No doubt most people are familiar with Intel and AMD, Qualcomm, Texas Instruments, and even VIA — but there’s another pioneering chipmaker that you should be familiar with. For the better part of the past decade, Cyrix, which successfully brought personal computers to ordinary people’s homes on the cheap, was scuppered by its best product and inability to run a popular game. was acquired by a larger partner.
The early 1990s were strange times for the desktop computer industry, and especially for the microprocessor world. Despite the intense competition in the processor space, it still looks like Intel is the winner. At that time, Apple switched to IBM’s PowerPC, Motorola’s 68K chips were slowly dragging Commodore’s Amiga PC family to a dead end, and Arm was just a small flame lit by Apple, VLSI technology, and Acorn Computer, which at the time almost Fully focused on developing the right processor for the infamous Newton.
Around the same time, AMD took the first step out of the negative halo of being considered a second supplier. After cloning generations of Intel CPUs, AMD came out with its own architecture, which by the late 90s was well-received in terms of price and performance.
This success can be attributed at least in part to Cyrix, a company that had a chance to take over the home PC market and kick Intel and AMD out of the game, but sadly, Cyrix failed and quickly disappeared among tech companies.
Cyrix was founded in 1988 by Jerry Rogers and Tom Brightman as a manufacturer of high-speed x87 math coprocessors for the 286 and 386 processors. They have great talent coming out of Texas Instruments, and they have huge ambitions to go up against Intel and to beat Intel on their own track.
Rogers began an aggressive search for America’s best engineers, but went on to become a notoriously tough leader, leading a team of 30 on impossible tasks.
Initially, the company’s first math coprocessors outperformed their Intel counterparts by about 50 percent and cost less. This made it possible to pair an AMD 386 CPU with a Cyrix FastMath coprocessor and get 486-like performance at a lower price, which caught the attention of the industry and encouraged Rogers to take the next step into the CPU market.
In 1992, Cyrix introduced its first CPUs, the 486SLC and 486DLC, designed to compete with Intel’s 486SX and 486DX. They’re also pin-compatible with the 386SX and 386DX, meaning they can be used as drop-in upgrades to older 386 motherboards, and they’re also used by manufacturers to sell cheap laptops.
Both products perform slightly worse than the Intel 486 CPU, but perform much better than the 386 CPU. The Cyrix 486 DLC can’t compete with the Intel 486SX clock cadence, but it’s a full 32-bit chip, has 1KB of L1 cache, and costs a lot less.
At that time, enthusiasts liked the fact that they could use the 486DLC running at 33 Mhz to achieve comparable performance to the Intel 486SX running at 25 Mhz. That’s not to say it’s not without problems though, as it can cause stability issues with some older motherboards that don’t have extra cache control lines or CPU register controls to enable or disable on-board cache.
Cyrix also developed a new “drop-in” version called the Cx486DRu2, followed by a “clock-doubling” version called the Cx486DRx2 in 1994, which integrated cache synchronization circuitry into the CPU itself.
However, by that time Intel had released its first Pentium processor, the price of the 486DX2 started to drop, and the Cyrix gradually lost its appeal as it was cheaper to upgrade to a 486 motherboard than to buy a Cyrix upgrade processor for an older 386 motherboard. When the “tripled the clock” 486DX4 came out in 1995, it was too small and too late.
Large PC makers such as Acer and Compaq did not favor Cyrix’s 486 CPUs, opting instead for AMD’s 486s. But that didn’t stop Intel from spending years accusing the Cx486 of infringing its patents, never winning a case.
Cyrix and Intel eventually settled out of court, in which Intel agreed that Cyrix had the right to produce its own x86 designs in factories that held Intel’s cross-licensing, such as Texas Instruments, IBM, and SGS Thomson (later STMicroelectronics).
Never repeat the same trick twice…unless you’re Cyrix
In 1993, Intel introduced the original Pentium processor based on the P5 microarchitecture and eventually became known by a market-friendly name. But more importantly, it raised the bar for performance, ushering in a new era of personal computing.
A superscalar architecture allows it to complete two instructions per clock, a 64-bit external data bus enables more data to be read and written per memory access, and a faster floating-point unit capable of reaching 15 of the 486 FPU throughput times. There are other advantages as well.
Cyrix once again faced the challenge of not being able to create a middle ground again for Socket 3 motherboards with new Intel Pentium CPUs until the model wasn’t ready to ship. In the middle is the Cyrix 5×86, which at 75 MHz has many of the features of fifth-generation processors like the Pentium and AMD K5.
Cyrix 5×86 CPU with heatsink. Source: NostalgiaNerd
The company even makes 100 MHz and 133 MHz versions, but they don’t really have all the performance-enhancing features advertised because enabling them can cause instability and have limited overclocking potential. All of these products were short-lived, and after six months Cyrix decided to stop selling them and move to other processor designs.
Cyrix via Quake
In 1996, Cyrix introduced the 6×86 (M1) processor, which was expected to replace the older Intel CPUs once again on well-performing Socket 5 and Socket 7 motherboards. But it’s not just an upgrade path for budget systems, it actually does some wonders of CPU design, which was considered impossible at the time – it combines many design aspects of RISC cores and CISC. Meanwhile, it continues to use native x86 execution and normal microcode, while Intel’s Pentium Pro and AMD K5 rely on dynamic translation to micro-ops.
The Cyrix 6×86 is pin-compatible with the Intel P54C, and comes in six variants with confusing naming schemes that are supposed to indicate expected performance levels, but not actual indicators of clock speed. For example, the 6×86 PR166+ only runs at 133 MHz and is marketed as equal to or better than the Pentium running at 166 MHz, which is a strategy AMD will pursue later.
The problem, though, is that the 6×86 actually identifies itself as a 486 CPU because it doesn’t support the full Intel P5 instruction set. This will quickly become a serious problem, as most application development is slowly migrating to P5 pentium-specific optimizations to take advantage of the new instructions to squeeze more performance. Cyrix finally improved compatibility with Pentium and Pentium Pro with 6x86MX and 6x86MII.
A huge selling point of the 6×86 is that its integer performance is significantly better than that of the Pentium, which is a nice advantage when most applications and games rely on integer arithmetic. For a while, Cyrix even tried to charge extra for the added performance, but that strategy failed shortly after.
Cyrix 6x86MX CPU Die Image
It turns out that the 6×86 FPU (Floating Point Unit) is just a slightly modified version of Cyrix’s 80387 coprocessor, and as such, is significantly slower than the new FPU design integrated into Intel’s Pentium and Pentium Pro.
To be fair, it’s still 2 to 4 times faster than the Intel 80486 FPU, and the Cyrix 6×86 beats Intel’s offering in overall performance. But when many software developers, especially those making 3D games, saw the growing popularity of the Pentium and opted to optimize their code in assembly language around the advantages of the P5 FPU, the whole balance collapsed.
When id Software released Quake in 1996, gamers with 6×86 processors found they were getting standard frame rates of up to 15 frames per second, unplayable unless they wanted to drop the resolution down to 320 x 200. Only the highest-end Cyrix 6x86MX PR2/200 CPU gets the 29.7 frames per second playability. But gamers with Intel systems have no problem running the game even at playable frame rates of 640 x 480.
John Carmack found that he could overlap integer and floating-point operations on the Pentium chip because the P5 core uses different sections for all operations except instruction loading. This technique does not work on Cyrix cores, exposing the weakness of its FPU. Reviewers at the time found that the 6×86 CPU would outperform the Pentium by 30 to 40 percent in all other benchmarks or performance tests.
Back in the mid-90s, when no one knew the exact direction of computing, Cyrix thought it would be better to prioritize integer performance, so it produced a processor without instruction pipelining, a feature that would become an essential part of any processor. Instruction pipelining is a technique used to divide a task into a set of smaller operations that are then executed concurrently by different parts of the processor in a more efficient manner. The FPU of the Pentium processor is pipelined, which allows floating point operations to process seismic graphics with very low latency.
The problem itself is easy to fix, and software developers have released patches for their apps and games. But id Software spent too much time designing Quake around the P5 microarchitecture and never offered such a fix. AMD’s K5 and K6 CPUs performed slightly better than Cyrix, but they were still inferior to Intel’s offerings on Quake. Quake is a very popular game and the flagship of the new generation of 3D games.
This made the Cyrix CPU’s performance gap so harsh that the company almost lost its credibility in the eyes of many enthusiasts. It’s a particularly hard blow when Cyrix’s fierce customer base is made up of these zealots, as the company can’t sign contracts with the big PC OEMs.
To make matters worse, Cyrix is a fabless chip maker that relies on third parties to make its processors, and those companies use the most advanced product lines for their own products. As a result, Cyrix processors are manufactured on the 600 nm process node, while Intel’s processors are manufactured on 300 nm.
Efficiency takes a big hit, which is why Cyrix CPUs can get so hot—so hot that enthusiasts are designing hotplates to use them as heating elements. They are too sensitive to low quality power supplies and have limited overclocking potential, but that hasn’t stopped someone like the author of this article (who has a Cyrix 6×86-P166+ CPU inside of his second PC) from pushing them a little bit, and slowly lead to their demise.
The fall of the first real rival to Intel’s CPU supremacy
By 1997, Cyrix went to great lengths to form partnerships with companies like Compaq and HP, because integrating its CPUs into systems would generate a steady stream of revenue. It also tried to sue Intel for infringing its patents on power management and registration renaming technology, but the matter was quickly settled through a mutual cross-licensing agreement so that the two companies could continue to focus on making better CPUs.
The famous National semiconductor advertisement
The lawsuit took its toll on an already cash-strapped company. Facing the prospect of bankruptcy, Cyrix agreed to merge into National Semiconductor. It’s a real blessing because it will finally have access to a proper manufacturing facility and a strong marketing team that will be able to get a lot of orders. IBM’s manufacturing deal was put on hold for a while, but Cyrix eventually moved all production to National Semiconductor.
However, it turns out that this acquisition will seal the fate of Cyrix. National Semiconductor was not interested in making high-performance PC parts, but wanted to use low-power SoCs (systems on a chip).
Sure enough, Cyrix introduced the much-hated 5×86 MediaGX, a chip that integrates features like audio, video, and memory controllers, and features 5×86 cores running at 120 or 133 MHz. It was underwhelming, but it convinced Compaq to use it in its low-end Presario computers. This sparked demand for 6×86 CPUs from other OEMs, including Packard Bell and eMachines.
The shift in focus hasn’t stopped Cyrix from trying to produce more high-performance CPUs, but it has brought hope and other goals. National Semiconductor eventually sold Cyrix to Taiwanese chipset maker VIA Technologies, but by then the key players had left, and the MII CPUs became a worthless piece that didn’t find a buyer.
Cyrix’s last design was the MII-433GP, which ran at 300mhz, and because of this unfortunate naming scheme, it ended up being compared to a processor running at 433 MHz, which performed much better.
AMD and Intel are busy rushing to 1 GHz and beyond, and it will be 20 years before Arm can challenge the two giants in the desktop and server markets — let alone completely dominate the mobile computing market.
VIA replaced the Centaur brand with the Cyrix name, which was its last fang, because its processor actually used the IDT-designed WinChip3 core. National Semiconductor continued to sell MediaGX for several years until it renamed it Geode in 2003 and sold its designs to AMD. Three years later, AMD demonstrated the world’s lowest-powered x86-compatible CPU, requiring just 0.9 watts of power and based on Geode cores, which was considered a testament to the ingenuity of the Cyrix design team.
Why Cyrix’s Legacy Matters
Regardless of whether you own a PC with a Cyrix CPU or not, the company’s history and lessons are worth remembering despite relatively little impact on the PC industry during its decade of existence. Cyrix’s failure proved that improving instructions per clock was a more productive effort for processor manufacturers than increasing raw clock speeds.
To this day, both Intel and AMD have tried to push the nominal clock speed higher with each generation, but after the 3 GHz milestone, most of the real improvements have come from rethinking core parts of their respective microarchitectures. The most famous current example is AMD’s Zen upgrade, which has improved single-threaded performance by 68% in less than four years.
Cyrix was able to survive and overcome a lot of legal (and financially, broadly speaking) pressures from Intel, which sued almost everyone in the CPU space that existed in the 1990s. It has twice shown that lawsuits are not good for a healthy market, and that cross-licensing deals result in a lot of overlap between engineering work at different companies, which can prove beneficial. Unfortunately, AMD and Intel’s cross-licensing agreement on x86 keeps third parties out of this market, and the only company still trying to break this duopoly is China’s Shanghai Zhaoxin.
In Cyrix’s glory days, it was also a fabless company. Today, this is standard practice for most Silicon Valley giants, including AMD, Qualcomm, Broadcom, Nvidia, Apple, Marvell, Unigroup China and HiSilicon, among others, all of which rely on other companies to make chips.
Before the merger with National Semiconductor, the company’s marketing strategy was never very good. In the 2000s, AMD made the same mistake with its Athlon and Sempron processors. These processors are marked as faster than Intel processors, but run at lower clock speeds, but this doesn’t always translate well in benchmarks or real-world performance tests. AMD dropped this option, but arguably, things are still a bit confusing to this day.
You’re unlikely to find Cyrix processors these days other than in the recycling business and enthusiasts’ vintage computer products. There’s evidence online that Cyrix-based desktops haven’t been around until at least 2010, meaning they’ve been around for 10 years after the company has essentially melded into VIA Technology’s offerings. It’s unlikely that VIA’s Zhaoxin arm will still use anything from the original Cyrix design, but only time will tell if they’ve learned their lesson to honor Cyrix’s legacy.