Believe it or not, your terrifically fast Core i7 fresh off Intel's assembly line contains DNA that dates back over three decades. The same is true if you roll with AMD's latest silicon, the Phenom II X4. We're of course referring to the longstanding x86 microprocessor architecture that has dominated the desktop and mobile scene since before some of you were even born, and will probably be a mainstay still yet for many more years to come.
Invented by Intel in 1978, the x86 architecture has evolved through the ages, not only getting faster, but increasingly flexible as more and more extensions and instruction sets accompany each new release. It's been a wild ride the past 30 years, and whether you lived through it all or have only recently picked up your first processor, we invite you to join as we look back at not only the most popular x86 CPUs in its history, but ones you may never even have heard of. Of course, if we've missed any of your favorite CPUs, chime in on the comments section and point out any omissions!
In the beginning, Intel created the 8086 and its first 16-bit microprocessor.
And Intel said, Let there be x86: and there was x86.
And Intel saw the x86, that it was good.
No, we're not about to anoint Intel as a deity, but the gargantuan chip maker did give birth to the x86 processor. More than 30 years later, or roughly 3,000 calendar years in computer time, x86 continues to evolve (see how easy it is for creationism and evolution to co-exist?) from its modest start in 1978. That was the year Intel created the 8086, a 3-micron chip chugging along at 4.77MHz, while later versions would run at up to 10MHz. The 8086 had just 29,000 transistors, which was still nearly four times as many as the 8085 released in 1976, and was Intel's first 16-bit microprocessor and responsible for kicking off the 16-bit era (note that the 8086 wasn't the first 16-bit chip). Backwards compatibility with software written for the 8008, 8080, and 8085, and the ability to address 1MB of memory natively made the 8086 a near instant success.
(Image Credit: CPU-World.com)
Clockspeed: 4.77MHz - 10MHz
Through some sneaky industrial espionage, the Soviet Union was able to reverse engineer and replicate the 8086 into their own pin-compatible K1810BM86.
The 8086 and later the 8088 closed out the 70s and were the opening act in the early 80s. That is, until Intel entertained the computing world with the 80286 in 1982, a 1.5-micron part with a mind-boggling 134,000 transistors and 16MB of addressable memory. The first 286 pedaled along at 6MHz and, like the original 8086, would later double in speed. However, clock for clock, the 286 boasted twice (or more) the performance of the 8086, a generational leap in the x86 architecture that has never been duplicated to this day. Throughout the decade, the 286 became synonymous with IBM PCs, and within 6 years of its release, Intel estimates there were 15 million 286-based PCs installed worldwide.
Introduced with the 286 was a feature called protected mode, which controlled how memory was accessed. This feature allowed all 16MB of memory to be addressed, but there was no easy mechanism in place for the 286 to switch from protected mode back to the backwards compatible real mode, so it was never widely used.
(Image Credit: Pipux.net)
Clockspeed: 6MHz - 12.5MHz
Bill Gates famously dubbed the 286 as a "brain dead chip," since it wasn't able to run multiple MS-DOS applications under the Windows environment.
Much has been made recently over the x86 licensing agreement between Intel and AMD, and we have to travel back to 1982 to see how it all began. That was the year AMD inked a deal allowing them to manufacture and sell both 8086 and 8088 processors. The very next year, AMD released the Am286, an exact clone of Intel's 286 processor, right down to the pin count, but with a higher frequency. And not just faster, but almost twice as fast at 20MHz. In some respects, the Am286 can be viewed as the first punch thrown in a fight that has been going on for almost 30 years.
(Image Credit: CPU-World.com)
Clockspeed: 8MHz - 20MHz
Like the Intel 286, the Am286 was built on a 1500nm fab process. Today's CPUs are made with a process that's 33 times smaller.
With PC gaming starting to take hold, Intel's 386 arrived not a moment too soon. Even adventure gaming could sometimes be a chore on a 286. Every try maneuvering Leisure Suit Larry in remade VGA form across the screen on a 286? We have, and it wasn't fun.
The 386, which was later named 386DX to avoid confusion with a lower cost 386SX variant that would debut three years after launch, initially ran at 16MHz and, once again, would eventually double in speed to 33MHz. It also doubled the number of transistors from its predecessor to 275,000 and was Intel's first 32-bit processor. The 386 could address up to 4GB (not MB) of memory, could switch between protected mode and real mode, and added a third 'virtual' mode, which allowed the execution of real mode applications that were unable to run in protected mode.
(Image Credit: yjfy.com)
Clockspeed : 16MHz - 33MHz
The 386 was the first widespread microprocessor to be initially single-sourced. That is, PC makers could only buy the chip from Intel, a policy which contributed to the company's success in the CPU market.
Starting to see a pattern? Before the greatest decade ever (1980s) came to end, Intel released one more x86 processor, the 486DX. The first CPU to include a built-in math co-processor, the 486 raced along at 25MHz (and later 50MHz) and was also the first chip to breach the 1 million transistor mark with 1.2 million transistors. Like the 386, it could address up to 4GB of memory, and with the addition of on-board cache, optimized instruction set and enhanced bus interface unit, the speedy 486 found a home in both desktop and server environments.
For gamers, the 486 picked up where the 386 left off, and most old-school gamers probably have fond memories of blowing up Tie Fighters while using a 486DX2-66. Fast as it was (at the time), the 486 couldn't keep up with the pixel pushing power required by the advent of 3D graphics.
(Image Credit: Wikimedia.org)
Clockspeed : 25MHz - 100MHz
Initially launched as the i486DX, the 486 design included numerous variations, including the i486SX, i486SL, and the widely popular i486DX2.
If the Am286 represents AMD's first punch thrown in the x86 bout, the AM386 was the uppercut that followed. Released in 1991, the AM386 was another carbon-copy of Intel's 386 CPU, and once again it came clocked faster than Intel's own silicon. It also came with a marketing edge as the first processor to come adorned with Microsoft's "Windows Compatible" logo, a move The New York Times described as "clearly intended to add credibility to [AMD]'s clones of Intel's microprocessors."
Intel did its best to prevent AMD from selling its new processor, alleging that the x86 agreement the two signed only applied to 80286 and previous processors (sound familiar?). AMD won the ensuing court battle, and though Intel had already released its 486 CPU, the Am386 offered near the same performance for much less. The flurry of sales that followed solidified AMD as a true competitor to Intel.
(Image Credit: hattix.co.uk)
Date Released: 1991
Clockspeed: 12MHz - 40MHz
The Am386 was ready for release before 1991, but was tied up in court due to a dispute over AMD's x86 license, which Intel claimed over covered the 80286 design.
Cyrix started off by producing math coprocessors for 286 and 386 systems in 1988, and in 1992, the company released its first x86 CPUs, the 486SLC and 486DLC. While the name might suggest otherwise, both of these were pin-compatible with the 386SX/DX, offering 386 platform owners an attractive upgrade option.
Manufactured by Texas Instruments, Cyrix's 486 series didn't come with a math coprocessor, though one could be added. Throughout its lifespan, the Cx486 series would come with anywhere between 1KB and 8KB of L1 cache and ramp as high as 100MHz.
(Image Credit: cpu-museum.de)
Now in its fifth generation, Intel's Pentium processor brought the x86 architecture to new heights, as well as brought along a new naming scheme. Unable to patent numbers, Intel avoiding dubbing its newest chip the 586.
The Pentium introduced several improvements designed to address the performance bottlenecks of previous processors. Chief among them was a 64-bit wide date bus, two execution units, a much improved floating point unit (FPU), and faster clockspeeds. Intel's Pentium processor launched at 60MHz, but it didn't take long for faster chips to follow before it eventually topped out a 233MHz. During its lifespan, the Pentium also shrunk from a 0.8-micron manufacturing process to 0.35-microns and increased its transistor count from 3.1 million to 4.5 million.
In 1996, Intel began selling Pentium MMX processors. The MMX instruction set added additional registers to the architecture and were designed to give multimedia and communications applications a boost.
(Image Credit: Microprocessor.sscc.ru)
Clockspeed: 60MHz - 233MHz
The last clone in the true Clone Wars, AMD's Am486 debuted almost a full four years after Intel's 486 came out, and one month after the Pentium. To compete with the existing 486 chip, AMD undercut the competition by selling its version for less, while clocking it higher than Intel's 486. Some of AMD's faster 66MHz chips even gave Intel's newly released Pentium a run for its money.
(Image Credit: liafa.jusseu.fr)
Despite the minor nomenclature update, the Pentium Pro represented a pretty significant update over the original Pentium, bringing with it a new microarchitecture rather than just a set of minor enhancements. The Pro added another million transistors to its die (5.5 million), but more importantly was the addition of its L2 cache, 256KB to start and later up to 1MB. While not yet integrated into the processor core itself, the Pentium Pro's L2 cache still ran at the same clockspeed as the CPU, anywhere from 150MHz to 200MHz.
But while the Pentium Pro's L2 cache gave it an advantage, it also presented problems for the chip. Putting the cache on a separate die introduced manufacturing defects that led to lower yields, and ultimately higher prices. This might have been okay, had it not been for the fact that the Pentium Pro shined brightest on 32-bit OSes when 16-bit still reigned supreme.
(Image Credit: Tomshardware.com)
Clockspeed: 150MHz - 200MHz
Still a relative newcomer to the x86 market, Cyrix proved it wasn't a one-hit wonder by following up the Cx486 series with its Cx5x86. And once again Cyrix preyed on end-users looking for a drop-in replacement to their existing setup by making the Cx5x86 compatible with 486 Socket 3 motherboards. The same couldn't be said for Intel's Pentium chip, giving Cyrix a leg up in this respect.
Stability issues forced Cyrix to disable some of the features it touted for its new series, including branch prediction and other performance enhancements. However,the Cx5x86's time on the market ended prematurely not because of design issues, but because Cyrix didn't want the chip cutting into sales of its 6x86 chip, which was released just six months after the 5x86.
(Image Credit: cpu-world.com)
Date Released: 1995
Clockspeed: 100MHz - 133MHz
Cyrix rated the speeds of its chips rather liberally, and very few of the Cx5x86 processors actually ran at 133MHz.
Offering an easy upgrade path for 486 computers, AMD's Am5x86 was really a 486DX with an internal x4 multiplier. This allowed the chip to run at 133MHz and ensured compatibility with most existing 486 motherboards, while offering performance on par with ,and sometimes better than, Intel's Pentium 75.
But what really stands out about the Am5x86 is that it became the first chip to make use of AMD's Performance Rating (PR), a tactic that would play an even bigger role later on. In this case, AMD sold the processor as an Am5x86-P75, letting customers know it was AMD's equivalent of a Pentium 75.
(Image Credit: Extrahardware.cz)
Date Released: 1995
Intel, having helped spur the competition with a licensing agreement that paved the way for faster clones, didn't make the same mistake with its Pentium line. As a result, AMD (and everyone else) could no longer simply clone Intel's silicon and sell it as their own. And thus was born the K5, AMD's first attempt at a next generation CPU developed in-house.
As might have been expected, design problems reared their ugly heads forcing AMD to delay the K5's launch. After working out the kinks, AMD released the K5 in 1996. Technically superior to Intel's Pentium, the K5 contained 4.5 million transistors, five integer units, a much larger branch prediction unit, and 16KB of cache, or twice the amount as found on the Pentium.
Unfortunately for AMD, the K5 suffered from low clock rates and failed to deliver any knockout blows to Intel's Pentium, nor was it a sales success.
(Image Credit: x86-guide.com)
Clockspeed: 75MHz - 133MHz
Formerly called the M1, the Cyrix 6x86 was both pin- and voltage-compatible with Intel's Pentium processor. However, it wasn't a reverse-engineered Pentium clone and instead an original design, which made it not 100 percent Pentium compatible.
Early versions shipped with 16KB of cache and showed impressive benchmark performance, often outperforming faster clocked Pentium chips in some scenarios. This led to Cyrix adopting a Performance/Pentium Rating of its own, despite its comparatively poor FPU performance.
Later versions of the 6x86 would be renamed MII. The MII revision brought with it lower heat output, paving the way for faster clockspeeds. This sometimes came at the expense of compatibility because it required non-standard bus speeds at 75MHz or 83MHz on socket 7 boards.
(Image Credit: Recycledgoods.com)
Date Released: 1996
Clockespeed: 80Mhz - 385MHz
While AMD's K5 turned out to be a forgettable CPU, the K6 rollout went much more smoothly and to a warmer reception. This was thanks in part to co-development efforts by Vinod Dham, who is known as the "Father of the Pentium" for his work in developing the Pentium CPU. Dham left Intel in 1996 before ultimately landing at a company called NexGen, later acquired by AMD. NexGen actually designed what would be the K6, which included MMX instructions and an FPU.
Launched in April 1997, the K6 served as a drop-in replacement for Pentium's Socket 7 motherboards. The K6, along with the NexGen acquisition, once again underscored AMD's position as a major chip competitor.
Clockspeed: 166MHz - 300MHz
To improve yields, Intel moved the L2 cache to an external cache chip. Doing so meant running the cache at half the speed of the CPU, which Intel tried to negate by doubling the amount of L2 cache from 256KB to 512KB on the lowest end Pentium II. This not only eventually led to lower prices (Pentium II cost a pretty penny at launch) and sub-$1,000 PCs, but also prompted Intel to package its new processor in a Single Edged Contact Cartridge to be plugged into the new Slot1 motherboards.
From a design standpoint, the Pentium II first showed up using a 0.35-micron manufacturing process that was later reduced to 0.25-microns, contained 7.5 million transistors, and could address 64GB of memory.
Additionally, the Pentium II also gave birth to the first Xeon-brand processors, released in 1998. But unlike the regular Pentium II, the Xeon version ran its L2 cache at full speed, up to 2MB of it.
(Image Credit: isosystems.eu)
: 1997 (Xeon in 1998)
Clockspeed: 233MHz - 450MHz (Xeon 400MHz - 450MHz)
Facing financial trouble, Cyrix found relief when it was bought out by National Semiconductor in 1997. It also found a change in philosophy, as National Semiconductor was much more interested in the value market than it was in trying to compete at the high end. The result of this new mentality was the Media GX, a processor based on the Cyrix 5x86 with integrated graphics, memory controller, and PCI controller. This would be paired with a companion chip that would contain the IDE controller, sound functions, and other tasks.
(Image Credit: ukcpu.net)
Date Released: 1997
Clockspeed: 120MHz - 300MHz
Pat yourself on the back if you remember the WinChip or, better yet, can follow the cluster**** of business acquisitions and collaborations between VIA, Cyrix, National Semiconductor, IDT, and Centaur Technology, among others who are in some way intertwined.
In this case, Centaur Technology produced and sold the WinChip, a Socket 7 processor. Deviating from traditional x86 processor design, Centaur used what it knew about RISC processors and created a chip with a smaller gate count and reduced die size. It was a simple, energy efficient design best suited for less demanding tasks. It didn't have any L2 cache, but did have 64KB of L1 cache. It also supported MMX and 3DNow!, but Intel's low-cost and faster performing Celeron ended any hopes Centaur might have had for the WinChip.
(Image Credit: watch.impress.co.jp)
Date Released: 1997
Clockspeed: 180MHz - 250MHz
Intel had done well serving the higher end and server markets up through the Pentium II and Pentium II Xeon, but the company lacked a true entry-level chip targeted specifically at the value PC sector. Enter the Celeron, a much lower performance part with a much lower price that first appeared in 1998.
Later in the x86 game, some specific Celeron models would make a compelling alternative for end-users looking to save a buck or three without sacrificing too much performance, particularly for those willing to overclock, but first-run Celerons based on the Pentium II core met with poor initial reaction from the general populace. This was due, in part, to the lack of L2 cache at debut, a trait which helped cripple performance. Later on, Intel would release another version with 128KB of L2 cache, and this would, in some cases, double the performance over the cacheless version. The combination of full speed L2 cache and ability to run beyond its stock speed made these second-run Celeron chips a hit among the overclocking crowd.
Throughout the years, Intel's Celeron line has accompanied Intel's mainstream processors, with the latest Celerons being built around the Allendale architecture and sporting two-cores.
Clockspeed: 266MHz - 3.2GHz
Continuing the success of the K6, AMD's K6-2, released in 1998, brought another MMX unit to the table as well as a new SIMD instruction set famously known as 3DNow! This gave AMD a slight head start in tearing through 3D applications before Intel fired back with its SSE instruction set. The K6-2 held appeal as a cost-conscious upgrade for Super Socket 7 motherboard owners.
Later on, AMD would follow suit with the K6-2+, which added 128KB of L2 cache and a smaller manufacturing process (180nm versus 250nm).
(Image Credit: CPU-World.com)
Clockspeed: 233MHz - 50MHz
SIMD, aka 3Dnow!, stands for "Single Instruction, Multiple Data." It's also known commonly as "vector instructions."
The last of the K6 line, AMD's K6-3 showed up in early 1999 and was the last Socket 7 processor ever made. K6-3 didn't have much time to bask in the the limelight, as Intel released a new processor, the Pentium III, just a few days after AMD's launch.
Essentially a K6-2 with 256KB of L2 cache and more than twice as many transistors (21.3 million versus 9.3 million), the K6-2 was initially successful, but quickly forgotten once AMD released its Athlon series.
(Image Credit: Knowplace.org)
Clockspeed: 350MHz - 570MHz
Things really started cooking for Intel with the release of the Pentium III in 1999. The addition of SSE instructions led to the ability to process up to four single-precision floating point numbers simultaneously and it was better at handling 3D, imaging, streaming content, and other multimedia tasks than the Pentium II.
Later on, Intel would release the Pentium III Coppermine. Coppermine featured 256KB of integrated full-speed L2 cache, a tweaked pipeline, and other enhancements that led to a sometimes significantly faster processor than the first Pentium III.
Yet another PIII chip -- called Tualatin -- would appear, which boasted higher clockspeeds, more cache, a die shrink, and lower temp voltage requirements leading to lower temps. Tualatin would provid the initial framework for Intel's mobile Pentium-M processor, which would later go on to inspire today's Core i7 CPUs.
As for the Pentium III Xeon, Intel's server chip didn't differ dramatically from its desktop brethren, although later run PIII Xeons would add more cache (up to 2MB) and support quad-processor configurations.
(Image Credit: Deskpicture.com)
Clockspeed: 450MHz -1.4GHz
Arguably the most significant series in AMD's CPU history, and certainly the most important in the company's recent history, AMD's Athlon line hit Intel square between the eyes and was such a success, even the Intel faithful found themselves building an AMD system for the first time. Dirk Meyer, who would later rise in rank as AMD's CEO, led the design team that developed Athlon, at first a cartridge-based processor with 512KB of L2 cache. Debuting at 500MHz, AMD beat Intel to the 1GHz mark with its Athlon processor, an important (and much anticipated) milestone at the time.
Over time, AMD would tweak its Athlon processor for the better, starting with the Thunderbird. The new core revision brought with it faster cache along with a few other tweaks that made it highly desirable. It also kicked off AMD's Socket A (462), one of the most successful motherboard sockets of all time.
(Image Credit: Wikipedia)
Date Released: 1999
Clockspeed: 500MHz - 1.4GHz
An evolution of the Media GX processor, the Geode picked up where Cyrix left off, but not for long. In 2003, National Semiconductor sold its Geode business to AMD, who continued to tweak the system-on-a-chip processor. Earlier versions can be found in some OLPCs, however AMD's latest Geodes (Geode NX) are based on the company's Athlon XP Thoroughbred core and include 256KB of L2 cache. It can also operate at up to 1GHz with passive cooling.
As of this year, AMD has stopped working on the Geode line.
Date Released: 1999
Clockspeed: 166MHz - 1.4GHz
A newcomer to the x86 chip market, Transmeta's Crusoe debuted in 2000 amid much hype. Crusoe was designed as an energy conscious, cool-running mobile part consuming anywhere from 1W to 3W during typical usage. At first built on a 180nm manufacturing process (and later 130nm), it owed much of its power savings to a software emulation layer. This, combined with the lack of emulation for SSE instructions, led Intel's Don MacDonald to say "You should check whether the Transmeta chip is 100 percent x86-compatible."
Over time Transmeta would revise its Crusoe chip, but the lack of comparable performance to Intel's and AMD's offerings combined with only negligible real-world energy management led to limited success for Transmeta's x86 CPU. So in 2004, Transmeta released a second x86 chip, the Efficeon. The new Efficeon microarchitecure was based on a 256-bit VLIW (Very Long Instruction Word) processor rather 128-bit like the Crusoe. Through Morphing Software, it also added much better x86 software compatibility, including MMX, SSE, and SSE instructions.
All told, Efficeon offered significant performance gains over Crusoe, sometimes as high as 200 percent, but faced increasing competition from Intel and AMD in the mobile market. After losing hundreds of millions of dollars over several years, Transmeta stopped making chips and instead focused on selling its technology. In January 2009, Transmeta was acquired by Novafora.
(Image Credit: x86-guide.com)
Date Released: 2000
Clockspeed: 300MHz - 2GHz
Cyrix once again passed hands, this time having been sold to VIA in 1999 who then released the Cyrix III x86 CPU in early 2000 for Socket 370 motherboards. Cyrix had already been working on the Cyrix III, but several design issues led to VIA issuing a core revision reducing the number of transistors from 22 million to 11 million. By doing so, the Cyrix III was primed to reach higher clockspeeds, which would come to label the chip in place the Performance Rating Cyrix had previously been using.
Yet another revision, codenamed Samuel 2, would add 64KB of L2 cache to the chip and switch from a 180nm to 150nm manufacturing process, again paving the way for faster clockspeeds. VIA would also later change the name of the Cyrix III to simply C3, as Cyrix technology was no longer part of the architecture.
Date Released: 2000
Clockspeed: 350MHz - 1.4GHz
Making a bid for the performance crown is only half the battle, and so in 2000 AMD released its Duron processor to compete with Intel's Celeron and win over the budget market. Essentially a crippled Athlon Thunderbird, early Durons featured a slower frontside bus at 100MHz and, as is almost always the case with low-cost chips, reduced cache. Durons came with just 64KB of L2 cache at a time with 256KB and 512KB were becoming the norm. These ranged in frequency from 600MHz to 950MHz.
Second generation Durons would be based on the Athlon XP architecture and add SSE support, and a final Duron revision based on the Thoroughred Athlon XP would see a faster frontside bus (133MHz) and clockspeeds up to 1.8GHz.
(Image Credit: x86-guide.com)
Date Released: 2000
Clockspeed: 600MHz to 1.8GHz
Sometimes change is a good thing, and other times -- as is the case here -- change can turn out to be bad. But even worse is when everybody knows something is bad, yet nothing is done about it for a long, long time. Enter the Pentium 4 processor and Intel's new NetBurst architecture.
At the risk of oversimplifying, Intel's Pentium III processor thrived on a highly efficient design that, had Intel continued to tweak, probably would never have allowed AMD to build as big of an enthusiast following as it has done. Instead, Intel left the door open for AMD by fixating on increasingly higher clockspeeds, a goal it planned to reach by introducing an exceedingly long-staged pipeline in the Pentium 4. While this paved the path for higher clockspeeds, a longer stage pipeline meant a bigger performance penalty when a set of instructions had to be thrown out and started over from scratch. Imagine putting almost all the pieces of a car together on an assembly line, only to find out towards the end that the windshield doesn't fit. But rather than ordering a new windshield, you have to scrap the entire car and all the work that went into it.
Pentium 4 wasn't all bad, though, and it did introduce both SSE2 and SSE3 instruction sets. Combined with HyperThreading, the Pentium 4 excelled with multimedia and content creation tasks, as well as code optimized for the new core. And with 3D graphics cards continuing to increase in power, a P4 chip provided a serviceable foundation for gaming rigs. Overclockers took a keen interest in the Northwood core released in 2002. With a capable motherboard and RAM, even beginning overclockers could set their sights on 1GHz overclocks using air cooling.
Still, for a Pentium 4 to really shine, it needed to ramp up clockspeeds to unprecedented levels. Intel envisioned this happening with the much anticipated Prescott core, the first chip to be built on a 90nm manufacturing process. But Prescott ran uncomfortably hot at launch, offered negligible performance gains, failed to live up to its pre-release hype, and was getting trounched by AMD silicon in gaming benchmarks.
(Image Credit: windowsdevcenter.com)
Clockspeed: 1.40GHz - 3.8GHz
Still part of the Athlon family, AMD's XP revision would add SSE instructions and represent a more aggressive approach to marketing. XP stood for eXtreme Performance and tied in nicely with Microsoft's Windows XP operating system, but it didn't stop there. AMD also went back to using a Performance Rating (PR) system for labeling its processors. Officially, AMD's PR rating (yes, we realize that's redundant) was used to denote an XP chip's performance to that of an equivalent Thunderbird core, so an AMD Athlon XP 1800+ would, in theory, offer the same performance as a Thunderbird running at 1.8GHz (1,800MHz). However, in practice it was much more widely used as a point of reference next to Intel's Pentium processor, even if this was incorrect, leading many to refer to the abbreviation as a Pentium Rating.
Yet another revision -- the Thoroughbred, or T-Bred -- would be released and drop the manufacturing process down from 180nm to 130nm. Later models would also increase the frontside bus from 100MHz (Thunderbird) and 133MHz (XP) to 166MHz (T-Bred).
But the most popular Socket A Athlon was the one built around the Barton core. Barton chips first appeared in 2003 and represented a tremendous value for enthusiasts and overclockers alike. Of particular interest were early revision Barton 2500+ processors, which shipped with an unlocked multiplier. By upping the multiplier, most Barton 2500+ chips could easily be transformed into AMD's flagship 3200+ model. Not only were Barton CPUs affordable, but so were the high-end motherboards they ran on, namely the Asus A7N8X Deluxe and Abit NF7-S Rev2, the two top overclocking boards of the time. When AMD began locking the multiplier, these and other top-end boards allowed the 2500+ to run like a 3200+ with a quick frontside bus adjustment.
From a technical standpoint, the Barton core doubled up its L2 cache to 512KB and increased the number of transistors from around 37 million (depending on the core) to 54.3 million.
Date Released: 2001
Clockspeed: 650MHz - 2.25GHz
Picking up where the Duron left off, AMD's Sempron brand replaced the Duron as the company's budget chip and Celeron competitor. And like the Duron, Semprons featured reduced L2 cache, or at least most of them did. Standing as an oddity among the rest was the Sempron 3000+ Not much more than a name change at first, early Semprons were basically Athlon XP chips with some of its cache disabled. However, the Sempron 3000+ contained 512KB of L2 cache and ran at 2.0GHz with at 166MHz frontside bus. For all intents and purposes, the Sempron 3000+ was really a Barton 2700+, if such a chip existed. It contained the same amount of L2 cache as a Barton, ran the same speed frontside bus, and was clocked between a Barton 2600+ (1.9MHz) and 2800+ (2.08GHz).
Semprons would continue to evolve alongside (and under) AMD's mainstream processor lines and still exist today.
Date Released: 2004
Clockspeed: 1.4GHz to 2.3GHz
If you're an AMD loyalist, you may want to skip this section, as it's like remembering an old friend before he was later gunned down. AMD's highest point came with with release of its Athlon 64 series, the first 64-bit processors aimed at mainstream users. While Intel was busy trying to ramp up its NetBurst-based P4 processors, AMD stole the performance crown with a much more efficient architecture and an integrated memory controller.
Not without some intial growing pains, A64 was first made available for Socket 754 motherboards, which lacked dual-channel memory support, and Socket 940, a server-oriented socket requiring buffered RAM. But like a perfect storm, Socket 939 brought with it dual-channel memory support, PCI-E, and a long-term upgrade path that helped cement AMD as the performance leader.
Even though A64 offered native 64-bit support, it was also fully backwards compatible with 32-bit code without any noticeable performance penalty. This was huge for Windows users, who were still living in a 32-bit world (the same holds true today, although 64-bit Vista is lightyears ahead of 64-bit XP).
Date Released: 2004
Clockspeed: 1.0GHz to 3.2GHz
The ill-fated NetBurst architecture made its last stand in Intel's Pentium D brand. Pentium D processors consisted of two single-core CPU dies fused to a multi-chip module. While not as elegant as AMD's dual-core design, the Pentium D offered decent multitasking performance, good overclocking performance, and comparatively low prices. The Pentium D provided a solid alternative to the Intel faithful who refused to invest in AMD.
Clockspeed: 2.66GHz to 3.73GHz
Continuing its dominance on the desktop, AMD's Athlon 64 X2 series consisted of two CPU cores on a single die sharing a crossbar that connects them to the integrated memory controller. These internal data links paid huge performance dividends compared to Intel's dual-core configuration, which had each core pushing communication through a shared frontside bus. SSE3 instructions were added to the X2 series, but most importanly, AMD managed to keep the new chip on Socket 939. While not all boards were compatible, many 939 mobos could handle an X2 upgrade with nothing more than a BIOS update, which meant AMD could tap into an existing install based for its new processors.
(Image Credit: Flickr Fr3d.org)
Date Released: 2006
Clockspeed: 1.0GHz to 3.2GHz
The Athlon 64 4000+ was the last single core model in the Athlon 64 series, but single-core Athlons lived on in the FX line.
Waking out of its Netburst slumber, Intel took the CPU world by storm with its Core 2 architecture. Instead of remaining fixated on higher clockspeeds, Intel refocused its attention on being more efficient with its pipeline. This meant a return to lower clockspeeds, however it also meant a return to prominence as the performance king. After Prescott failed to live up to its hype, the media remained cautiously optimistic that Core 2 could live up to Intel's promised performance gains, but much to the chagrin of AMD, Core 2 lived up to its billing, and then some.
The first Core 2 Duos burst out of the gates with 167 million transistors, a 65nm manufacturing process, 2MB of L2 cache, and a 1,066MHz frontside bus. Despite debuting at just 1.86GHz and 2.13GHz (E6300 and E6400, respectively), Core 2's performance made it instantly attractive, and Intel's aggressive pricing sealed the deal.
Later on, Core 2 would move to a 45nm manufacturing process with its Penryn revision, pack up to 820 million transistors into a quad-core package, and reach as high as 3.2GHz.
(Image Credit: Flickr BodHack)
Clockspeed: 1.8GHz - 3.2GHz
Intel actually made single-core Core 2 chips for its mobile line, based on the Merom and Penryn designs.
It might seem strange to resurrect the Pentium name at this stage in the game, but that's exactly what Intel did. Somewhat confusing, the Pentium Dual-Core is based on Intel's Core technology, and not earlier Pentium chips, nor is it a derivative of the Pentium D.
The first Pentium Dual-Core processors took aim at the notebook market before later being ported over to the desktop. These were intended to fill a gap between the Celeron and Core 2 series.
(Image Credit: TomsHardware)
Clockspeed: 1.4GHz - 2.8GHz
Having given up the performance crown to Intel's Core 2 architecture, AMD's previous success led to optimism that something special was brewing with Barcelona, the codename for what would later be called Phenom. Hype was so high that it may have been impossible to live up to even if Intel never released Core 2, but a delay in Phenom's release would only foreshadow more trouble to come.
When Phenom was finally released, it failed to win back the performance crown, even though Core 2 had been out for a year. Early Phenom revisions also contained a bug, and overclocking efforts were often met with frustration. And if that weren't enough, Intel's Nehalem architecture was just around the corner.
Lest we give the wrong impression, Phenom wasn't (and still isn't) a bad architecture when judged on its own merits. Several SIMD instruction sets were present, including MMX, Enhanced 3DNow!, SSE, SSE2, SSE3, and SSE4a, it contained four cores, and performance overall was very good. It just wasn't on the level of Intel's latest silicon, made even worse (for AMD) by Intel's aggressive pricing strategy.
Date Released: 2007
Clockseepd: 1.8GHz to 3.0GHz
AMD's quad-core Phenoms were the first true monolithic quad-core chips, a feature mirrored in Intel's Core i7 CPUs.
If you're an AMD fan, go ahead and cue the Evil Empire music. The Core i7's march onto the desktop has spelled nothing but trouble for AMD, who is still struggling to keep up with Intel's previous generation Core 2 architecture. Meanwhile, Core i7 (formerly known as Nehalem) stands in a league of its own.
Adding insult to injury, Intel borrowed a page from AMD and finally retired the traditional frontside bus in favor of a QuickPath Interconnect, which is Intel's equivalent of AMD's HyperTransport. This point-to-point interconnect allows for much faster communication between the CPU and various subsystems. This has also meant that overclockers have had to 'relearn' how to overclock, which includes learning several new terms.
So far there are just three Core i7 processors available -- Core i7-920, Core i7-940, and Core i7-965 -- each built on a 45nm manufacturing process with 731 million transistors and 8MB of L2 cache.
Clockspeed: 2.66GHz - 3.2GHz
Many feel that Phenom II is what the original Phenom should have been all along. With triple the amount of L3 cache (6MB versus 2MB), DDR3 support, and the elimination of the a 'cold bug' that was the bane of extreme overclockers, the Phenom II closed the performance gap between itself and Intel's Core 2 line. The only problem with that is Intel has since moved on to Core i7, which currently is untouchable by the best AMD has to offer.
Unable to compete for the performance crown, AMD has been forced to price its processors well below where it would like. Whereas the Athlon 64 X2 chips had a tendency to run high, AMD's flagship processor today, the Phenom II X4 940, streets for just $215, well below the $1,000 mark that flagship processors typically command.
(Image Credit: Flickr JoongDal)
Date Released: 2008
Clockspeed: 2.5GHz - 3.0Ghz
You may have noticed we left a few mobile-specific processors off our x86 timeline and have instead brought them up where appropriate. But we can't ignore Intel's Atom series, which has been a driving force in the uber-popular netbook (mobile) and nettop (desktop) sectors. Why is this important? Because despite a global economic downturn, worldwide PC sales have remained on an uptick thanks in large part to the explosive growth of netbooks, the vast majority of which sport an Intel Atom processor inside.
On the hardware front, these low-power chips only boast 47 million transistors, 512KB of L2 cache, and a top clockspeed of 1.86GHz. A dual-core variant exists for the desktop, but so far not for mobile PCs.
Clockspeed: 800MHz - 2GHz
Almost 15 million Atom-based netbooks were shipped in 2008, with growth expected in 2009.
While Intel's Atom series appears to have a death grip on the low-power computing market, don't count VIA out. VIA's Nano line might not enjoy the same level of sales as the Atom, but clock for clock, some benchmarks show the Nano performing better, albeit while also consuming slightly more power.
Available anywhere from 1GHz to 1.8GHz on a 533MHz or 800MHz frontside bus, VIA's Nano includes up to a whopping 1MB of L2 cache. It also supports several instruction sets, including MMX, SSE, SSE2, SSE3, and SSSE3.
Furthing adding to Nano's appeal is the promise of a dual-core variant aimed at netbooks in 2010. Should it beat Intel to the punch -- and all indications suggest it will -- the Nano could prove to be a game changer.
Date Released: 2008
Clockspeed: 1GHz - 1.8GHz