The K5 is AMD’s first x86 processor to be developed entirely in-house. Introduced March 27, 1996, its primary competition was Intel’s Pentium microprocessor. The K5 was an ambitious design, closer to a Pentium Pro internally than a Pentium. However, the final product was closer to the Pentium regarding performance, although faster clock-for-clock compared to the Pentium. The “K” in the name K5 stood for Kryptonite. AMD, being an underdog compared to Intel, hoped the K5 be the undoing for the Superman-like Intel.
Why AMD built the K5

I’ve seen speculation that AMD built the K5 because it wasn’t good at reverse-engineering Intel CPUs. This is based on the observation that AMD’s 386 and 486 CPUs came out several years after Intel.
But the main roadblocks for AMD’s 386 and 486 were legal, not technical. In each case, AMD reverse engineered Intel’s design in less than two years, but the courts delayed the release.
One of the stipulations of the long Intel-AMD court battle was that after the 486, AMD could no longer reverse-engineer Intel designs and produce its own versions. Any post-486 chip had to be an independent design. That was what led to the K5, AMD’s answer to the Pentium. It arrived on the market about 4 years after the first Pentium did, almost to the day. The problem was, when it arrived, its speed topped out at 90 MHz. Intel had reached 200 MHz by then. Cyrix beat AMD to market by 5 months and had already reached the equivalent of 166 MHz. Even struggling startup NexGen had reached 133 MHz.
AMD K5 technical details
The K5 had a highly parallel RISC processor architecture with an x86 decoding front-end. The K5 offered good x86 compatibility. All models had 4.3 million transistors, with five integer units that could process instructions out of order and a single floating-point unit. The branch target buffer was four times the size of the Pentium’s and it had register renaming to help overcome register dependencies. The chip’s speculative execution of instructions reduced pipeline stalls. It had a 16 KB four-way set-associative instruction cache and an 8 KB data cache.
It could execute instructions out of order and had a larger instruction cache than the Pentium. Its internal capabilities were closer to the Pentium Pro than the original Pentium.
The floating-point unit was more precise than the Pentium, but the Pentium was faster. By some measures, it was twice as fast.
On paper, the K5 was a formidable chip. In practice, it never quite lived up to its initial promise.
AMD K5 performance
The K5 was ambitious, to say the least. Where AMD’s 386 and 486 were strictly catch-up efforts, with the K5, AMD had the opportunity to take technical leadership from Intel. Although the chip addressed the right design concepts, the actual engineering implementation had issues.
The early 75 and 90 MHz K5 models performed roughly on par with the Pentium. The later models were significantly more performant clock for clock than the Pentium, and used the same PR rating as Cyrix to state their integer performance relative to the Pentium. The AMD K5-PR200 ran at 133 MHz on a 66 MHz bus. Cyrix had to run its 6×86 at 150 MHz on a 75 MHz bus to reach the same level of performance as the K5. Cyrix also struggled to reach higher clock speeds but did eventually reach 285 MHz.
Between limitations of AMD’s manufacturing process and the K5’s many levels of logic, AMD struggled to scale the K5 struggled to high clock speeds, never reaching a clock rate higher than 133 MHz. Additionally, while the K5’s floating-point performance was faster than the Cyrix 6×86, it was only about half as fast as the Pentium. This made the Pentium better than the K5 for running 3D games like Quake and professional tasks like CAD, but the Cyrix was better than the K5 for productivity applications since it could reach higher clock rates. This left the K5 with nowhere to go but the low-end, low-margin fringe of the market AMD designed the K5 to escape.
What PC brands used the AMD K5
Because the K5 was late to market and did not meet performance expectations, the K5 never gained the acceptance among large computer manufacturers that the earlier Am486 and later AMD K6 enjoyed. The K5 was five months behind Cyrix, 18 months behind NexGen, and a full two years behind Intel.
Delays in the K5 forced AMD to release the AMD 5×86 CPU as a stopgap in late 1995, which despite its name was really a 4th-generation CPU, a slightly modified 486DX4.
Budget PC maker Monorail was one of the few name brands to use the K5, using the 75 MHz model in its $999 PC from late 1996.
Other brands seemed to leave the option open, even if they never used it. When I tried a K5 in an IBM PC 350 in the late 90s out of curiosity, it recognized it as an AMD 5K86 and worked perfectly. That suggests bigger brands like IBM were willing to entertain the possibility.
Models
There were two revisions of the K5 architecture, internally called the SSA/5 and the 5k86, both released with the K5 label. SSA stood for super scalar architecture, fifth generation.
The original set of “SSA/5” CPUs had its branch prediction unit disabled and additional internal waitstates added. These issues were remedied with the “5k86”, resulting in up to 30% better performance clock for clock. The “SSA/5” line ran from 75 to 100 MHz while the “5k86” line ran from 90 to 133 MHz. However, AMD used what it called a PR rating, or performance rating, to label the chips according to their suggested equivalence in integer performance to a Pentium of that clock speed. Thus, a 116 MHz chip from the second line was marketed as the “K5 PR166”. Manufacturing delays caused the PR200’s arrival to nearly align with K6’s release. Since AMD did not want the K5 and K6 competing, the K5-PR200 only shipped in small numbers.
The K6 was released April 2, 1997. AMD quietly phased out the K5 after the K6 reached the market.
Relation to the K6
The AMD K6 is based on the Nx686 microprocessor that NexGen was designing when AMD acquired NexGen. Despite the name implying a design evolving from the K5, it is in fact a totally different design that was created by the NexGen team, and adapted after the AMD purchase.
Between the K6 design and licensing fabrication technology from Motorola, AMD was able to overcome its fifth-generation problems and be more competitive with Intel in subsequent generations, especially with the Athlon, Athlon 64, and Ryzen.

David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.

was NexGen chief processor architect Greg Favor, created the the Athlon
As far as I can tell Greg Favor left AMD around the year 2000.
is Athlon based on Next Gen and a further development of K6 or something entirely new? I know they hired Dick Meyer from Dec Alpha
The Athlon wasn’t entirely new, there was some K6 in it, but AMD did hire a lot of Alpha engineers (including Meyer) so they would have influenced the Athlon design.
so is Athlon based on a internal DEC Alpha RISC back end and a Next Gen translate code designed x86 front
The Athlon used the same bus as the Alpha but that doesn’t mean it was an Alpha internally. I don’t know the specifics of what the RISC portion of the Athlon looks like but given the short time duration between the K6 and the Athlon, the Athlon almost certainly had to build on the K6.