Cyrix was a scrappy, up and coming CPU manufacturer in the 1990s. They never had Intel’s name recognition, but for a few years they made life more difficult for its larger rivals, Intel and AMD. For a while, Cyrix processor chips were a popular choice for value-conscious PC buyers.
Cyrix contributed a lot of confusing alphabet soup to the 1990s CPU market, and their chips usually weren’t the highest-performing chips available. But they usually did provide good value for the money, even though Cyrix never was a premium brand.
More than the 6×86
The 6×86 CPU was the chip that really put Cyrix on the map, but Cyrix had products before the 6×86.
Cyrix was founded in 1988 by Jerry Rogers and Tom Brightman in Richardson, Texas, a suburb of Dallas. Its first chip, released in 1989, was a 387-compatible math coprocessor that was up to 50% faster than Intel’s chip while costing less. Rogers assembled a team of 30 engineers, most of whom came from Texas Instruments. This TI pedigree led some pundits to speculate in the early 1990s that Cyrix might have been a front for TI to lower the risk of competing with Intel. TI and Intel have some history, going back to the early 1970s. TI’s and Cyrix’s own troubled relationship put an end to that speculation after 1993.
Cyrix launched its IPO in 1993 at $13 per share, and quickly jumped to $19 per share. And by the end of 1993, Cyrix had captured 5 percent of the CPU market.
Cyrix’s first products, introduced in 1988, were 287- and 387-compatible math coprocessors that outperformed Intel’s equivalent chip. This is ironic, since Cyrix spent much of its life chasing Intel’s floating point performance. But in the 386 era, Cyrix outperformed Intel’s math coprocessor, and the combination of an AMD 386 and a Cyrix FasMath both running at 40 MHz outperformed Intel 486s running at 20 or 25 MHz while costing less.
Cyrix soon decided to enter the fray and sell CPUs, not just math coprocessors.
The Cyrix 486SLC and 486DLC
The first Cyrix CPU chips were the 486SLC and 486DLC CPUs, which were pin-compatible with the 386, not the 486, and could use the 387 math corprocessors. They had the 486 instruction set and a small 1K memory cache, so they outperformed the 386, but couldn’t really keep pace with an Intel 486 with its larger cache and more efficient data bus. Then again, Cyrix sold its 40-MHz version for $199, when Intel charged $404 for a chip clocked at 33 MHz.
The SLC and DLC CPUs mostly found use in off-brand PCs, from small companies who didn’t mind taking a chance on an unknown CPU maker, or locally owned clone shops of mixed repute. The only company with any name recognition I can find who used them was Leading Edge, a value brand whose best days were behind it.
They ran at 25, 33, and 40 MHz. The SLC was really a 486-class machine in name only. The DLC chip tended to perform one speed grade back from a comparable Intel chip. A DLC running at 40 MHz held its own with an Intel 486 running at 33 MHz. Since they plugged into 386 sockets, integrators could use 386 motherboards with Cyrix CPUs to deliver very low-cost PCs.
Intel sued Cyrix, but since other companies who had patent cross-licenses with Intel made the chips, Intel’s suit was unsuccessful. Over the course of its history, Intel sued Cyrix 17 times, though Cyrix always won in court.
IBM had a similarly named 486SLC2 CPU, and while IBM and Cyrix did collaborate at times and the chips were conceptually similar, IBM’s SLC2 chips and Cyrix’s SLC chips were not related to one another. IBM’s SLC2 chips were directly derived from Intel technology while Cyrix’s chips were independent designs.
The Cyrix 486DRX2
In 1993, Cyrix also released an upgrade processor that plugged into a 386 socket, providing a clock-doubled 486-compatible CPU. While a 386 upgraded with one of Cyrix’s upgrade chips had limitations, it did bring aging 386 systems into 486-class territory.
This chip confuses a lot of people. Did it come with a heatsink? And what was the point?
Yes, normally this chip came with a green heatsink that was glued on, and it definitely needs a heat sink to work properly. Sometimes the chip shows up without a heatsink. It’s not at all inconceivable that Cyrix sold a few that way. Cyrix was almost always desperate for cash. If someone wanted to buy a few bare 486DRX2s and resell them that way, Cyrix wasn’t in a position to say no.
But the Cyrix 486DRX2 processor was really aimed at the upgrade market. The target was individuals and businesses who had 33 MHz 386s. Replacing the whole computer wasn’t in the budget. Maybe replacing the motherboard wasn’t either. On some systems, replacement motherboards that would fit might not be available. Even if they were, swapping a processor is easier than swapping a motherboard.
Cyrix hoped to carve out a niche by undercutting the cost of replacing motherboards. But the scarcity of the chip today suggests they didn’t sell as many of them as they hoped.
Think of the Cyrix 486DRX2 processor as the overdrive chip Intel didn’t want to make. It did indeed speed up 386-based systems, but the limitations of the 386 bus kept it from performing like an Intel DX2 CPU of the same speed.
The Cyrix 486
By 1993, Cyrix released full 486-class chips that were fully pin-compatible with Intel and AMD 486s. The first was the oddball Cyrix 486S, which wasn’t quite as fast as a 486SX, but lacked a coprocessor. Later in the year Cyrix followed with 486SX and DX CPUs that were actually close enough to the Intel chips to be worthy of the name.
Unlike AMD’s design, which was derived from Intel’s design, Cyrix’s 486 was a clean-room implementation. The Cyrix 486s were about 10 percent slower than an Intel or AMD counterpart, but they were cheap. They also were a cheap alternative to an Intel DX2 Overdrive CPU since they ran at 5 volts. AMD CPUs couldn’t run at 5 volts as part of their settlement with Intel. But you could buy a Cyrix DX2 for half the price and get 90 percent of the benefit. I don’t know how many people knew that, but I upgraded a few 25 and 33 MHz 486s with Cyrix DX2s in the mid 90s to extend their usefulness.
And like AMD and Intel, Cyrix eventually released clock-tripled 486DX4 CPUs.
The Cyrix 5×86
Cyrix also released an oddball 5×86 CPU in August 1995. This was really a throwback to the 486DLC, in that it wasn’t really a 586-class CPU. It had a 486 data bus with a 4x clock multiplier and some advanced fifth-generation features, such as branch prediction, but it didn’t quite have the full Pentium instruction set. Cyrix only sold the chip for about six months, discontinuing it when the 6×86 was ready for market. AMD’s similarly named 5×86 was a less advanced CPU but it stayed on the market longer because it took longer for AMD to release a Pentium-class Socket 7 CPU.
AMD and Cyrix thrived for a time selling these “586” chips in the post-486 confusion. Intel countered by trying to position itself as the safe, reliable, compatible choice. One of the main reasons this strategy worked was because Intel started producing and selling quality motherboards.
But the Cyrix 5×86 processor core made a comeback a couple of years later, in an unexpected place. More on that in a minute.
The Cyrix 6×86/M II series
The Cyrix 6×86 was a revolutionary CPU for its time, offering about 30% better performance than a Pentium, at least when it came to integer performance. But Cyrix used its controversial “P-rating,” selling 133 MHz chips as equivalent to Intel 166 MHz parts. This could be deceptive, because the motherboard you used could affect the overall performance, and the Cyrix chip’s math performance was nowhere near Intel’s. The Cyrix chip was indeed fantastic for word processing and e-mail, but it wasn’t as good as an Intel CPU for 3D gaming. It was faster and better than AMD’s disappointing K5 CPU but was no match for AMD’s K6.
The Cyrix chips sold pretty well at first, because the performance beat Intel in some situations and AMD’s competing chip was late to market. Intel struck back by ramping up performance and lowering prices, and its better math performance helped it hang on to the enthusiast market. This relegated Cyrix mostly to entry-level machines. Compaq used their chips occasionally, as they were sympathetic to any alternative to Intel, and Packard Bell used them in its least expensive machines. So did Emachines, from time to time.
Ultimately Cyrix released three major variants of the 6×86: the initial design, the lower-voltage 6x86L, and the 6x86MX which added MMX instructions. The 6×86, later renamed the M II, couldn’t scale in clock speed like Intel and AMD, and adding more cores wasn’t really an option in the Windows 95 and 98 days, so Cyrix had to find other ways to try to compete.
The problem with Cyrix chips
Cyrix chips had a reputation for poor compatibility and/or performance, but that was often due to factors beyond its control. Since it specialized in the low-cost space, many PC makers who used Cyrix chips paired it up with the cheapest video and sound cards they could find. It wasn’t Cyrix’s fault when integrators used sub-par Sound Blaster clones and your games didn’t work right.
I knew a lot of people in the 90s who insisted that Intel chips worked better, even though the majority of the difference they perceived was because the people we knew with Cyrix CPUs had off-brand video and sound cards, while the people we knew with Intel CPUs had Diamond video cards and Sound Blasters.
When I paired up Cyrix processors with comparable name-brand components, I had no compatibility issues with them. The only difference was Intel had a better floating point unit. If you were playing 3D games, that could be important.
The MediaGX series
Once it was clear Cyrix wouldn’t match Intel and AMD for performance, it created a highly integrated CPU called the MediaGX that was essentially a system-on-a-chip that used a modified Socket 7. Released in 1997, it offered entry-level Pentium performance at best.
That’s because it’s not really a Pentium-class CPU. Cyrix took its 5×86 CPU core, which was really an enhanced 486, added the MMX instructions to it to modernize it a bit more, and ramped up the clock speed. Then they added integrated graphics and sound to it. The result allowed for very compact and inexpensive machines. It found use in subcompact laptops and very low-end desktop machines, usually available during the back-to-school season.
The Cyrix MediaGX processor had a much longer lifespan as an embedded chip. But even though it came out in the Pentium II era, it’s really just the world’s fastest 486.
Today, if you treat a MediaGX system like you would a 233 MHz 486, you’ll be happier with it than you will be if you expect 233 MHz Pentium performance out of it.
Fabless before fabless was cool
When Jerry Sanders ran AMD, he famously said, “Real men have fabs,” when comparing his company to Cyrix. AMD didn’t have as much manufacturing capacity as Intel, but they did have their own. Cyrix did not, so Cyrix designed its CPUs, then relied on other companies to manufacture them. At various times in its existence, Cyrix used Texas Instruments and SGS-Thomson (later ST Microelectronics) as contract manufacturers. In 1994, Cyrix added IBM as a third manufacturer. As part of their agreement, these companies could make the chips themselves and sell them under their own brand name. But this meant Cyrix frequently found itself competing with its own product.
Cyrix and TI traded lawsuits, with TI alleging breach of contract and Cyrix alleging TI didn’t give it enough manufacturing capacity. The two companies settled, with TI paying Cyrix and licensing its technology. Cyrix turned to IBM and ST to get the capacity it needed. Unfortunately for Cyrix, ST had difficulty making the 6×86 line of CPUs reliably, so most of the 6×86 CPUs were made by IBM.
Oddly, IBM chose to sell its share of the 6×86 chips on the open market rather than use them in its own PCs, even though IBM was never shy about busting Intel’s chops when problems occurred in its chips. IBM-branded 6×86 chips often sold at a slight discount compared to its Cyrix counterparts, although they were identical except for who’s name was etched on the top. Analysts had predicted IBM might use its share of the chips in its own PCs to cut its overhead, and anticipated it would help boost Cyrix’s reputation. Neither happened.
The end of the line for Cyrix
The legal wrangling with Intel and TI and the difficulty in securing manufacturing capacity took its toll. In 1997, National Semiconductor purchased Cyrix although production had to remain at IBM. Cyrix’s chip line didn’t thrive under National Semiconductor, however, and Natsemi never successfully transitioned production to its own fab in Portland, Maine. In 1999, Natsemi sold Cyrix to VIA Technologies but retained the MediaGX series, which it continued to sell as the Geode line. Natsemi sold the Geode to AMD in 2003, who used it in embedded applications, such as thin clients and industrial control systems. The AMD Geode GX and LX CPUs were based on Cyrix technology.
VIA acquired the design team for IDT’s x86-compatible CPUs around the same time it acquired Cyrix from Natsemi. Cyrix had better name recognition, so VIA discontinued the Cyrix design but branded the IDT-derived chips as VIA Cyrix. The VIA Cyrix chips competed successfully with AMD’s Geode GX and LX CPUs, which were derived from Cyrix technology. The remnants of Cyrix enjoyed a fairly long and successful life in the embedded market.
In 1993, Cyrix was selling its 40 MHz 486DLC CPU with a 387 coprocessor for $199. By the time Cyrix sold out to National Semiconductor, you could buy a whole computer for $399. Cyrix was much more aggressive than either of its two larger competitors at driving prices down. The 6×86 line gave Intel a bit of competition at the high end for a time, but Cyrix really was the king of good-enough/cheap-enough CPUs while it lasted.