach a level of
performance equal to or higher than current high-level
workstations, it has preimuschestvmi, which denied conventional
workstations: full compatibility with more than 50,000 software
applications with the cost billions dollars that were written
under ahitekturu company INTEL. In addition, Pentium processor
allows the use of all major operating systems that are
available today's desktop PCs, workstations and servers,
including UNIX, Windows-NT, OS / 2, Solaris, and NEXTstep.
- Pentium processor. Technological innovations -
Numerous innovations - the hallmark of
Pentium processor in the form of a unique combination of high
performance, interoperability, data integration and
scalability. This includes:
- Superscalar architecture;
- Separate caching code and data;
- Block predicting the correct address of the transition;
- High performance block floating-point calculations;
- Advanced 64-bit data bus;
- Multi-mode;
- Means set the size of memory pages;
- Means of detecting errors and functional redundancy;
- Performance Management;
- Scalability with Intel OverDrive processor.
- Superscalar architecture -
Superscalar architecture Pentium processor is
is only compatible with INTEL dvuhkonveernuyu industrial
architecture that enables the processor to achieve new levels
of productivity by performing more than one team for one clock
cycle. The term "superscalar" means a microprocessor
architecture, which contains more than one computing unit.
These computing units, or conveyors, are the nodes where there
are all the basic processes of data and commands.
The emergence of a superscalar architecture Pentium
processor represents a natural evolution of previous family
processors with 32-bit architecture firm INTEL. For example,
Intel486 processor can perform several of their teams for
one clock cycle, but the previous processor family firm INTEL
require many clock cycles to perform a single command.
Ability to execute multiple commands in one clock cycle
exists due to the fact that the Pentium processor has two
conveyors, which can perform two instructions simultaneously.
Just as with the Intel486 one pipelines, double conveyor
Pentium processor performs a simple command for five phases:
preconditioning, the first decoding (decoding command), the
second decoding (generation of address) execution and reverse
discharge. This allows multiple teams are in various stages of
implementation, thereby increasing the processing power. Each
conveyor has a arithmetic and logic unit (ALU), a set
generating device address and interface caching. So
well as the CPU Intel486, Pentium processor uses
hardware-execution of commands, replacing the set of
microinstructions used in the previous families of
microprocessors. These instructions include downloading,
storing, and simple operation ALU, which can run the hardware
processor, without the use of microcode. This improves
performance without affecting compatibility. In the case of
more complex commands for the further acceleration of
productivity performance advanced Pentium processor microcode
to execute the commands you use both conveyors superscalar
architecture.
As a result of these architectural innovations, compared
with previous microprocessors, is much more
commands can be executed for the same time.
- Separate caching code and data -
Another important revolutionary improvements introduced in
the Pentium processor, is the introduction of a separate cache.
Caching improves performance by increasing the temporary
storage place for frequently used program code and data from
fast memory, replacing the possible appeal to the external
system memory for some commands. Processor Intel486, for
example, contains one 8-KB unit on-chip cache used to cache
both code and data.
The designers of the company INTEL bypassed this
limitation using an additional circuit, performed by 3.1
million transistors Pentium processor (for comparison, Intel486
contains 1.2 million transistors) to create a split inside
caching code and data. This improves performance by eliminating
the conflicts on the bus and makes Dual Caching available more
frequently than was previously possible. For example, during
the phase of preconditioning is used command code received from
the instruction cache. In the case of one block cache, a
conflict between the process of preliminary training commands
and data access. Performing separate cache for instructions and
data excludes such conflicts, allowing both teams to run
simultaneously. Cache memory code and data Pentium processor
contains information on 8 KB each, and each is organized as a
set of dual-associative cache - designed to record only
previewed specified 32-byte segments, and faster than the
external cache. All these especially the expansion of
performance required for 64-bit internal data bus, which
provides the possibility of double caching and superscalar
processing conveyer simultaneously download the data. Data
cache has two interfaces, one for each of the pipelines that
allows him to provide two data odelnye instructions within one
machine cycle. Once the data gets the from the cache, they are
written into the main memory in writeback mode. This technique
gives better cache performance than simply caching with direct
entry, in which processor writes data simultaneously to the
cache and main memory. However menne, Pentium processor is
capable of dynamically configured to support caching with
direct entry.
Thus, data caching uses two different
great solution: write-back cache and an algorithm, called the
MESI (modified, exclusion, distribution, release) protocol.
Cache Write-Back allows you to record in the cache without
reference to main memory, in contrast to previously used direct
simple caching. These solutions increase proizvoditlnost
through the use of a converted bus and a warning exception of
the bottleneck in the system. In turn MESI-protocol allows the
data in the cache memory and external memory match - a great
solution for improved multiprocessor systems, where different
processors can used for the same data.
The recommended amount of total cache for desktop systems,
based on the Pentium processor is equal to 128-256 K, and for
servers - 256 K and above.
- Block predicting the correct address of the transition
-
Block prediction correct address transition - is the next
great solution for computing, which increases productivity
through complete filling pipelines teams based on
pre-determining the correct set of commands that must be met.
Pentium processor - this is the first and only PC-compatible
processor that uses block prediction, which until then had been
traditionally associated with mainframe computing platforms.
To better understand this concept, consider a typical
software application. After each program
cycle, the program performs the appropriate checks to determine
whether to return to the top of the loop or out and continue
with the next step. These two solutions, or paths, called
prediction addresses the transition. Block prediction correct
address transition predicts which branch of the program is
needed, based on the assumption that the previous branch that
has been passed, will be used again. Pentium processor performs
the prediction of the correct addresses of the transition,
using a special branch prediction buffer (BTB). Unlike
alternative architectures, a software and formulaic innovation
makes it possible to recompile the code, increasing the speed
and performance existing application software. If a team
manages the branching program, the buffer BTB stores the
command and address to which you want to go, and predicts what
branch instructions in the next moment will be used. When the
buffer contains a correct prediction, the transition is
executed without delay.
- High performance block floating-point calculations -
Growing wave of 32-bit software applications,
including a lot of intensive computing, graphically based, in
programs that take a lot of CPU resources execution of floating
point operations, providing the mathematical calculations.
Since the requirements for personal computers by the software
floating-point calculations are constantly rising, those needs
can be improvements in microprocessor technology. Processor
Intel486 DX, for example, was the first microprocessor
integrated on one substrate with a math coprocessor. The
previous processor family firm INTEL, if necessary, the use of
floating-point calculations used external math coprocessor.
Pentium processor allows you to perform mathematical
calculations at a higher level through the use of advanced
built-in block floating-point calculations, which includes
vosmitaktovy conveyor and the firmware of the basic
mathematical functions. Four-stroke Conveyor team floating
point complement four-stroke integer konveerizatsmyu. Most of
the teams calculations Floating-point can be performed in an
integer pipelines, and then served in the conveyor
floating-point calculations. Normal functions of floating-point
calculations, such as addition, multiplication and division are
implemented in hardware to accelerate computation.
As a result of these innovations, Pentium processor
performs team floating point five times faster than
33-MHz Intel486 DX, optimizing them for high-speed numerical
computations, which are an integral part of such advanced video
applications, such as CAD and 3D-graphics.
Pentium processor at 66 MHz operates as a
"Chislodrobilka" with a rating of 64.5 on the test SPECint92,
almost not yielding to RISC-processor Alpha Company Digital,
but with the clock frequency twice as high.
Overall performance Pentium processor exceeds 6
times 25 MHz Intel486 SX and 2.6 times - 66 MHz Intel486 DX2.
Index iCOMP rated for 66 MHz Pentium processor, which executes
112 million instructions per second is 567. Index on iCOMP
(Intel COmparative Microprocessor Peformance) performs relative
performance comparison of 32-bit processors the company INTEL.
- Extended 64-bit data bus -
Pentium processor outside a 32-bit
device. External data bus to memory is 64-bit,
doubling the amount of data transferred during one bus cycle.
Pentium processor supports several types of bus cycles,
including batch mode, during which the portion of data from 256
bits to the data cache and, within one bus cycle.
The data bus is the main thoroughfare that passes
information between the processor and memory subsystem. Thanks
This 64-bit data bus, Pentium processor significantly increases
the transmission rate compared with an Intel486 DX - 528 MB /
sec to 66 MHz, compared to 160 MB / sec to 50 MHz CPU Intel486
DX. This extended data bus facilitates high-speed computing
thanks to the support of simultaneous make-up commands and data
processing unit superscalar calculations, which results in even
greater overall performance Pentium processor as compared with
an Intel486 DX.
In general, having a wider data bus, Pentium processor
provides konveerizatsiyu bus cycles, thus increasing bandwidth.
Konveerizatsiya bus cycles allows the second cycle start ranshe
completion the first cycle. This allows the memory subsystem
more time to decode the address, which allows you to use slower
and less expensive memory components, reducing the resulting
total system cost. Accelerated reading and writing,
parallelilizm address and data, and decoding in a single cycle
- together, improves throughput and increases the capabilities
of the system.
- Multiprocessing -
Pentium processor - it's ideal for narostayuschey wave
multiprocessor systems, as well as the highest level of
performance and computing power of modern computing facilities.
Multiprocessor applications, which combine two or more Pentium
processors - is well served by through advanced architecture of
the crystals, separate built-in caching code and data, and well
as chipsets for the management of external cache memory, and
sophisticated means of data integrity.
As discussed above, Pentium processor supports structured
cache with its MESI protocol. When one processor accesses data
that is cached in a different processor, it is able to receive
correct data. And if the data is modified, all processors are
able to access to receive data in a modified form. The latest
Pentium processor INTEL firm also determines which commands are
recognized by the system according to ipolzuemym way of
programming. This strictly definitely tells how the software
software developed for single-processor system to work
correctly in a multiprocessor environment.
- Means of shared memory on the page -
Pentium processor offers the option to support any of the
traditional page size - 4 KB or larger, 4 MB page. This option
allows you to calculate the frequency swap pages of complex
graphics applications, the buffers frames, as well as the
kernels of operating systems, where the increased size of the
page now allows users to reschedule the wider originally bulky
objects. Increased page gives a result in the form of increased
productivity, all of which are reflected in the application
software.
- Identify errors and functional redundancy -
Good data protection and ensure their integrity through
internal resources becomes extremely important in applications,
critical data loss due to the spread of modern client-server
environment. Pentium processor contains two improvements that
traditionally inherent in the design Class Mainframe - internal
error detection and control by functional redundancy (FCR) - it
helps ensure the integrity of the data emerging today based
systems on desktop computers.
Internal error detection adds parity
internal code and data caching, the shear associative
page table, microcode, and the destination buffer the
transition, helping to identify errors in such a way that it
remains invisible and for the user, and for the system. At the
same time, control by functional redundancy is optimized for
applications Critical data loss, where the Pentium processor to
work in the basic configuration / monitoring. If two processor
detects a disagreement, the system shall be notified error. As
a result, the discovery of more than 99% errors.
In addition, the substrate processing device is located
built-in testing. Self-test covers more than 70%
nodes Pentium processor, does not require a crystal clear
and is a procedure commonly used in diagnosis systems. Other
embedded solutions is the implementation of standard, IEEE
1149.1, which allows to test the external connections proessora
and debug mode, enabling the software to view the registers and
processor status.
- Performance Management -
Performance Management - a feature Pentium processor that
allows system developers and application extensions to optimize
their hardware and software by identifying potential
bottlenecks for software code. and rabotchiki can watch and
take measures for the domestic Event processor, such as the
read and write data caching matches and deletions, interrupt
and bus utilization. This allows them to measure the
effectiveness of which has a code in a double Pentium processor
architecture and its products and to perform fine-tune their
applications and systems to achieve optimum performance.
Benefit to end users - a higher esteem and higher
performance, and all this thanks to a good interaction
with a Pentium processor, a custom system and application
software.
Enabling developers to design systems with power
management, protection, and other properties Pentium processor
supports system management mode (SMM), such a regime
architecture Intel SL.
- Expandability -
Along with all that is done for the new 32-bit
microprocessor architecture firm INTEL, Pentium processor is
designed for easy expandability with the use of architecture
building company INTEL. These innovations will protect your
investment users by increasing productivity, which helps to
maintain the level of productivity-based systems, processor
architecture firm INTEL, more than the life expectancy of the
individual components. Technology building makes it possible to
take advantage of most processors improved technologies into
existing systems with a simple installation means a
single-crystal scale performance. For example, the first means
of capacity - it OverDrive processor, designed for processors
Intel486 SX and Intel486 DX, which uses technology simply
doubling the clock frequency used in designing microprocessors
Intel486 DX2.
By building one of these additional processors in the
socket, located near the central microprocessor on most
motherboards Intel486, users can increase overall system
performance by more than 70% for virtually all software
applications.
Technology building by OverDrive processors possible for
systems based on the Pentium processor through a simple setup
in the future, the processor performed on advanced technology.
In turn, technology Pentium processor is the basis of
additional processor, developed for systems based on the
Intel486 DX2.
At present, Pentium processor enables superior performance
at very reasonable price, fully maintaining compatibility with
previous microprocessors, family X86.
Other articles:
Similar articles:
В этот день... 27 April