Now, let's start dealing with some of the misinformation above — and there is quite a lot of it to deal with!
Bill is quite right in saying that Cyrix was always a fabless manufacturer, more or less, but both Bill and the usually well-informed iGary get the details pretty compregensively garballed.
There were essentially three stages to the Cyrix fabbing arrangements.
The maths co-pros, the 486SLC and the 486DLC were fabbed by Texas Instruments, and also (I think) by SGS-Thompson. TI were certainly the main source. They were pretty much all sold with a Cyrix badge.
The 486DX/2, 486DX/4, 5x86, and 6x86 Classic were
mostly but not exclusively fabbed by IBM. The IBM-made chips were sold under both brands (IBM and Cyrix) in roughly equal numbers. And yes, Bill, there was a no-money-changes-hands arrangement which meant that for at least some of the several contracts, IBM got paid in kind: Big Blue got to keep half the finished parts. But there were various different contracts with different terms, the details of which I don't know.
By no means all of the Cyrix chips of this era were IBM fabbed, however. Somewhere at home I have a 486DX/2-80 colorfully marked with the Texas Instruments name, and I assume (but don't know for sure) that it was manufactured by TI. More significantly, SGS-Thompson also made Cyrix-designed parts, notably the 6x86 Classic. This may or may not have been sold with a Cyrix badge, but the SGS-Thompson branded and manufactured 6x86-166 sold in reasonable numbers late in the market life of the various 166 classic CPUs - i.e., the 6x86-166, Pentium Classic 166, and K5-166. These were all replaced quite soon afterwards by the Pentium MMX 166, 6x86 Classic 200, and K6-166. The 6x86MX came along a few months later.
The 6x86MX began in the same mould: Cyrix design, IBM manufacture, sold both ways. This applied to the 166, 200, and 233. I don't believe that anyone else except IBM fabbed these parts.
But starting with the 266, things changed. For starters, the original release of the 83MHz bus 6x86MX-266 was, for the first time, not shared equally between the two companies.
LiamC said:
IBM avoided the 83MHz bus chips and opted for another 0.5 multiplier @75MHz to get the same PR rating--> Cyrix would offer 2 x 83 = 166MHz PR200 whilst IBM would offer 2.5 x 75 = 187MHz PR 200.
Wrong. IBM wanted to release the part, but Cyrix didn't believe it was stable enough yet. So IBM released it, Cyrix stayed with 233. Cyrix were right, IBM were wrong. The part was very quick (by the standards of the day) but a very tricky damn thing to get right. (83MHz was an absolute bastard of a bus speed.) The manufacturing, once again, was all IBM. Cyrix did release a 266 (renamed M-II 266) a couple of months later, but sold only a handful of them as by that time the (vastly improved) 300 part was where the action was.
The 6x86/M-II 300 was the last of the great Cyrix chips. For the first time in Cyrix's history, their PR rating was way too optimistic - the 300 was barely PR-266 - but (unlike the 266) it was simplicity itself to work with, as reliable as sunrise, and priced like a PR-200. This last of the traditional knock-down, drag-out Cyrix bargain parts was manufactured by IBM to begin with, but then by National Semiconductor as well. Eventually, all M-II production was on the National Semiconductor fabs. By this time, Cyrix had been under a new management team for quite some time, and was headed up by an ex-Compaq CEO with a strictly sales background. This was a significant change, it saw good sales growth, but staff morale plummeting and the brightest stars on the design team leaving for greener pastures. So the last of the old-school Cyrix products was selling well, but there was nothing workable by way of new parts in the pipeline. They got up to the 333 .... and stayed there .... forever.
After that, there was nothing except promises. National Semiconductor was an absolute disaster for Cyrix. No continuity, no sense of direction, no new product: before too long there was no design team either. NatSemi made a handful of M-II 366 parts, and then degenerated into an organisation barely capable of making a press release.
It was a very sad day for the industry. Cyrix brought PC prices down further and faster than any other company before or since. Cyrix was also the first company in two decades to produce a desktop CPU faster than Intel's best. That alone was a huge achievement. It would not be matched again until the K6-233, which only held the speed crown for 6 weeks and was in any case a tricky damn thing at the best of times, then the K6-3/450, which had a similarly short life as the fastest desktop chip on the planet. The 6x86-200, in contrast, was the fastest thing money could buy for six straight months. It wasn't until the Athlon arrived in 1999 that a longer-lasting challenge to the former monopolist would succeed.