Last Updated on March 4, 2026 by Dave Farquhar
What do you do if you are an Intel competitor in 1995, facing an aggressive marketing campaign spinning 5th generation CPUs as a necessity, and your own 5th generation CPU is 18 months away from delivery? You hot rod a 486 CPU core, call it 5th generation, and hope for the best. That is what the AMD 5×86 CPU, announced in late September 1995, was. AMD released it in November of that year.
Same name, different internals
Both AMD and Cyrix released CPUs they called 5×86. But they were not at all related.
The Cyrix 5×86 design had a better claim for being a 5th generation CPU in a 4th generation package. It had many of the performance improvements the Pentium CPU had, but not all of them were enabled. The 5×86 also implemented some, but not all, of the new instructions introduced in the Pentium. Cyrix also had a weaker floating point unit than Intel’s.
Cyrix only sold its 5×86 for about six months, discontinuing it as soon as its 6×86 CPU, which was pin-compatible with Intel’s Pentium, was ready.
AMD’s 5th generation design wasn’t as far along as the Cyrix equivalent, so its 5×86 design was 5th generation in name only. It was a 486DX4 with a 4X multiplier implemented, as opposed to the 3X multiplier in a normal DX4, so you could plug it into a 486 with a 33 MHz bus, and it would run internally at 133 MHz. Give it 16 KB of level 1 cache, and it was still a good 35% slower than a Pentium at an equivalent clock speed, but that meant a 133 MHz 5×86 could keep pace with a 90 or 100 MHz Pentium.
Naming convention aside, the AMD 5×86 was the spiritual descendant of the 386DX-40 in every other possible way: Take Intel’s previous-generation design and push it to higher speeds than Intel was willing to do in that generation. These were not AMD’s glory days.
The P rating
The problem for both companies was they couldn’t keep pace clock for clock with Intel, at least not in this generation. And they both expected to be able to slightly outpace Intel clock for clock in their next generation, but not necessarily reach 200 MHz like Intel was. AMD didn’t outpace Intel in both regards until the Athlon generation.
So the two competitors agreed to cooperate. They agreed on a benchmark to use, then the benchmarked their products against Intel. Initially, they called it a P rating, labeling a CPU as a 5X86–P75 or 5×86-P90. The problem was, everyone called a 90 MHz Pentium a P90. Intel couldn’t claim a trademark on the letter P followed by a number, but both companies knew it would hurt them more than it would hurt Intel if it tried. So they quickly backed off from that naming scheme and used the letters PR and an equivalent Pentium clock speed. Officially, PR stood for “Processor Rating.” Intel had successfully sued Chips & Technologies and UMC out of the x86 processor business and had previously sued both AMD and Cyrix before.
Arguably this was collusion. But since the two companies combined still had significantly less than 50% of the x86 CPU market, it would have been hard to call it anticompetitive. So they didn’t have to worry about antitrust issues. If Intel had cooperated with one of them but not the other to create such a similar rating, that probably would have raised antitrust issues.
AMD poked Intel by dropping the letter x in “586” the same way Intel dropped the letter “e” in its trademark at the time.
Advantages of the AMD 5×86
The advantage of the AMD design was that since it was just a hopped up 486, the changes to make a motherboard work with it were minor. The motherboard needed to be aware of the 4X multiplier and aware of the 16 KB level 1 cache. But outside of that, it was just a fast 486. In some cases, the BIOS just worked, and even gave the improved performance, just misidentifying the chip. If that was the case, the motherboard just needed a cosmetic change. In other cases, it acted like a somewhat slower 486 and needed a bit more adjustment.
Due to having some Pentium-like features, the Cyrix design needed some BIOS changes to work with existing motherboards. But it wasn’t long before all of the major motherboard manufacturers were supporting both types of 5×86 CPU.
Disadvantages of the AMD 5×86
The disadvantage of both 5×86 chips was that they were a dead end from an upgrade standpoint. If you bought one in 1995 or 1996, you were going to have to swap the motherboard in order to upgrade in the future. As long as you knew this going in, it was a reasonable deal. As I recall, a 5×86 CPU and motherboard cost about half as much as a Pentium motherboard and CPU, so you were betting that by the time you needed to upgrade, a true Pentium class motherboard and CPU would cost half as much as it did in 1995.
The cost of Socket 7 motherboards and CPUs took a plunge in 1997, so the people who took that bet did fine.
Because of these limitations, you didn’t typically see the AMD 5×86 in big name computers, especially the brands sold at retail. It was the small independent computer stores that sold the majority of 5×86 CPUs. These stores were very capable of swapping the motherboard for you when the time came to upgrade.
The 5×86’s legacy
This CPU really was a stopgap solution to give AMD a halfway marketable product to sell until the AMD K5 CPU was ready, which finally happened in the spring of 1996. But it helped AMD stay afloat until its K5 and K6 CPUs were ready. Plus, the faster 486 class CPUs like the 5×86 survived for many years after they went obsolete as desktop CPUs. AMD continued to sell the 5×86 through 1999. They also used the 5×86 as the basis for its AMD Elan microcontroller, which it produced until 2003.
Ironically, the two 5×86 cores crossed paths in 2003. AMD discontinued the Elan in 2003 when it acquired the Geode CPU from National Semiconductor. The Geode was originally a Cyrix design, based on Cyrix’s 5×86 CPU core with MMX instructions bolted on.
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.