Controversies & Anomalies
The official Cyrix 5x86 website boasted about several features of the chip that were disabled by default in the final versions. The most controversial of these features was the "branch-prediction" feature, which was enabled in the benchmarks results on the company website when comparing the chip to Intel's Pentium processor. While it was possible to enable the extra features using a special software utility, it usually resulted in an unstable system, especially on earlier steppings of the chip when running 32-bit code.
There are also many rumours surrounding a 133 MHz, clock-quadrupled version of the Cyrix 5x86. The 133 MHz version is very rare, however, and producers of upgrade kits were given preferential access to it, notably Gainbery. Some of the 100 and 120 MHz parts also contain support for the 4X multiplier setting, and some of these chips may also work at 133 MHz. However, the 5x86 is not known to overclock well, and 120 MHz is generally considered to the pushing the limitations of the process on which it was fabricated. An 80 MHz (2x40MHz) 5x86 also exists, but is unclear as to whether or not it was ever officially released.
IBM's 5x86C was considered to be more conservatively rated than the Cyrix branded parts, and operated at a lower voltage (3.3V). For example, what Cyrix would rate as a 100 MHz part, IBM would mark as 75 MHz. IBM 5x86C was available as 75MHz and 100MHz parts. A few examples of 120MHz parts also exist, but they have early production dates indicating that they may have been produced prior to IBM's decision to scale back clock speeds. 5x86C also had a much longer production run than the Cyrix branded parts. IBM continued to produce 5x86C at least until late 1998 (QFP only), whereas Cyrix's own part was discontinued in 1996. No parts known to exist implement the 4X multiplier or Stepping 1 Rev 3 cores.
Read more about this topic: Cyrix Cx5x86