The evolution of overclocking processors AMD and Intel. Overclocking dashing 90s to the new time

Striving for perfection - perhaps this is the way to describe the development of mankind. Take, for example, female beauty. What just do girls to be the most beautiful. But in the end they can turn into real silicone dolls.

The same applies to men. Moreover, this mania is manifested to a greater extent, not only in relation to his loved ones. Already moving to the subject of this material, it is easy to see how overclocking evolved - the male kind of desire for the ideal. And along with overclocking, both enthusiasts and companies producing iron. Today, for the masters of overclocking, the main thing is to demonstrate the maximum, momentary success. Therefore, now non-standard types of cooling are used in the form of systems of phase transition or the use of liquid nitrogen. Although everyone understands that components will not work constantly under such extreme loads - and it will be expensive to store a nitrogen tank.
')
But before overclocking existed solely for the benefit. For a man wanted to get more performance for less money. It all started with processors - they were the first components capable of overclocking. Later, the chase began for megahertz in the world of video cards and RAM.

Part 2: Modding: History. Difficult path

First attempts

It all began with the development of overclocking clock generator. Namely, the unit was controlled by closing certain FS-contacts. A set of different signals (high or low) allowed to get the values ​​of logical zero and one. The result is a table with certain processor frequencies. Later on, motherboards began supplying jumpers that changed the clock signal. As a rule, the middle contact was responsible for the number of FS-legs, and the other two - for the ground and voltage. CPU overclocking was performed in the same way. At first, the increase in frequencies did not promise big dividends. Prehistoric nuclei could be crammed to 5-10 extra megahertz.

The first official evidence of overclocking on the now legendary resource hwbot.org was the AMD Am386 processor, released back in 1991.



This "stone" was intended to compete with Intel 80386. Although the word "competition" is too much said. Performed according to the 1000-nanometer process technology, "three hundred and eighty-sixth" was a complete copy of the crystal heirs of Gordon Moore. However, if such borrowing of ideas took place today, Intel would have eaten “micro-devices” with giblets with its court cases. Whatever it was, AMD Am386 had a 32-bit data bus, and was also equipped with 80387 FPU. And this is with a resource of 275,000 transistors! The frequency of the "stone" varied depending on the clock generator, but not much - only 12-40 MHz. Moreover, the previously mentioned rival crystal Intel 80386 worked at a maximum speed of 33 MHz. As you can see, eternal “friends” still have no measure of causal places.

The most productive AMD Am386 processor was an AMD Am386DX-40 device. From the name it is clear that the silicon device functioned with a clock frequency of 40 MHz. But the Portuguese overclocker WoOx3r [Pt] at one time managed to overclock the “stone” up to 50 MHz!

Let now the result provided and sounds ridiculous, but then it was a record of records. By the way, on similar characteristics, the Super Pi test with a pattern of one million decimal places was completed in some 2 days 21 hours 36 minutes and 32.992 seconds. Quick, right?


Both representatives

The next generation CPU turned out to be more productive: AMD Am486 and AMD Am5x86. The first family of processors appeared in 1993. On the silicon "hump" of the novelty, 1 185 000 transistors were located, thanks to the transition to the 800-nanometer process technology. Naturally, the frequencies went up. If at first models came out with a low speed of up to 40 MHz, then the clock frequency of the “stone” rose to 120 MHz. Enthusiasts were not ashamed to overclock the new processors. For example, DrSwizz overclocker was able to run AMD Am486DX-25 at 33 MHz. The Super Pi benchmark test calculated a million decimal places in 2 hours 4 minutes and 59 seconds (compare with the result of Am386).

Already in 1995, enthusiasts played enough with the AMD Am5x86-P75 processor.
Thus, the core under the code name X5 was able to overclock to 162 MHz - more than twice. As a result, the Czech overclocker orange overcame the Super Pi test in just 36 minutes exactly.

Intel time

Simultaneously with the release of the AMD Am5x86 chip, the brand of Intel Pentium processors appeared, which later became a cult. Among overclockers, the Pro or P6 series chip has become very popular.

In fact, under this brand was hiding a crystal with a completely different architecture than the usual “stump”. First, through the use of a dual-independent bus architecture, memory bandwidth limitations were removed. For what it was necessary to develop a special slot - Socket 8. The technology of placing two chips was also used for the first time.
One of them was, strictly speaking, a CPU with 5.5 million transistors, made according to a 250-nanometer process technology. A second chip played the role of a second-level cache. Over time, Pentium Pro models were released with 256, 512 and 1024 Kbytes of SRAM memory. The design worked at the expense of the 387-pin SPGA-case at a supply voltage of 3.3 volts. Among overclockers, the Intel Pentium Pro model with a 256 KB cache in the second level, operating at 200 MHz, has become popular. For example, our compatriot Veld overclocked P6 to 245 MHz. But faster than all the test Super Pi passed again Russian frag_: Intel Pentium Pro at a frequency of 225 MHz calculated a million characters in 7 minutes 44.700 seconds.

An interesting situation. Many overclockers decide to mock the "iron" after some time. For fun or in a rush of nostalgic feelings. Never mind. But in 2009, the Ukrainian RomanLV, with a pair of Intel Pentium Pro operating at 240 MHz, managed to pass the wPrime 32m test in 6 minutes and 41.190 seconds.

AMD again

The following year, after the announcement of the Intel Pentium, AMD broke out with another generation of its processors. At this time, there was no plagiarism, and the family of silicon dreamers with the sonorous name K5 acquired its own individual features. In fact, this CPU is the first separate product of the corporation. Naturally, AMD "stone" positioned itself as the main (and how else?) Competitor to the Intel Pentium. It was then that an interesting mess appeared with the names of the processors. Since the AMD K5 PR133 with a clock frequency of 100 MHz was considered an analogue of the Intel Pentium chip, operating at a speed of 133 MHz (the so-called PR rating has gone since that time). All in all, there were “stones” with the 75, 90, 100 and 116 MHz signals in the model range of “green”. There were also funny situations when absolutely identical AMD K5 PR90 and AMD K5 PR120 chips were produced, operating at a frequency of 90 MHz. Crystal "framed" in accordance with the 350-nanometer process technology, which allowed to place 4.3 million transistors. The first level cache was divided into 8 KB for data and 16 KB for instructions.

But the shared memory of the second level was not even planned. She unsoldered on the motherboard. The energy consumption level of the fifth generation of processors exceeded 10 psychological watts. And for their cooling (processors, well, and watts, too) required the use of not only passive, but also active air cooling. However, this did not scare overclockers at all. So, the Brazilian boy RIBEIROCROSS became the best among the overclocking of AMD K5 PR133. He managed to start the “Pyaterochka” at a frequency of 142.5 MHz and pass the benchmark Super Pi 1m in 12 minutes and 48.640 seconds. The top-end AMD K5 PR166 processor (@ 116 MHz), under the scrutiny of a well-known retro-caster, orange won the 150.5 MHz mark. With the help of the same device, Croatian extreme skydec passed the Super Pi 32m test in 18 hours 52 minutes and 40.392 seconds.

Evolve together

On May 7, 1997, Intel announced the continuation of the Pentium processor line. The second “stump” was nothing more than the processing of the P6 core, the potential of which was discussed above. The upgrade of the crystal was to increase the cache of the first level from 16 Kbytes to 32 Kbytes, as well as the appearance of a block of SIMD instructions MMX. Therefore, Intel Pentium MMX should not be considered the first processor with exclusive (at that time) multimedia extensions. By the way, at the same time as the P6 reincarnation, the SDRAM standard and the AGP (Accelerated Graphics Port) AGP interface became very popular.

In total, the second "stump" existed in five guises. The first is originally considered the core of Klamath. Processors based on it had a FSB bus with a frequency of 66 MHz, and the CPU itself operated at a speed of 233-300 MHz. In this case, the external cache of the second level (512 Kbytes) worked at half the core frequency. The design of the device itself was a cartridge with elements soldered on it. Later, such a case had to be abandoned in favor of a textolite plate, very similar to today's processors.
The next core of Deschutes was still located in the cartridge installed in Slot 1. Differences from Klamath consisted in switching to a 250-nanometer process technology. Hence, the consumed voltage of the processor decreased from three volts to two, and the frequencies increased to 450 MHz. The “stone” Pentium II 350 MHz has become very popular. Overclocker Jonh from sunny Argentina even managed to get the model model up to 601 MHz! The test in the form of Super Pi 1m with such characteristics of the CPU was overcome in an average of 200 seconds.

Name, sister, name!

Among the Intel Pentium II later, P6T (OverDrive) and Tonga / Dixon mobile cores appeared. However, they did not delight transcendental clock speeds. But overclocking would not be so popular, if the first Celeron processor did not appear on April 15, 1998. These budget cards without a second-level cache literally conquered the hearts of overclockers around the world.
And some retro-bankers still indulge themselves in dispersing “celery” (this is how people call Celeron because of the very close resemblance to the word Celery).

The performance of this chip was at a very low level. But acceleration at the absolute maximum frequency could not but rejoice. At the same time, similar results began to be called popcorn. So the Slovenian Moonman managed to swing the Intel Celeron 433 MHz (based on the Mendocino core) to 780 MHz. To do this, we had to increase the bus speed to 120 MHz. The “stone” multiplier was kept at x6.5 units.

Just K6

Meanwhile, AMD is not inactive. In 1997, the corporation introduced the K6 processor (Model 6).
As usual, the new processors positioned themselves as an alternative to the Intel Pentium.
Therefore, the names of the crystals were adjusted according to the frequency potential of competitors.

The core after the transition to the 350-nanometer process technology has got 8.8 million transistors. And later came the variation of Little Foot (or Model 7), processed by the "file" to 250 nanometers. The first-level cache was 64 KB, equally divided into data and instructions. The processor worked with frequencies 166, 200 and 233 MHz. “Lapa”, as it was affectionately called, was able to reach 300 MHz. Why the seventh model was unclaimed by overclockers is a mystery. But the Model 6 perfectly yielded to overclocking. The record belongs to the Austrian Turrican, who launched a 233 MHz chip at a frequency of 310 MHz.

Similarly, the K6 new family - K6-2 - was designed to compete with the Intel Pentium II. "Stone" consisted of 9.3 million transistors, for which the area of ​​the crystal had to be increased from 68 to 81 square millimeters. The processor's heat emission has also risen, reaching 28.4 watts depending on the model. However, the true "soldier" Socket 7 did not require an active cooling system. And with the help of the usual 120-millimeter turntables, the Belgian retroman Massman dispersed the AMD K6-2 (Model 8) to 720.5 MHz.
Our compatriot, qwerty84, forced the processor to pass the Super Pi 1m test at 650 MHz in 5 minutes and 12.44 seconds.

Later (November 16, 1998), AMD released the core Chomper Extended. True, the frequency of such "stones" increased not much. The top device functioned at a speed of 550 MHz. The best result of overclocking belongs again to Turrican: 744.6 MHz.

Finally, the era of the decline of the K6 line was marked by IA-32 microarchitecture processors introduced to the public in February 1999. The Sharptooth and K6-III-P cores have acquired a second-level full-speed cache, etched right on the chip. By the way, for 256 Kb fast “brains” of the chip, we had to spend 21.3 million transistors, but without upgrading the technical process.

Chip frequencies did not differ from the sixth, seventh and eighth models. Unfortunately, the overclocking potential of the new CPU is not pleased. Overclocker GtaduS managed to squeeze 575.1 MHz from the AMD K6-III 450 MHz model (Model 9).

At the border of millennia

Probably, it would not be entirely logical if the Intel and AMD processors did not make a huge leap forward on the border of old and new times. From the first side, the Intel Pentium III processor became this jump. Released on February 26, 1999, the Katmai core initially did not have supernatural characteristics. Frequencies were generally at the level of 450-600 MHz. One of the few differences of the Deschutes modified crystal was the optimization of work with memory and an extended set of SSE commands.
Later, the third "Penek" was updated in the form of a Coppermine chip. The processor frequencies have finally reached gigahertz! This miracle happened on March 8, 2000. True, among overclockers, the conquest of this milestone was celebrated a little earlier. To be more precise, back in 1999 (officially, the “stone” was introduced on October 25), when the Intel Pentium III processor with a frequency of 733 MHz subdued the cherished milestone due to overclocking.

To date, the record belongs to the Dutch enthusiast _Datura_: the guy managed to remove the validation at 1181.3 MHz on the core. It is noteworthy, but to achieve a similar result, the overclocker had to use a phase transition system (read freonku). The SDRAM test bench memory functioned at a frequency of 215 MHz, for which the water block had to be mounted on the module.

As always, the excellent overclocking potential was demonstrated by the “stones” of the Celeron line. Based on the same Coppermine core, the processors had 128 Kbytes of a 4-channel second-level cache and a 66 MHz FSB bus. As a result, the memory latency has doubled compared to a regular Pentium III.
But the overclocking potential of the silicon device did not cause any complaints. All thanks to the high multiplication factor x8. As a result, the model with a nominal frequency of 800 MHz was launched at 1406 MHz. At the same time, the overclocker DDC, who came from the country of tulips, did not have to install anything except a more powerful fan for a stock cooler.

Last word

In the summer of 1999, AMD designated the Athlon processor line with the K7 microarchitecture. As usual, Argon, Pluto and Orion crystals were made after Intel "stones". And as usual, they positioned themselves as an equivalent replacement. Only at the beginning of the seventh generation did not work out with overclocking. The potential of the first Athlon was very low. When the 700 MHz “stump” easily conquered psychological gigahertz, the similar Orion barely overcame the 800 MHz mark.

This is evidenced by the result of the overclocker mafler, installed 10 years later: AMD Athlon 700 MHz was launched at a frequency of 889.15 MHz.

More excitement caused the release of processors on the core of Thunderbird. Model AMD Athlon 1000 won the all-time mark of 2184 MHz! For what it is worth saying thanks to the French overclocker cpulloverclock.

It was on such a positive note that the overclocker cakeug met a new millennium. In many ways, its success and indicated, like a compass, the direction of development of central processors, Intel and AMD. And ahead were two thousandths. Ahead was an interesting and intriguing road.

New Millennium

The industry has met the new millennium with enthusiasm.
In November 2000, the Pentium 4 came out. Work on the processors of this line began in 1998, but, due to many difficulties, the development lasted until the end of 2000. The new processors were created on the NetBurst micro-architecture, which had fundamental differences from the P6 micro-architecture, on the basis of which the Pentium II and Pentium III processors were built, so they received the new name - Pentium 4.



The first modifications of Pentium 4 processors were not very successful. They lost in performance top models of the Pentium III and competing processors from AMD. And the prices of these processors were high. However, over time, when faster modifications of processors of this line appeared, Pentium 4 began to win its niche in the market of computer technology.

But the Pentium 4 was not bad at all and it supported the SSE2 and SSE3 instruction sets. And in combination with HyperThreading, Pentium 4 perfectly coped with both multimedia and content tasks, as well as with codes optimized for the new core. And the use of graphics cards for 3D graphics improved performance even more, so the P4 processor laid the foundation for the development of gaming tools.

Overclockers showed great interest in the Northwood core, released in 2002. With a suitable motherboard and memory, even beginning overclockers could raise the clock frequency by 1 GHz with air cooling.

But in order for the Pentium 4 to really shine, it was necessary to raise the clock frequency to record numbers. Intel assumed that this could be achieved with the Prescott core - the first chip manufactured using 90nm technology. But Prescott gave only a slight increase in performance, as opposed to loud advertising promises, and in game tests it was significantly inferior to AMD processors.
Pentium 4 was the first processor that in all versions was already within the concept of Socket. Socket 478 - came into use for a long time, the cartridge system was forgotten.

AMD era

At this time, AMD, with a line of Athlon XP and a new system for describing the clock frequency (1800+) entered the market. Part of the Athlon family, after revising XP and adding SSE instructions, was another aggressive step in AMD marketing. XP supported eXtreme Performance and got along fine with Windows XP. In addition, AMD has returned to using the Performance Rating (PR) system for marking processors. Officially, AMD's PR had to characterize the performance of the XP processor in relation to the Thunderbird core, so theoretically the AMD Athlon XP 1800+ should have the same performance as the Thunderbird at 1.8 GHz. However, in practice this abbreviation was mistakenly used much more widely, for example, as a pointer to the corresponding Intel processor - largely due to the coincidence of the abbreviations "Pentium Rating" and "Performance Rating".



The most popular Socket A Athlon was created based on the Barton core, which appeared in 2003 and promised tremendous overclocking capabilities. In particular, interest was caused by the first version of the processor - Barton 2500+, which came with an unlocked multiplier. By increasing the multiplier value, most Barton 2500+ processors could easily achieve the performance of the flagship AMD 3200+ model.

Of course, AMD engineers could not afford the luxury of how to remove protection against overclocking. The new Athlon XP / MP on the Palomino core was a great example of high-quality work that the chip maker is capable of. Before that, it was possible to connect tracks to “transform” the processor into a more powerful model. This method was very effective in past Athlon with the core of Thunderbird. Thus, the dreams of the steep accelerators, who had already made plans for overclocking before the purchase of the processor, dissipated. But overclocking power was phenomenal without it!





The next jump occurred again in the ranks of AMD. The success of AMD is the top of its 64-bit Athlon 64 processor, designed for the majority of users. While Intel engineers were trying to create a P4 processor based on NetBurst, AMD started producing chips with a more efficient architecture and an integrated memory controller.

Although A64 offered its own 64-bit base, it was also fully compatible with 32-bit encoding without any noticeable loss in performance. This was very important for Windows users who still lived in the 32-bit world.

Intel is not appeased. The unlucky architecture of NetBurst finally surrendered its positions in the latest Intel Pentium D brand. Pentium D processors, containing two single-core processors, were subsequently transformed into multi-core modules.Not as elegant as AMD's dual-core development, the Pentium D offered decent multitasking performance, good overclocking capabilities at a relatively low price. Pentium D has provided Intel followers with a confident alternative to AMD.

Continuing to dominate the desktop market, AMD's Athlon 64 X2 processor series contained two cores in a single chip, sharing an integrated memory controller. This internal structure of data exchange provided a huge performance advantage over the Intel dual-core configuration, in which the cores communicated via a common bus. In the X2 series, SSE3 commands have been added.

Intel vs AMD

Awakened by the "hibernation", Intel begins to storm the processor world with its new Core 2 architecture.

Instead of concentrating on achieving the maximum clock frequency, Intel focused on the higher performance of its processor pipeline. This meant a return to lower clock speeds, but on the other hand, it increased processor performance. But after the failure of Prescott was revealed, the media cautiously reacted to the promises of Intel regarding the performance of Core 2. But, to the deep disappointment of AMD, Core 2 fully complied with the stated capabilities.

The first Core 2 Duo literally blew up the market. Despite the debut with low frequencies of 1.86 GHz and 2.13 GHz (E6300 and E6400, respectively), performance, as well as aggressive pricing, made Core 2 desirable and popular.

Later, Core 2 was transferred to 45 nm manufacturing technology. So there was a version of Penryn, in which 820 million transistors were packaged in a quad-core processor operating at a frequency of 3.2 GHz. The minus was in the temperatures of the processor.

AMD, handing over the performance benchmark to Intel's Core 2, nonetheless hoped to make a breakthrough in the market with the future Barcelona processor, which was later renamed Phenom. But early versions of the Phenom contained bugs and often malfunctioned. And in the back of his head Intel's Nehalem architecture was already breathing.

It cannot be said that Phenom was such a bad architecture - it certainly had its own advantages: several SIMD instructions, including MMX, Enhanced 3DNow !, SSE, SSE2, SSE3 and SSE4a, 4-core processor and good performance. But all this is incomparably inferior to the level of the latest Intel processors, moreover, AMD has lost Intel in price policy.



The Core i7 processor, released in 2008, further reinforced the concerns of AMD, which was still hoping to compete for an architecture capable of competing with Core 2. Meanwhile, the Core i7 (previously known as Nehalem) remained out of competition.

And Intel, meanwhile, finally withdrew from the traditional bus in favor of QuickPath Interconnect, which was an analogue of AMD's HyperTransport. This point-to-point interconnect allows communication between the processor and various subsystems much faster. However, because of this, overclockers had to “improve their skills”, including mastering several new terms in order to learn how to implement overclocking properly.

At first, these were processors that were difficult to overclock, and also Intel first started blocking overclocking from the box. After these processors, special modifications for overclockers began to appear - a series with notes (K - for overclocking with unlocked multipliers), while the rest were already trimmed.

Many believe that the Phenom II is what the original Phenom should have been. Together with tripled third-level cache (6 MB instead of 2 MB), DDR3 support, and removal of “cold bugs” that poisoned overclockers, Phenom II closed the performance gap with the Intel Core 2 line. But AMD still had Problem: Intel has already taken the next step, but AMD has nothing to offer to users as a competitor to the Core i7.

Being unable to compete with Intel in performance, AMD had to lower the prices of its processors much more than we would like. While the Athlon 64 X2 tended to be high, the Phenom II X4 940 had a retail price of just $ 215, noticeably less than $ 1,000, which they usually asked for for flagship processors.

Intel: Pros and Cons

Since the advent of Corei7, a new era began, the number of overclockers and groups after the peak of the first processors on Bloomfield began to fall. And Intel actively began to promote the idea of ​​the integrated video core in the processor. Blocked frequency on all versions except the K series did not add to the popularity of overclocking processors, as a result, the main frequency records of those years won the AMD PHENOM II X2.

But enthusiasts still remained, nitrogen was also used, but with the advent of i7, this is a completely different era, deserving of a separate article.

Instead of an afterword

Many thanks to the magazine "Iron", which now does not exist, on his articles and information grew my vision of the world of Hardware. Overclocking, modder scene - generally remain one of the most memorable things.
The modder scene is unfortunately not so well-known, even in IT circles, although Russian modders are making different adaptations from cyberpunk and steam-punk to various fan fiction.

To be continued, of course, if you are interested.

UPD:
Fixed bugs.