NEWS
REVIEWS
VIDEO
HOW TO CHOOSE
iPhone SE
18+
Articles and news
Prices
Laptops and tablets
Phone numbers
Photos and videos
Audio
Iron
Peripherals
Networks
Games
Science and technology
Smart Home
Televisions
appliances
Fitness and health
Auto
E books
Computers
Applications
Sites of the day
All Obzoryzhelezovse reviews of computer hardware
Topics: INTEL processor AMD MOTOROLA NEC CPU history evolution DEC
The evolution of processors.
Part 3: 32 bit processors in the second half of the 1980s
15.07.2014
Oleg Kolenchenko, info@ferra.ru
Print version
The first half of the 1980s was marked by 16 bit processors, and two companies stood out among the manufacturers: Intel and Motorola.
Soon they focused on the release of 32 bit solutions, which this part of the series of articles is dedicated to.
Table of contents
Intel i386 Intel i486 Motorola 68020, 68030, 68040 NEC V60, V70, V80 DEC VAX Instead of conclusion
Previous parts:
Part 1: The 8 bit Era;
Part 2: The 16 bit era.
Intel i386
Three years after the release of Intel 80286, its follower, a crystal with the index 80386, saw the light.
"Three hundred and eighty sixth" became the first 32 bit processor of an American company.
Despite the fact that the Intel 80386 was still based on the x86 architecture and maintained backward compatibility with the Intel 8086 and 80286 processors, it underwent many changes.
According to some estimates, the x86 architecture did not receive such significant changes as in the case of the 80386 "stone" for many years.
Therefore, it is worth telling more about them.
As we have already said, the i386 processor has retained backward compatibility with its 8086 and 80286 predecessors.
That is, it can execute absolutely all programs written for previous processors, and it does it more efficiently.
Higher performance was achieved due to higher clock frequencies, as well as a smaller number of synchronization cycles when executing programs.
For example, the multiplication of two 16 bit numbers was performed in 9-22 clock cycles.
For comparison, the 80286 processor performed this operation in 21 clock cycles, and the 8086 crystal - in 118-133 clock cycles.
The advantage of the i386 was obvious!
In addition, the increased command pre selection buffer, the volume of which was 16 bytes, played a role.
Intel i386 processor
Image source
Of course, the main innovation of the i386 was that the processor became 32 bit.
The entire x86 architecture has been expanded to 32 bits.
The registers became 32 bit, and, of course, the processor received support for a set of 32 bit instructions.
Importantly, the protected mode of operation, which first appeared in 80286, was significantly improved.
The principle of operation of the protected mode remained the same, but the mode received three important innovations: removing the limit on the segment size, Page Addressing mode (Page Addressing) and Virtual 8086 mode (Virtual 8086 Mode).
In protected mode, the i386 used the same architecture with memory segments as in previous Intel solutions.
However, if earlier the maximum size of the memory segment was 64 KB, which for a long time did not suit programmers, now it has increased to 4 GB.
This greatly facilitated the development of 32 bit applications that could be executed without switching between different memory segments.
Also, in i386, it became possible to quickly switch between real and protected modes without simulating a CPU reboot.
As for the virtual 8086 mode, it does not represent anything special.
Interestingly, when creating the "three hundred and eighty sixth", a rather big mistake was made.
So, the processor incorrectly performed the multiplication operation of 32 bit numbers.
However, at the time of the chip's release, there were no 32 bit operating systems and applications yet, so the error was discovered only 18 months later — in April 1987.
All released but not sold processors were re labeled by Intel with the note "only for 16 bit operations".
However, the "stones" released after the error was detected were marked with a double symbol "sigma" (ΣΣ).
The i386 processor was released in many different versions, which differed in performance, form factors, power consumption and other characteristics.
The i386 was produced using CHMOS III technology, which combined the speed of HMOS technology and the low power consumption of CMOS technology.
At the same time, a 1.5 micron process was used, and the number of transistors was 275 thousand pieces.
i386DX processor (left)
Image source
The first i386 was introduced on October 13, 1985 and had a clock frequency of 16 MHz.
Subsequently, this modification of the" stone " received the prefix DX the model became known as the 386DX immediately after the launch of the cheaper 386SX in June 1988 The DX prefix stood for Double word eXternal, which emphasized the processor's support for a 32 bit external data bus.
The clock frequency of the 386DX has increased over the years.
So, in 1987, the frequency was increased to 20 MHz, in 1988 - to 25 MHz.
And in 1990, a modification with a frequency of 33 MHz went on sale.
At the same time, the processor's power consumption remained at a fairly low level — even lower than that of the "eighty sixth".
The 386DX was available in several cases: for example, in the PQFP 132 and in the ceramic PGA 132.
The main disadvantage of the 386DX was its high cost.
Intel wanted to increase the number of sales of a new generation of processors, and so soon the "stripped down" 386SX crystal saw the light.
The chip was released in 1988 and eventually became the most popular in the i386 line.
In its architecture, it was a complete analog of the DX version, with the exception of data buses and addresses.
So, instead of a 32 bit external data bus, a 16 bit one was used.
The bit depth of the external address bus was 24 bits.
At the same time, the processor itself remained completely 32 bit.
The reduction of the external data bus led to the fact that the exchange of information with the 386SX was carried out at half the speed than in the case of the 386DX.
This reduced the performance of the crystal by about 25%.
i386SX processor
Image source
The first 386SX had a frequency of 16 MHz, which then increased to 20, 25 and 33 MHz, respectively.
The SX version was intended for entry level desktop computers and portable systems.
In fact, the processor is "registered" in a huge number of home and office systems.
In addition to the SX and DX modifications, one of the first energy efficient 386SL processors was presented, designed primarily for laptops.
The "stone" had a frequency of 20 or 25 MHz and (unlike the 386SX) contained many built in controllers: for example, a RAM controller, a bus controller and an external cache memory controller, the volume of which varied from 16 to 64 KB.
In addition, the 386SL supported various "sleep" modes, as well as system management modes (System Management Mode).
Compaq Deskpro 386 computer
Image source
The first computer using the i386 processor was the Compaq Deskpro 386.
At that time, Compaq became the first "third party" company in history to make significant changes to the PC platform.
Until then, IBM was always the first to release new computers.
It could have been the first one this time, but IBM had a long term contract for the use of 286 processors, and the company preferred to devote some more time to the 16 bit platform.
As history has shown, this step was quite a big mistake.
Deskpro 386 sold well, so by the time the first IBM computers based on the 386 processor were launched, the company had already lost its leading position.
As a result, Compaq managed to "reshape" the entire desktop market a little.
Thus, competition increased, and the influence of IBM was no longer so significant.
Andy Grove former CEO of Intel
Image source
As before, i386 clones soon appeared on the market.
Several companies were engaged in their production: AMD, Cyrix and IBM.
However, Intel's own policy regarding clones has changed.
CEO of the company Andy Grove decided not to issue licenses for the production of modifications of the i386 to third party companies, but later they still appeared.
The first clones were released by AMD in March 1991.
The processors were ready long before this date, but Intel was sure that the license for the production of "duplicates" provided by AMD extended only to 80286 and earlier processors, so the case went to court.
The lawsuits lasted for quite a long time, but in the end AMD won the case, and the AMD Am386 processor family still saw the light.
The line included clones of both 386DX and 386SX processors.
The top model Am386DX received a clock frequency of 40 MHz, that is, 7 MHz more than the most productive Intel modification!
The performance of such a processor was at the level of the next — generation chip from Intel i486, which had already been released by that time.
At the same time, the cost of the AMD solution was much lower than that of Intel models.
Due to the favorable combination of price and speed, the processor has found application in many desktop systems.
AMD Am386DX Processor
Image source
As for the 386SX clone the Am386SX model it was not so much an ordinary copy, but a redesigned version of the" Intel " crystal.
So, the chip was produced according to a thinner 0.8 microns technological process and used a static core, which made it possible to achieve energy efficient operation of the processor.
On average, the Am386SX was 35% more economical than the original 386SX.
And even more economical than the 386SL processor designed specifically for portable devices.
At the same time, the clock frequencies of the Am386SX were, as a rule, higher than those of the 386SX (the maximum clock frequency was 40 MHz).
By the way, despite the fact that the Am386SX is a clone of the "Intel" chip, it is considered the first independent development of AMD.
And after the launch of the Am386 line, AMD was rightfully considered one of Intel's competitors.
AMD Am386SX Processor
Image source
i386 clones produced by Cyrix have been used in laptops and inexpensive desktop systems.
The line of "stones" consisted of two models: 486SLC and 486DLC.
Despite the index in the name, the processors were copies of the 386SX and 386DX, respectively.
However, it should be noted that Cyrix solutions have received support for the i486 instruction set.
An interesting architectural feature of the line was the presence of a first level cache memory with a volume of 1-8 Kbytes.
As for the clock frequency of the processors, its maximum indicator was 40 MHz, as in the case of AMD Am386.
At the same time, the power consumption of the Cx486 was at a very low level.
The processors could not compete with the AMD line.
Over time, Intel reduced the prices of its products, and the i486 managed to finally displace Cyrix crystals.
Cyrix 486DLC processor
Image source
IBM also did not stay away from the production of clones.
In 1991, it introduced the 386SLC and 386DLC processors, which were clones of the 386SX and 386DX, respectively.
They were used in IBM PS/2 and PS/ValuePoint desktop computers, as well as in the IBM ThinkPad laptop.
IBM ThinkPad Laptop
Image source
In addition to the above models, Intel has released processors for embedded systems: 80376 and 386EX.
The first crystal was released in January 1989.
It differed from the 386SX in the lack of support for the real mode of operation (the "stone" worked only in protected mode) and the process of replacing pages in the memory management unit.
The clock frequency of the 376th was 16/20 MHz.
After 5 years, 386EX replaced 80376.
The processor supported 26 bit memory addressing, had a static core that provided high energy efficiency, and many peripheral devices: for example, counters, timers and an interrupt controller.
Basically, the 386EX was used in the computer systems of various orbiting satellites, as well as in a NASA project called FlightLinux.
Intel i486
When developing processors of the next (read — the fourth) generation, Intel engineers faced serious problems.
The previous generation of integrated circuits reached the performance ceiling, and the technologies used at that time did not allow placing even more transistors on the same area.
The developers had no choice but to rework the existing architecture, or rather, to supplement it.
So, for the first time, i486 processors got such components as cache memory, a pipeline, an integrated coprocessor and a multiplication factor (multiplier).
Thanks to them, the new generation of CPUs has become faster than its predecessors.
But about everything in order.
"What is a cache memory?" — our readers know the answer to this question perfectly well.
It is located "between" the processor and RAM and stores copies of the most frequently used data from the main memory.
The access time to it is much less than to the main memory.
Therefore, when the necessary data is contained in the cache, the average memory access time is significantly reduced.
The i486 processor received an 8 KB cache memory.
The first 486 processors worked with the cache on the Write Through principle, that is, data was always written to the main memory, even if they were already present in the cache.
Then the "stones" learned how to work with the cache using the Write Back function.
When using this principle (if there is a copy of them in the cache), they were written only to the cache memory, no write was made to the RAM.
i486 processors also worked with the so called external cache, which was located on the motherboard.
Its volume at that time ranged from 256 KB to 512 KB.
One of the processors of the i486 family
Image source
The presence of a level 1 cache memory has significantly complicated the chip itself.
The i486 processor contained almost 1.2 million transistors.
About half of them were on the cache memory.
The complexity of the chip caused its high power consumption and heat dissipation.
For example, active cooling was used for the first time in systems using i486.
In addition to this, the complexity of the processor caused an increase in defects during production.
Consequently, because of this, the cost of the device has also increased.
i486 processors also have computing pipelines, the essence of which is to divide the processing of a computer instruction into a sequence of independent stages with the results saved at the end of each stage.
Something similar was implemented back in the Zilog Z8000.
The i486 pipeline consisted of five stages: sampling, decoding, decoding operand addresses, executing a command, and recording the result of executing an instruction.
The appearance of pipelines not only increased the speed, but also simplified the processor architecture to some extent.
It is also worth noting that the appearance of conveyors has had a positive effect on the acceleration potential CPU.
As for the coprocessor (FPU, Floating Point Unit, Floating Point Operations Module), it was a module that helps to perform mathematical operations on real numbers.
In the i486, it was built into the chip itself.
However, not all fourth generation crystals had an integrated FPU.
Multiplication coefficients did not appear in the first i486 processors at all.
The 486SX and 486DX models dispensed with it and worked at the system bus frequency.
Support for the multiplier appeared only in the 486DX2 crystal.
At a system bus frequency of 33 MHz, the clock frequency of the processor itself was 66 MHz.
That is, the multiplier was equal to two.
In 486DX4, the multiplication factor was increased to three.
Along with the introduction of multipliers, overclocking appeared.
i486DX Processor
Image source
As with the third generation, Intel initially introduced only two models to the market: the 486SX and 486DX.
As we have already said, the only difference between these processors was the lack of an integrated FPU.
Otherwise, the crystals were completely identical.
By the way, due to the large percentage of defects in production, some 486SX models were 486DX with a faulty coprocessor.
Thus, Intel tried to reduce production costs.
The chips were produced according to 1 microns of the technological process, and a little later, according to 0.8 microns of technorms.
The clock frequency of both models varied from 25 to 50 MHz.
The maximum power consumption reached the mark of 5 watts.
Intel i486SX processor
Image source
Despite the fact that Intel was increasingly protecting its developments with patents, many i486 clones appeared on the market.
AMD, Cyrix, IBM, Texas Instruments and others were engaged in the production of copies.
The basis of the AMD Am486 clone line was the Am486SX and Am486DX models.
The processors were produced according to a thinner 0.7 microns technical process, and then in accordance with 0.5 microns and 0.35 microns technological standards.
Technically, the Am486SX and Am486DX were complete analogs of Intel crystals.
Processors that did not use multipliers had frequencies from 25 to 40 MHz, and "stones" with a multiplication factor worked at a clock frequency from 66 to 100 MHz.
In 1995, AMD introduced the fastest i486 compatible processor called the Am5x86.
The crystal was produced according to a 350 nm process technology and had 1.6 million transistors.
The volume of the cache memory of the 1st level was increased to 16 KB, and the multiplication factor was equal to 4.
The processor worked with a bus with a frequency of 33 MHz, that is, the clock frequency of the crystal itself was 133 MHz.
In terms of performance, the Am5x86 was comparable to a Pentium processor with a frequency of 75 MHz.
The Am5x86 was AMD's top solution before the release of the new K5 generation processors.
A clone processor from AMD
Image source
The i486 clones produced by other companies were no different.
They architecturally repeated the original and, of course, had the same performance.
Motorola 68020, 68030, 68040
In 1984, a few months before the introduction of the i386, Motorola released its first fully 32 bit chip, the 68020.
The processor was produced according to a 2 micron process technology and had 190,000 transistors.
Its clock frequency ranged from 12 to 33 MHz.
In comparison with its predecessor, the 68010, the new processor has received many improvements.
First of all, it should be noted that the "twentieth" worked with full fledged 32 bit external data and address buses, and also supported new instructions and addressing modes.
At the same time, the execution time of some instructions was reduced.
Also, the 68020 became the first processor in the Motorola 68k line with built in cache memory of the first level.
However, its volume was only 256 bytes.
Alas, the 68020 did not have a built in FPU, although the crystal interface provided support for up to 8 coprocessors.
As for performance, at a frequency of 33 MHz, the result was 5.36 million instructions per second.
The main area of application of the processor again became Apple computers: Macintosh II and Macintosh LC.
In addition, the Motorola 68020 is also "registered" in the systems Sun 3, Hewlett Packard 8711, Sinclair QL and Alpha Microsystems AM 2000.
Macintosh LC Computer
Image source
At the same time, Motorola introduced a "stripped - down" version of the processor called the 68EC020.
So, the crystal had a 24 bit address bus and therefore could address only up to 16 MB of memory.
Kodak and Apple used it in their printers, and Commodore used it in Amiga 1200 computers and CD32 game consoles.
In 1987, the next Motorola processor appeared on sale — the 68030.
It was fully 32 bit.
The data bus in the 68030 became dynamic, it could function in 8 -, 16 and 32 bit modes.
There was also a synchronous mode of operation of the data bus and the address bus, which increased the data transfer speed.
Processor performance has also increased due to the additional 256 bytes of the first level cache, reducing the access time to the instruction cache and adding a memory management unit.
As in the case of the model 68020, the "thirtieth" did not have a built in coprocessor.
As for the technical characteristics, they largely coincided with those of the predecessor.
The processor clock frequency ranged from 16 MHz to 50 MHz.
In the second case, the performance of the "stone" was about 18 million instructions per second.
Motorola 68030 processor
Image source
The 68030 was used in the same Apple Macintosh II and Commodore Amiga computers, as well as in the Next Cube, Sun 3/80, Atari TT and Atari Falcon systems.
A stripped down version of the processor called 68EC030 was also released.
The Motorola 68040 processor, which became available in 1990, brought much more architectural changes than its predecessors.
So, for the first time there was a built in coprocessor.
In the "fortieth", support for the memory management unit, which appeared in the previous generation of "stones", has been preserved.
The size of the instruction memory cache and the data memory cache has been increased to 4 Kbytes each.
The principle of operation of the processor was based on computing pipelines, which consisted of six stages.
With the advent of an integrated coprocessor and an increase in the amount of cache memory, the chip has become much more complicated.
At the same frequency, the performance of the 68040 model exceeded the CPU speed of previous generations by more than four times.
At the same time, the crystal was very hot, and the developers were not able to solve this problem.
Hence, the processor's clock frequency never exceeded the 40 MHz mark, although Motorola itself had plans to launch a 50 megahertz version.
In terms of its capabilities, the Motorola 68040 was comparable to the Intel i486 and even surpassed it in speed at the same frequency.
However, overclocking was gaining momentum at that time, and the 486 differed favorably from the "Motorola" chip just by its overclocking potential.
Motorola did not forget about the younger versions of the 68040, of which this time there were two: 68EC040 and 68LC040.
Both models lacked a built in coprocessor, and the 68EC040 version also lost the memory management unit.
Despite this, the processors showed decent performance.
So, the 68LC040 was not inferior in speed to the 68020.
Macintosh Quadra Computer
Image source
As before, Apple made extensive use of Motorola processors in its computers.
For example, the 68040 became the basis of the High End Macintosh Quadra system.
Macintosh Centris and Performa also used processors of the 68040 family.
Plus, Amiga 4000/4000T computers were built on the basis of the Fortieth, as well as some Alpha Microsystems and HP servers.
NEC V60, V70, V80
Japanese manufacturers are known for paying a lot of attention to the domestic market.
The situation with NEC processors was about the same: the first 32 bit solution called V60, released in February 1986, was available only in the Land of the Rising Sun.
If all the previous developments of the company were based on the x86 architecture, now the design of the processors partly resembled the RISC model with a large number of general purpose registers.
Nevertheless, the "stone" inherited a lot from the CISC architecture.
x86 platform emulation was also supported.
The V60 was not fully 32 bit — the width of the external data bus was 16 bits, and the address bus was 24 bit.
The processor itself was produced according to a 1.5 microns process technology and had 375 thousand transistors.
It used computing pipelines with six stages, and also had a built in coprocessor and a memory management unit.
At a clock frequency of 16 MHz, the performance of the V60 was 3.5 million instructions per second.
Sega used this "stone" in its arcade platforms System 32 and Model 1.
NEC V60 (bottom row, left) and V70 (bottom row, right)processors
Image source
A year later, NEC released the V70 processor.
The new chip was able to work with full fledged 32 bit data and address buses — this was its main difference from the V60.
At first, the processor was produced according to 1.5 microns of the technical process, but then 1.2 microns of technological norms were mastered.
The maximum clock frequency of the V70 was 20 MHz.
At this speed, the chip's performance was 6.6 million instructions per second.
Like its predecessor, the V70 was not very popular, but Sega continued to use it in its platforms.
In the spring of 1989, the V80 processor saw the light.
This crystal already had an instruction cache and a data cache of 1 KB each, as well as a branch prediction block.
The production of the crystal was transferred to the 0.8 microns process technology, and the chip itself contained 980 thousand transistors.
The clock speed indicator also increased: V80 it worked at a frequency of 25/33 MHz.
The flagship's performance was 16.5 million instructions per second.
NEC plans included the release of the V80 with a frequency of 45 MHz, but the project remained on paper.
The V80 was the last NEC "stone" with the CISC architecture.
The next line of processors, the V800, was based on the RISC architecture.
DEC VAX
In 1985, DEC offered its line of 32 bit processors based on its own VAX architecture.
The first swallow was the MicroVAX 78032, which was also the first device manufactured by the company's own forces.
The processor contained 125 thousand transistors and was manufactured according to 3 microns NMOS technological standards.
At a frequency of 5 MHz, the chip's performance when working with integers was comparable to the speed of the VAX 11/780 superminicomputer released in 1977.
To speed up operations with floating point numbers in a duet with MicroVAX 78032, it was proposed to use the MicroVAX 78132 coprocessor.
The 78032 crystal was used in DEC computers, starting with the MicroVAX II model.
DEC MicroVAX II computer
Image source
And in 1987, the CVAX processor was introduced — the next generation of "stones" with the VAX architecture.
Although it is more correct to call this solution not a CPU, but a chipset.
CVAX consisted of the processor itself with the index 78034, a coprocessor for accelerating CFPA floating point calculations, and various support chips (for example, a memory controller and a Q Bus controller).
The clock frequency of the VAX 78034 was 11.11 MHz or 12.5 MHz.
The processor was produced using the first generation CMOS technology.
The crystal received a common cache of instructions and data with a volume of 1 KB.
While Intel and Motorola solutions used SRAM memory as the cache, the DEC processor cache used DRAM type memory.
The 78034 also supported 64 KB of external cache memory.
The first systems using CVAX appeared in 1987.
Among them were both high performance solutions (VAX 6000 Model 200) and Low end MicroVAX 3500 minicomputers.
DEC CVAX Chipset
Image source
The last chipset that DEC managed to release in the 80s was DEC Rigel.
Like CVAX, it consisted of a processor (REX520), a DC523 coprocessor, a DC592 cache controller, and support chips.
The volume of the level memory cache was increased to 2 Kbytes, and the size of the external cache was 128 Kbytes.
The REX520 was manufactured using the second generation CMOS technology and contained 320 thousand transistors, more than half of which accounted for cache memory.
Like its predecessors, the processor was used only in DEC systems (VAX 6000 Model 300/400 and VAXstation 3100 Model 76).
Instead of a conclusion
By the end of the 1980s, the situation on the processor market, as you can see, had changed slightly.
Intel only strengthened its position, and Motorola, in turn, tried to keep up.
A little later, in the early 1990s, AMD loudly declared itself.
However, these companies relied on time tested CISC architectures, while many others began to produce processors based on the RISC architecture.
There were many interesting and promising developments among them.
We will tell you about them next time.
To share:
Did you see a typo?
Select the text and press Ctrl+Enter
The main thing for the week
29.03.2016 The FBI hacked the iPhone without the help of Apple
01.04.2016 YouTube launched SnoopaVision
01.04.2016 A new Android Trojan is distributed using GPS
31.03.2016 Military strategy War Thunder: Conflicts is available on iOS and Android
29.03.2016 Konami showed Contra for iOS and Android
News of the section
17:29 NVIDIA Quadro M5500 has become the fastest graphics for laptops
14: 35 Intel has confirmed the Core i7-6950X Extreme Edition processor
01.04 14: 18 MSI introduced the modular motherboard The One
01.04 09: 24 Intel introduced 14 nm Xeon Broadwell EP processors
30.03 17: 13 Ferra.ru will broadcast the presentation of Microsoft Build 2016
30.03 13:44 MediaTek is preparing a new ten core Helio X30
24.03 13: 30 Intel canceled the "Tick Tock" strategy
24.03 11: 56 The performance of the Russian Baikal T1 processor is revealed for the first time
23.03 11: 27 NVIDIA introduced the Quadro M6000 with 24 GB of video memory
21.03 19:13 AeroCool presented the PGS V Cruisestar gaming case
All news
Mobile application
Mailing
Receive news and reviews by email:
Testlab
Legal information
Contacts
Forum
RSS
Ferra.ru ("Ferrat.Ru"), 18+
Email address: info@ferra.ru * - Information about advertisers
