The Big Processor Guide - AMD Cores
Advanced Micro Devices, Inc., has historically been a small but notable figure in the microprocessor market. While never matching the enormous size of long-time rival (and one-time partner) Intel, they have firmly established themselves as the second biggest player in the marketplace. In the days of the 486, they were noted for producing chips able to clock to higher speeds than those made by Intel, who originated the architecture. In more modern times they have established themselves as a leader (at least for now) in high performance, low power consumption products, and have leveraged innovations such as HyperTransport to prove that the little guy can score technological wins.
While computer processors are their most high-profile business, a substantial portion of AMD's revenues have historically come from flash memory production (since spun off as Spansion, of which they own a good portion). They are also a player in the embedded device processor market now. And July 24, 2006, they announced they would be merging with ATI, thus putting them in the chipset and video card markets.
Note: All Front Side Bus speeds are listed at the frequency of the clock that drives them; double-pumping is indicated by a "x2" and quad-pumping by an "x4". This page identifies sockets that support those features."HTT" indicates the processor uses HyperTransport.
Contents
Desktop
Mainstream
Phenom
The next generation, Phenom brings the new K10 microarchitecture to the desktop.
Toliman: Agena with one of the four cores disabled. All of the L3 cache is present. Because L2 cache in the K10 design is kept separate for the individual cores, the L2 cache associated with the disabled core is also disabled, so the total amount of L2 cache is reduced. Ditto for the L1 cache.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Agena | Phenom X3 | AM2+ | 3 | 512 KB | HTT | 65 nm | 2.4 GHz (8750) |
Agena: The first generation product, with 128 KB of L1 cache per core and a shared 2 MB L3 cache. Suffered a severe sales setback due to the TLB erratum issue. Adds support for SSE4a, which is not the same as Intel's SSE4.1.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Agena | Phenom X4 | AM2+ | 4 | 512 KB | HTT | 65 nm | 2.4 GHz (9750) |
Athlon X2
Exactly like the Athlon 64 X2, but with a shorter name. No, really.
Brisbane: A slightly later stepping, with slightly lower power consumption, than the Athlon 64 X2 Brisbane.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Brisbane | Athlon X2 | AM2 | 2 | 512 KB | HTT | 65 nm | 2.3 GHz (BE-2400) |
Athlon 64 X2
The Athlon 64 X2 is AMD's first dual core desktop processor. It was originally aimed at the enthusiast market, but (as such things do) has gradually moved toward the mainstream. These chips essentially combine two Athlon 64 cores into a single device, with some links to allow communication between the two cores. While having two cores does not mean double the performance, there are applications where it can make a substantial difference.
Brisbane: A die shrink of Windsor (now exclusively with 512 KB of L2 cache per core). As is often the case for die shrinks, Brisbane's most apparent advantage is that it produces less heat. Its reduced size also lets AMD put more dice on one wafer, which means reduced production costs. AMD hopes to eek out slightly higher clock speeds, but time will tell on that point. Brisbane was initially troubled by yield issues (probably due to AMD's fairly evolutionary manufacturing process) and by increased L2 cache latency (to provide the option of increased L2 cache sizes later on).
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Brisbane | Athlon 64 X2 | AM2 | 2 | 512 KB | HTT | 65 nm | 2.7 GHz (5200+) |
Windsor: Making the jump to Socket AM2, adds support for DDR2 memory. Dual Core.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Windsor | Athlon 64 X2 | AM2 | 2 | 512 KB | HTT | 90 nm | 2.8 GHz (5400+) |
| AMD | Windsor | Athlon 64 X2 | AM2 | 2 | 1 MB | HTT | 90 nm | 3.2 GHz (6400+) |
Toledo: Manchester, but with twice the cache.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Toledo | Athlon 64 X2 | 939 | 2 | 1 MB | HTT | 90 nm | 2.4 GHz (4800+) |
Manchester: Essentially 2 Athlon 64 Venice cores integrated into one chip.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Manchester | Athlon 64 X2 | 939 | 2 | 512 KB | HTT | 90 nm | 2.4 GHz (4600+) |
Athlon 64
The standards-bearer, the Athlon 64 is AMD's mainstream product.
Lima: Reduced-power Athlon 64. Essentially a single-core version of the Athlon 64 X2 Brisbane.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Lima | Athlon 64 | AM2 | 1 | 512 KB | HTT | 65 nm | 2.4 GHz (3800+ EE) |
Orleans: Making the jump to Socket AM2, adds support for DDR2 memory.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Orleans | Athlon 64 | AM2 | 1 | 512 KB | HTT | 90 nm | 2.4 GHz (3800+) |
San Diego: Venice with twice the cache. San Diego was first deployed in the Athlon 64 FX series, and eventually found its way into regular Athlon 64s.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | San Diego | Athlon 64 | 939 | 1 | 1 MB | HTT | 90 nm | 2.4 GHz (4000+) |
Venice: Boasting an improved memory controller, and using less power than Clawhammer and Newcastle, Venice is truly a force to be reckoned with. Venice is a Socket 939-only chip. The first Athlon 64 X2 (dual-core) chips, Toledo and Manchester, are based on Venice.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Venice | Athlon 64 | 939 | 1 | 512 KB | HTT | 90 nm | 2.4 GHz (3800+) |
Winchester: AMD's first 90 nm processors, Winchester was also the first Socket 939-only chip. With dramatically reduced power use and an improved memory controller, Winchester was a promising but short-lived step forward. It was quickly phased out in favor of Venice. Winchester brought SSE3 instructions to the line as well.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Winchester | Athlon 64 | 939 | 1 | 512 KB | HTT | 90 nm | 2.2 GHz (3500+) |
Newcastle: the immediate successor to ClawHammer, Newcastle is almost identical. However, it has half as much L2 cache and reduced P-ratings to compensate. Newcastle displaced Clawhammer very quickly.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Newcastle | Athlon 64 | 754 | 1 | 512 KB | HTT | 130 nm | 2.4 GHz (3400+) |
| AMD | Newcastle | Athlon 64 | 939 | 1 | 512 KB | HTT | 130 nm | 2.4 GHz (3800+) |
Clawhammer: the consumer version of the Opteron processors, and the world's first desktop 64-bit processors, Clawhammer chips were the first available for Socket 754, and were intended to be the successor to the Athlon XP. The original Clawhammers had 1 MB of L2 cache; later chips would be sold with only 512 KB, allegedly as a way of coping with manufacturing defects.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Clawhammer | Athlon 64 | 754 | 1 | 512 KB | HTT | 130 nm | 2.2 GHz (3400+) |
| AMD | Clawhammer | Athlon 64 | 939 | 1 | 512 KB | HTT | 130 nm | 2.2 GHz (3500+) |
| AMD | Clawhammer | Athlon 64 | 754 | 1 | 1 MB | HTT | 130 nm | 2.4 GHz (3700+) |
| AMD | Clawhammer | Athlon 64 | 939 | 1 | 1 MB | HTT | 130 nm | 2.4 GHz (4000+) |
Athlon XP
In its day, the Athlon XP was respected by the technical community, adored by the value-conscious, and revered by overclockers. With an initially slow start, the Athlon XP eventually came roaring forward to flex its muscle. The XP series were the first to bear "PR Rating" designations.
Thorton: Something of a rarity, Thortons are basically Bartons with half their L2 cache disabled and their bus speed reduced, giving them specs identical to Thoroughbred B units. A clever use of Bartons with partially-defective L2 cache.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thorton | Athlon XP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 2.2 (3100+) |
Barton: The last gasp of Socket A, allowing it to go out with a bang and not a whimper. The Barton was a modified Thoroughbred B with more cache and the capability to use higher Front Side Bus speeds. It was given higher performance ratings relative to its clock speed than its predecessors, partly because of its greater cache. Virtually all Barton chips are actually the same chip, set to operate at a different speeds- thus making it a very overclocker-friendly processor.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Barton | Athlon XP | A | 1 | 512 KB | 166 MHz x2 | 130 nm | 2.17 GHz (3000+) |
| AMD | Barton | Athlon XP | A | 1 | 512 KB | 200 MHz x2 | 130 nm | 2.2 GHz (3200+) |
Thoroughbred B: The "B" chip represented a shift: in order to increase scalability (the ability to make faster versions of the chip), an extra layer (the ninth) was added to the core. This was rather like dealing with heavy city congestion by building subways, or skywalks. All previous Athlon XP cores were declared obsolete at its release, and all Palomino or Thoroughbred A processors were gradually displaced by Thoroughbred B versions of themselves. Some Thoroughbred B chips also supported a higher FSB.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred B | Athlon XP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 2.13 GHz (2600+) |
| AMD | Thoroughbred B | Athlon XP | A | 1 | 256 KB | 166 MHz x2 | 130 nm | 2.25 GHz (2800+) |
Thoroughbred A: The "A" was added later to this successor to the Palomino, to separate it from the "B" revision. The Thoroughbred was a significantly more able processor than the Palomino, and cooler as well. This was in part due to a smaller process size.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred A | Athlon XP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 1.8 GHz (2200+) |
Palomino: The first big upgrade to the Thunderbird Athlon and the initiator of QuantiSpeed ratings, this processor offered much greater potential speed at the cost of increased heat. The Palomino produced a large amount of heat compared to its contemporaries, and gave the AMD line its current reputation of running warmer. Could not be unlocked in the same way as Thunderbird Athlons, though it was still possible.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Palomino | Athlon XP | A | 1 | 256 KB | 133 MHz x2 | 180 nm | 1.73 GHz (2100+) |
Athlon
In many ways, these were the chips that put AMD in a petition to really compete with Intel. Comparatively faster and cheaper, the Athlons firmly established AMD's foothold in the enthusiast market.
Thunderbird: The grand-daddy of the Socket A chips. It set the standards for the future generations, especially as regards its price advantage over Intel's offerings. First AMD processor with locked multipliers, though this could be undone with a common pencil.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thunderbird | Athlon | Slot A | 1 | 256 KB | 100 MHz x2 | 180 nm | 1.0 GHz |
| AMD | Thunderbird "B" | Athlon | A | 1 | 256 KB | 100 MHz x2 | 180 nm | 1.4 GHz |
| AMD | Thunderbird "C" | Athlon | A | 1 | 256 KB | 133 MHz x2 | 180 nm | 1.4 GHz |
Pluto/Orion: AKA K75. The second (and last) Athlon "Classic", Pluto was essentially a die shrink of Argon.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Pluto | Athlon | Slot A | 1 | 512 KB | 100 MHz x2 | 180 nm | 850 MHz |
| AMD | Orion | Athlon | Slot A | 1 | 512 KB | 100 MHz x2 | 180 nm | 1.0 GHz |
Argon: AKA K7. The first Athlon "Classic", Argon represented a substantial improvement over previous K6-based designs. With an improved Floating-Point Unit and more L1 and L2 cache, it helped greatly expand AMD's market.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Argon | Athlon | Slot A | 1 | 512 KB | 100 MHz x2 | 250 nm | 700 MHz |
K6-III
In essence the K6-III is a K6-2 with on-die L2 cache.
Sharptooth: AKA Model 9. Sharptooth added its own on-die L2 cache, essentially treating motherboard-mounted cache as L3 cache rather than L2 cache. While expensive to produce (its transistor count nearly matched the Athlon that succeeded it, and was far greater than the K6-II), it proved to be quite profitable. A solid performer brought to market while Intel was struggling with manufacturing issues, Sharptooth became popular even among traditionally Intel-only vendors.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sharptooth | K6-III | Super 7 | 1 | 256 KB | 100 MHz | 250 nm | 450 MHz |
K6-2
At its base, the K6-2 was the K6 with the addition of AMD's 3DNow! instruction set. The K6-2 struck a major blow against Intel by performing well and being very cheap.
Chomper: AKA Model 8. Available in several variants. The CXT variant added write combining.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Chomper | K6-2 | 7 | 1 | 0 KB | 66 MHz | 250 nm | 266 MHz |
| AMD | Chomper | K6-2 | Super 7 | 1 | 0 KB | 95 MHz | 250 nm | 333 MHz |
| AMD | Chomper | K6-2 | Super 7 | 1 | 0 KB | 100 MHz | 250 nm | 350 MHz |
| AMD | Chomper CXT | K6-2 | Super 7 | 1 | 0 KB | 66 MHz | 250 nm | 366 MHz |
| AMD | Chomper CXT | K6-2 | Super 7 | 1 | 0 KB | 95 MHz | 250 nm | 475 MHz |
| AMD | Chomper CXT | K6-2 | Super 7 | 1 | 0 KB | 97 MHz | 250 nm | 533 MHz |
| AMD | Chomper CXT | K6-2 | Super 7 | 1 | 0 KB | 100 MHz | 250 nm | 550 MHz |
K6
Built on recently-acquired NexGen's Nx86 chip, retooled for Socket 7, the K6 was AMD's bid to put itself back in competitive standing with Intel. Performance was comparable to the Pentium Pro. The K6 launched shortly before the Pentium II, but ultimately was similar in performance. For this reason P-ratings were dropped for this series. The K6 did nonetheless have the advantage of being cheaper, and fitting in the Socket 7 used by Intel's desktop offerings. The K6 added MMX support.
Little Foot: AKA Model 7. Little Foot brought the K6 to Super Socket 7, allowing increased Front Side Bus speeds (and AGP). It was also a die shrink of Model 6.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Little Foot | K6 | 7 | 1 | 0 KB | 66 MHz | 250 nm | 266 MHz |
| AMD | Little Foot | K6 | Super 7 | 1 | 0 KB | 100 MHz | 250 nm | 300 MHz |
Model 6: Little changed from the NexGen design, Model 6 saved AMD.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Model 6 | K6 | 7 | 1 | 0 KB | 66 MHz | 350 nm | 233 MHz |
K5
The K5 was AMD's attempt to compete with Intel's Pentium. And it was a dismal failure. Launched over a year late, it sold poorly. In addition to its other woes, it is noted to have produced quite a bit of heat for its time. K5 was an ambitious design, and like the Pentium Pro it contained a RISC core that translated x86 instructions into its internal "micro-ops" before execution. Confusingly, the K5 was refered to as the 5k86 until the release of the core codenamed 5k86... at which time it was refered to as the K5. The K5 was given P-ratings.
5k86: AKA Model 1. It should be noted that the PR200 is quite rare (and never entered regular distribution), and the fastest generally-available chip was the 116 MHz PR133.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | 5k86 | K5 | 5 | 1 | 0 KB | 60 MHz | 350 nm | 90 MHz |
| AMD | 5k86 | K5 | 5 | 1 | 0 KB | 66 MHz | 350 nm | 100 MHz |
| AMD | 5k86 | K5 | 7 | 1 | 0 KB | 66 MHz | 350 nm | 133 MHz (PR200) |
SSA5: AKA Model 0. Ambitious but weak, poor floating-point performance and other faults would doom this poor core.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | SSA5 | K5 | 5 | 1 | 0 KB | 50 MHz | 500 nm | 75 MHz |
| AMD | SSA5 | K5 | 5 | 1 | 0 KB | 60 MHz | 350 nm | 90 MHz |
| AMD | SSA5 | K5 | 5 | 1 | 0 KB | 66 MHz | 350 nm | 100 MHz |
Am5x86
A stopgap measure, AMD produced the Am5x86 while it awaited completion of the K5.
Am5x86: A very popular and very broadly compatible upgrade path for 486 owners. While unable to compete with the Pentium at a matching clock speed, the advantages over the 486 were substantial. Only one speed was manufactured in quantity, although a 150 MHz version reportedly exists.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Am5x86 | Am5x86 | 3 | 1 | 0 KB | 33 MHz | 350 nm | 133 MHz |
High End
Phenom Black Edition
As per usual, these chips are just the same as regular Phenoms, but with unlocked multipliers. And as per usual, they are sold at higher speeds before their "mainstream" counterparts. The major break from tradition is that they are sold at a much more modest upcharge; where a new Athlon 64 FX or Pentium 4 Extreme Edition could have cost $1,000 in its day, these have typically been at a more achievable $250 or so.
Agena: Just like the Phenom X4 Agena, but with the multiplier unlocked.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Agena | Phenom Black Edition | AM2+ | 4 | 512 KB | HTT | 65 nm | 2.6 GHz (9950 BE) |
Athlon 64 FX
Targeted at the enthusiast market, the Athlon 64 FX series is meant to represent the absolute best of AMD's desktop line- the state of the art. This sometimes means new technologies reach the FX series first, while in other cases it means simple amenities like unlocked multipliers and tighter testing tolerances. Initially offered only on the costly Socket 940 platform, the FX series eventually found its way to the more mainstream Socket 939.
Windsor: Based on the Athlon 64 X2 core of the same name, but with the guarantee of 1 MB of L2 per core and unlocked multipliers. The Socket F parts are for use in QuadFX configurations.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Windsor | Athlon 64 FX | F | 2 | 1 MB | HTT | 90 nm | 3.0 GHz (FX-74) |
| AMD | Windsor | Athlon 64 FX | AM2 | 2 | 1 MB | HTT | 90 nm | 2.8 GHz (FX-62) |
Toledo: Adapted from the Athlon 64 X2 series chip, these FX chips are essentially identical other than having unlocked multipliers, which make them friendlier to overclockers.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Toledo | Athlon 64 FX | 939 | 2 | 1 MB | HTT | 90 nm | 2.6 GHz (FX-60) |
San Diego: This was the first appearance of San Diego, before it made its way to the Athlon 64 line. The only demonstrable difference was that the Athlon 64 FX version had an unlocked multiplier.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | San Diego | Athlon 64 FX | 939 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (FX-57) |
Sledgehammer: The bastard child of Opteron, the Athlon 64 FX Sledgehammer is little different from contemporary Opteron Sledgehammers.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sledgehammer | Athlon 64 FX | 940 | 1 | 1 MB | HTT | 130 nm | 2.4 GHz (FX-53) |
| AMD | Sledgehammer | Athlon 64 FX | 939 | 1 | 1 MB | HTT | 130 nm | 2.6 GHz (FX-55) |
Value
Sempron
Continuing in the spirit of the Duron series, the Sempron chips are intended to appeal to value-conscious buyers. Semprons first appeared on Socket A, and have since migrated as needed.
Sparta: A lot like the Athlon 64 X2 Brisbane, but with just one core.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sparta | Sempron | AM2 | 1 | 512 KB | HTT | 65 nm | 2.2 GHz (LE-1250) |
| AMD | Sparta | Sempron | AM2 | 1 | 256 KB | HTT | 65 nm | 2.0 GHz (LE-1150) |
Manila: Making the jump to Socket AM2, adds support for DDR2 memory.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Manila | Sempron | AM2 | 1 | 128 KB | HTT | 90 nm | 2.0 GHz (3500+) |
| AMD | Manila | Sempron | AM2 | 1 | 256 KB | HTT | 90 nm | 2.0 GHz (3600+) |
Palermo: The replacement for the Paris core. They are basically retooled Winchester (later Venice) chips. They are on a smaller process than Paris. Beginning at the E3 stepping, they support SSE3. Beginning at the E6 stepping, they support the full AMD64 extensions; some refer to these as the "Sempron 64" series. All support the NX bit. The Socket 939 chips support an increased 1000 MHz HyperTransport speed.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Palermo | Sempron | 754 | 1 | 128 KB | HTT | 90 nm | 2.0 GHz (3300+) |
| AMD | Palermo | Sempron | 754 | 1 | 256 KB | HTT | 90 nm | 2.0 GHz (3400+) |
| AMD | Palermo | Sempron | 939 | 1 | 128 KB | HTT | 90 nm | 2.0 GHz (3400+) |
| AMD | Palermo | Sempron | 939 | 1 | 256 KB | HTT | 90 nm | 1.8 GHz (3200+) |
Paris: Modified Athlon 64 Newcastle with some cache disabled and HTT speed limited to 800 MHz. Initially, many Athlon XP-based Semprons were refered to as being Paris core, though this was incorrect.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Paris | Sempron | 754 | 1 | 128 KB | HTT | 130 nm | 1.8 GHz (3000+) |
| AMD | Paris | Sempron | 754 | 1 | 256 KB | HTT | 130 nm | 1.8 GHz (3100+) |
Barton: Essentially a remarked Athlon XP Barton.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Barton | Sempron | A | 1 | 512 KB | 166 MHz x2 | 130 nm | 2.0 GHz (3000+) |
Thoroughbred B: Essentially a remarked Athlon XP Thoroughbred B. AMD stopped marketing all Athlon XPs at the launch of the Athlon 64, presumably to avoid confusing customers, and instead turned them into Semprons.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred B | Sempron | A | 1 | 256 KB | 166 MHz x2 | 130 nm | 2.0 GHz (2800+) |
Duron
AMD's first "value" line, analogous to Intel's Celeron, the Duron was a cheap but surprisingly hardy chip.
Applebred: The last gasp of Duron, with a bump to the Front Side Bus. Despite modest performance (and modest clock speeds), able to best the entire range of regular Celerons. The Applebred was a down-tuned Athlon XP Thoroughbred (A or B) with reduced cache. In some cases this cache was defective; in others, it could actually be re-enabled.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Applebred | Duron | A | 1 | 64 KB | 133 MHz x2 | 130 nm | 1.8 GHz |
Morgan: A weakened Athlon XP Palomino.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Morgan | Duron | A | 1 | 64 KB | 100 MHz x2 | 180 nm | 1.3 GHz |
Spitfire: A down-tuned Athlon Thunderbird, Spitfire had very little L2 cache and a reduced Front Side Bus.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Spitfire | Duron | A | 1 | 64 KB | 100 MHz x2 | 180 nm | 950 MHz |
Laptop
Mainstream
Turion X2 Ultra
A new version of Turion introduced with AMD's Puma platform, the Turion X2 Ultra is meant to offer both better performance and lower power usage.
Griffin: A smaller process and more L2 cache are the most obvious changes from older products, but Griffin has more subtle differences as well. It moves from a single power plane for the entire chip to 3 (one for each of the two cores, and one for the integrated memory controller and everything else), allowing the chip to vary the voltage supplied to its different components. This means, most critically, that cores that are not very busy can be fed a lower voltage and thus consume less power. Griffin also moves to HyperTransport 3.0, which providesd more bandwidth but also the ability to change the link's "width" (from 16 lanes to 8) to save power. It is not yet clear if all versions will have 1 MB of L2 cache per core, or if some might have half that.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Griffin | Turion X2 Ultra | S1 | 2 | 1 MB | HTT | 45 nm | 2.4 GHz (ZM-86) |
Turion 64 X2
A dual-core implementation of the Turion on a new socket. The new socket brings support for DDR2 as well. AMD has indicated the Turion 64 X2 will fully replace the regular Turion 64 in a short time frame.
Taylor: Adds support for DDR2 memory. Dual core.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Taylor | Turion 64 X2 | S1 | 2 | 256 KB | HTT | 90 nm | 1.6 GHz (TL-50) |
| AMD | Taylor | Turion 64 X2 | S1 | 2 | 512 KB | HTT | 90 nm | 2.0 GHz (TL-60) |
Turion 64
Strictly speaking, Turion is a "platform" or "package" of the same breed as Intel's Centrino brand. However, the requirements are so loose that essentially any laptop with the right processor meets them- and so in practice, the name Turion 64 is applied to the chips themselves.
Lancaster: At long last, a truly mobile chip from AMD. Based on the Winchester/Venice/San Diego cores, Lancaster produces dramatically less heat than its predecessors.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Lancaster | Turion 64 | 754 | 1 | 512 KB | HTT | 90 nm | 2.4 GHz (ML-42) |
| AMD | Lancaster | Turion 64 | 754 | 1 | 1 MB | HTT | 90 nm | 2.4 GHz (ML-44) |
Athlon 64 Mobile
These mobile variants on the Athlon 64 line initially targeted only the desktop replacement market (due to the high power consumption of the Clawhammer core), but with the introduction of Lancaster began to expand their appeal. 1600 MT/s HyperTransport.
Newark: The "desktop replacement" chip based on the Athlon 64 Winchester/Venice/San Diego.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Newark | Athlon 64 Mobile | 754 | 1 | 1 MB | HTT | 90 nm | 2.6 GHz (4000+) |
Oakville: The "portability" chip based on the Athlon 64 Winchester/Venice/San Diego.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Oakville | Athlon 64 Mobile | 754 | 1 | 512 KB | HTT | 90 nm | 2.0 GHz (3000+) |
Odessa: Available as the 1.6 GHz 2800+ "desktop replacement" and several lower-power, lower-voltage versions. Based on the Athlon 64 Newcastle.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Odessa | Athlon 64 Mobile | 754 | 1 | 512 KB | HTT | 130 nm | 2.0 GHz (3000+) |
Clawhammer: the vaguely mobile variant of the Athlon 64 processor, officially launched alongside the desktop unit. Intended as a Desktop Replacement, rather than a truly mobile unit.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Clawhammer | Athlon 64 Mobile | 754 | 1 | 1 MB | HTT | 130 nm | 2.4 GHz (3700+) |
Athlon XP-M
The Athlon XP-M is not fundamentally different from the regular Athlon XP. XP-Ms are culled from the best of the best XPs, chosen largely for their ability to operate correctly with reduced voltage. They also have unlocked multipliers.
Barton: Based on the Athlon XP of the same name.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Barton | Athlon XP-M | A | 1 | 512 KB | 133 MHz x2 | 130 nm | 2.2 GHz (3000+) |
Thoroughbred B: Based on the Athlon XP of the same name.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred B | Athlon XP-M | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 2.13 GHz (2600+) |
Mobile Athlon XP
The first adaptation of the Athlon XP to the mobile market, the Mobile Athlon XP offered the promise of increased Front Side Bus speeds over the Athlon 4 Mobile.
Thoroughbred A: Based on the Athlon XP of the same name.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred A | Mobile Athlon XP | A | 1 | 256 KB | 100 MHz x2 | 130 nm | 1.6 GHz (1900+) |
| AMD | Thoroughbred A | Mobile Athlon XP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 1.6 GHz (1900+) |
Athlon 4 Mobile
A regular Athlon adapted for mobile use by way of reduced operating voltage and improved power management.
Corvette: Based on the Athlon Thunderbird.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Corvette | Athlon 4 Mobile | A | 1 | 256 KB | 100 MHz x2 | 180 nm | 1.4 GHz (1600+) |
K6-III+
Mobile variant of the K6-III. Extended 3DNow! and PowerNow!
Sharptooth: Based on the K6-III of the same name. Note that it is on a smaller process than the desktop chip.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sharptooth | K6-III+ | Super 7 | 1 | 256 KB | 100 MHz | 180 nm | 500 MHz |
K6-III-P
Early mobile variant of the K6-III.
Sharptooth: Based on the K6-III of the same name.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sharptooth | K6-III-P | Super 7 | 1 | 256 KB | 100 MHz | 250 nm | 450 MHz |
K6-2-P
Mobile variant of the K6-2. Reduced operating voltage.
Chomper: Based on the K6-2 of the same name.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Chomper | K6-2-P | 7 | 1 | 0 KB | 66 MHz | 250 nm | 333 MHz |
| AMD | Chomper | K6-2-P | Super 7 | 1 | 0 KB | 66 MHz | 250 nm | 433 MHz |
| AMD | Chomper | K6-2-P | Super 7 | 1 | 0 KB | 95 MHz | 250 nm | 475 MHz |
| AMD | Chomper | K6-2-P | Super 7 | 1 | 0 KB | 100 MHz | 250 nm | 450 MHz |
Value
Mobile Sempron
Cut-down Mobile Athlon 64s, the Mobile Sempron series is aimed for the smaller budget. With reduced L2 cache, they underperform their bigger brothers. HyperTransport speed of 1600 MT/s.
Sonora: Low power version of Georgetown targeted at thin and light notebooks.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sonora | Mobile Sempron | 754 | 1 | 128 KB | HTT | 90 nm | 2.0 GHz (3300+) |
| AMD | Sonora | Mobile Sempron | 754 | 1 | 256 KB | HTT | 90 nm | 1.8 GHz (3100+) |
Georgetown: Based on the Athlon 64 Winchester/Venice. Targeted at desktop replacements.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Georgetown | Mobile Sempron | 754 | 1 | 128 KB | HTT | 90 nm | 2.0 GHz (3300+) |
| AMD | Georgetown | Mobile Sempron | 754 | 1 | 256 KB | HTT | 90 nm | 2.0 GHz (3400+) |
Dublin: Available in desktop replacement and low power versions. It should be noted that for a time, these chips were sold under the Athlon XP-M branding.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Dublin | Mobile Sempron | 754 | 1 | 128 KB | HTT | 130 nm | 1.8 GHz (3000+) |
| AMD | Dublin | Mobile Sempron | 754 | 1 | 256 KB | HTT | 130 nm | 1.6 GHz (2800+) |
Duron Mobile
The "budget" line that sold alongside Athlons and Athlon XPs.
Morgan: A retooled Duron Morgan.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sharptooth | Mobile Duron | A | 1 | 64 KB | 100 MHz x2 | 180 nm | 1.3 GHz |
Spitfire: A retooled Duron Spitfire.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Spitfire | Mobile Duron | A | 1 | 64 KB | 100 MHz x2 | 180 nm | 700 MHz |
K6-2+
More closely related to the K6-III, the K6-2+ was essentially brought to market as a down-scaled K6-III+. Adds PowerNow!
Chomper: As the K6-III+ Sharptooth, but with reduced cache and in some cases a lower Front Side Bus.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Chomper | K6-2+ | Super 7 | 1 | 128 KB | 95 MHz | 180 nm | 475 MHz |
| AMD | Chomper | K6-2+ | Super 7 | 1 | 128 KB | 97 MHz | 180 nm | 533 MHz |
| AMD | Chomper | K6-2+ | Super 7 | 1 | 128 KB | 100 MHz | 180 nm | 550 MHz |
Server
Opteron
The Opteron line was AMD's bid to take over a substantial portion of the server market. It enjoyed much greater success than the Athlon MP line before it, in part due to its excellent thermal characteristics. After appearing on the more consumer-oriented Socket 939, Opterons enjoyed a period of considerable popularity among overclockers. The more rigorous testing (and for a time surprisingly low prices) of these processors as compared to Athlon 64 and X2 alternatives proved to be very appealing.
Barcelona: The first K10 product to market, Barcelona finally retired the K8 series of Opterons. Dual- and quad-processor-capable versions were available at launch, with single-processor parts to follow. The launch was unfortunately marred by a TLB erratum that caused AMD to scramble to do damage control.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Barcelona | Opteron | F | 4 | 512 KB | HTT | 65 nm | 2.0 GHz (2350/8350) |
Santa Rosa: Ushering in the era of "Revision F" Opterons, Santa Rosa makes the jump to Socket F. The switch came shortly after desktop chips moved to Socket AM2. Santa Rosa is available with support for 2 or 8 chips in a system (or less, of course). Santa Rosa Opterons are members of the Opteron 22xx (up to 2 processors per system) or 82xx (up to 8 processors per system) family.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Santa Rosa | Opteron | F | 2 | 1 MB | HTT | 90 nm | 2.8 GHz (2220/8220) |
Santa Ana: Almost exactly the same as Santa Rosa, but on a different socket and with a different memory controller. Santa Ana is made for Socket AM2, and consequently has a DDR2 memory controller. It also consequently supports single-processor use only. Santa Ana Opterons are members of the Opteron 12xx family.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Santa Ana | Opteron | AM2 | 2 | 1 MB | HTT | 90 nm | 2.6 GHz (1218) |
Egypt: Denmark with support for 8-socket systems. Egypt processors are members of the Opteron 8xx family, meaning they are suitable for use in quad-socket systems.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Egypt | Opteron | 940 | 2 | 1 MB | HTT | 90 nm | 2.6 GHz (885) |
Italy: Denmark with support for 2-socket systems. Italy processors are members of the Opteron 2xx family, meaning they are suitable for use in dual-socket systems.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Italy | Opteron | 940 | 2 | 1 MB | HTT | 90 nm | 2.4 GHz (280) |
Denmark: The first dual-core Opteron. Denmark is made up of two Troy cores in much the same way an Athlon 64 Manchester is made up of two Venice cores. Denmark is a member of the 1xx family, and is suitable for single-socket configurations
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Denmark | Opteron | 940 | 2 | 1 MB | HTT | 90 nm | 2.6 GHz (185) |
| AMD | Denmark | Opteron | 939 | 2 | 1 MB | HTT | 90 nm | 2.2 GHz (175) |
Athens: Venus with support for 8-socket systems. Athens processors are members of the Opteron 8xx family, meaning they are suitable for use in quad-socket systems.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Athens | Opteron | 940 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (854) |
| AMD | Athens | Opteron | 939 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (854) |
Troy: Venus with support for 2-socket systems. Troy processors are members of the Opteron 2xx family, meaning they are suitable for use in dual-socket systems.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Troy | Opteron | 940 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (254) |
| AMD | Troy | Opteron | 939 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (254) |
Venus: 90 nm manufacturing hits the Opteron. Like the Sempron Paris, a curiosity. Available on Socket 940 like other Opterons, but also on Socket 939. Part of the 1xx family, suited for single-socket use. Much like Sledgehammer Opterons, but manufactured on the newer 90 nm process.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Venus | Opteron | 940 | 1 | 1 MB | HTT | 90 nm | 2.2 GHz (148HE) |
| AMD | Venus | Opteron | 939 | 1 | 1 MB | HTT | 90 nm | 2.8 GHz (154) |
Sledgehammer: the first of AMD's 64-bit "Hammer" processors to hit the market. Unlike later cores, this generation of Opterons did not have different core names for versions supporting different numbers of sockets.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Sledgehammer | Opteron | 940 | 1 | 1 MB | HTT | 130 nm | 2.4 GHz (x50) |
Athlon MP
The MP series are virtually identical to their XP brothers, with the key difference that they could be run in multi-processor configurations. Setting the tone for the Opteron, the MP series sometimes received new technologies before they filtered down to the XP line. The Palomino core first launched under MP branding.
Barton: As the Athlon XP of the same name, but with multiprocessor support.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Barton | Athlon MP | A | 1 | 512 KB | 133 MHz x2 | 130 nm | 2.13 GHz (2800+) |
Thoroughbred B: As the Athlon XP of the same name, but with multiprocessor support.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred B | Athlon MP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 2.13 GHz (2600+) |
| AMD | Thoroughbred B | Athlon MP | A | 1 | 512 KB | 133 MHz x2 | 130 nm | 2.13 GHz (2800+) |
Thoroughbred A: As the Athlon XP of the same name, but with multiprocessor support.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Thoroughbred A | Athlon MP | A | 1 | 256 KB | 133 MHz x2 | 130 nm | 1.8 GHz (2200+) |
Palomino: As the Athlon XP of the same name, but with multiprocessor support. The Palomino core actually appeared first as an Athlon MP. Some had P-ratings and some did not (earlier ones did not); that is why this table has an extra row. Sometimes refered to as the Athlon MP Mustang.
| Manufacturer | Core | Series | Socket | Cores | L2 Cache/Core | FSB | Process Size | Top Speed |
| AMD | Palomino | Athlon MP | A | 1 | 256 KB | 133 MHz x2 | 180 nm | 1.2 GHz |
| AMD | Palomino | Athlon MP | A | 1 | 256 KB | 133 MHz x2 | 180 nm | 1.73 GHz (2100+) |
