The following four processors are much more similar than one might think:
486 aficionados will recognize the processors made by IBM, ST Microelectronics, and Texas Instruments to be essentially one and the same model—Cyrix 486DX.
These processors are not to be confused with the infamous Cyrix 486DLC, an upgrade processor designed for 386 boards. The Cyrix 486DX was a true 486 with built-in FPU, power management, and 8KB on-chip write-back cache, designed for 486 boards.
Since Cyrix was a fabless chip design company, the Cyrix processors were built by several established microelectronics manufacturers. The processors were sold under both the Cyrix brand as well as the IBM, ST, and TI brands.
It’s worth noting that while AMD produced 486-core processors well into the late 1990s (the 133MHz AMD 5×86-P75, billed as a cheap 75 MHz Pentium equivalent), Cyrix fully focused on systems with the Pentium socket and it doesn’t appear that Cyrix ever made 486s faster than 100MHz (the short-lived Cyrix 5×86 is a different story).
As a side note, the Texas Instruments DX4 is sure to be a top contender for the world’s most colorful x86 processor ever.
AMD also produced 150 MHz 5×86’s. These were badged as P75+ and are rare. I’ve seen much debate over the years about whether 160 MHz 5×86’s were also produced. I’ve never seen a definitive answer.
If I’m not mistaken (always possible!) the last 5×86’s were produced in 1999.
PS: Don’t forget the Cyrix DRX2’s – clock-doubled 486DLC’s ;).
Board manuals often contain settings for AMD X5 at 150 and 160 MHz, but that is known not to be a proof of a processor’s existence 🙂 That said, from what I understand the AMD X5s overclock fairly well, especially the versions rated to operate without a heatsink. So getting an AMD X5 running at 160MHz is not a big deal, whatever the part is rated for.
AMD definitely did produce the X5 for a while, 1999 sounds right. It was very cheap and reasonably fast. There was a similar story with the 40MHz AMD 386DX which was produced until 1995 or so I believe. If anyone knows when AMD stopped volume production of the Am386DX, please speak up.
The Cyrix DRx2 is a different beast as it’s not designed for a 486 socket. That said, I’d love to find one of those!
Michal,
I’ve seen that on a few boards (the X5 markers). I think the HOT-433 had it. Or it may have been one of the SOYO boards. One of the stories I was told is that the 150/160 MHz versions were produced very briefly and later on versions that would have been binned at 150/160 were just sold and branded as 133’s. (I always found it amusing that the early 5×86’s have “80486DX5” stamped on them and the later ones don’t.)
I’ve heard of stable 180 MHz systems (some of the boards using the SiS chipset you recently profiled as well as the UMC chipset had undocumented 60/66 FSB jumper settings). There was a guy in Germany, if memory serves, who got a 5×86 to POST at 200 MHz (66×3) but he had to disable the cache and it wouldn’t get much past the DOS prompt.
Long ago I had one of those DR2x chips. But I had no use for it and had no interest in collecting the stuff at the time and threw it away. (shorter version: I was an idiot)
486DX5 would have been accurate. After all, the X5 is nothing more than a write-back enhanced DX4 with a 4x instead of 3x multiplier. AMD’s own processor recognition guidelines are pretty funny–if you detect a 486 that runs at 133+ MHz, call it a 5×86…
The binning makes sense. I have two X5s right here, AMD-X5-133ADW and the less common 133ADZ. The difference is that the ADZ is rated for a considerably higher case temperature (85 vs 55 degrees Celsius) and does not require a heatsink/fan, unlike the ADW. I wonder if AMD’s reasoning was similar to Intel’s and they didn’t want to compete with their own Pentium-class CPUs. FWIW, the Am5x86 datasheet from March ’96 only mentions the 133 MHz variant, nothing else.
Back in the day I completely skipped the 386DX and 486 systems, I went from a 25 MHz 386SX straight to a 90 MHz Pentium. But I’m now slightly sorry that I sold my Cyrix and IBM 6×86 CPUs, though at least those I can buy cheaper now than what I sold them for back then 🙂
“I wonder if AMD’s reasoning was similar to Intel’s and they didn’t want to compete with their own Pentium-class CPUs”
Thinking about it, the reason they were able to do them was that the P5 bus was still not patented. When Intel patented the P6 bus, AMD ended up having to create “Super Socket 7” as a term for the 100Mhz P5 bus.
For what its worth, I had one of those TI 486s, the lower left-hand one if I remember correctly. I can’t remember any longer if I actually purchased it (unlikely) or got it in trade at a much later date. However I do remember that the colorful Windows logo would slowly turn brown over time as the inks and dyes were affected by the heat of the CPU, age, and nicotine. The last time I looked at it, it was this funky, burnt-looking deal. Kinda sad really.
I had an overclocked X5 to 160 Mhz with a 40 Mhz PCI bus and the difference showed on 3d bench. For that to work I needed some setting on the hard drive controller (I don’t remember what).
Do you know if the registers of those Cyrix DX2/DX4 CPUs are identical for ST Microelectronics, Texas Instruments, IBM, and Cyrix branded chips? Or do some brands have more register options or have them set differently?
The Am5x86 or X5 was produced until at least the 37th week of 2002. At least, that is the datecode one of of my chips. The later chips, from about 2000 onward, have a B1 revision engraved in them. I’m not sure what this stands for. The chips are sometimes rebranded as a DX2-66 with 16 MB of write-back cache, or a DX4-100 with 16 MB of write-back cache. The DX2 chips are 2x/3x capable and the DX4 chips are 3x/4x capable. Both the DX2 and DX4 chips tested seem to run at 133 MHz. The DX2 can be run at 66×2 for 133 MHz. The DX2 also seemed to work in DOS at 2×83 MHz for 166 MHz, though the FSB signal on the motherboard was rather low. The DX4 version seems like an X5 re-labelled and works at 160 MHz.
I’m fairly certain the Cyrix/IBM/ST/TI DX2/DX4 processors can’t be distinguished in software. They may have different physical pinouts but I think the core is identical. Let me get back to you on that though.
I can’t find any documentation for the non-Cyrix-branded Cx486DX models. The documentation exists, just doesn’t appear to be available.
However, sources tell me that the Texas Instruments 486 DX2 and DX4 CPUs can be distinguished through the DIR0 register (contains 80h/81h rather than values in the 1xh range which Cyrix CPUs use). So far I see no hints that IBM- and ST-branded CPUs differ from Cyrix.
Update: I did find some info on IBM DX2/DX4 486s and yes, they do use the same DIR0 values as Cyrix. Still can’t find anything about ST.
The information about DIR0 you received, is it from some sort of documentation, or did you test the CPUs yourself? And is DIR0 simply an identification register? I am a little surprised that no other tweaks were made, especially by TI. They were the ones who too advantage of Cyrix and added 8 KB of cache to the Cyrix DLC, clock-doubled, and went up to 66 MHz. They seemed like quite the rebel, but then to maintain consistency between the Cyrix DX4 and the TI DX4 seemed a little out of character. I do own each of these DX2/4 CPUs (IBM, Cyrix, TI), but will not have the time to investigate this for some years. For ST, I only have their “Cyrix” 5×86.
My main interest for this topic is concerning 386 upgrade paths. A few companies made PGA-132 to PGA-168 upgrade interposer boards. I have these from two such manufacturers, but haven’t had the opportunity to test them. I still have my IBM BL3 on the test bed and don’t even have time to get HIMEM working with that configuration. Anyway, my home was that perhaps the TI DX2/4 chips might have the registers either set or have added features which are more accommodating for use with these PGA-132 to PGA-168 upgrade PCB boards. Do you have any thoughts in this regard?
It ultimately would be interesting to benchmak the IBM BL3 at 100 MHz on the same motherboard which has a PGA-132 to PGA-168 upgrade interposer board containing a TI DX4 100. This way you have system board consistency and can compare the two CPUs directly. I do not think such a comparison has been done, or at least not one that I am aware of. I have completed exhaustive “ultimate” 486 and 686 benchmarks comparisons containing more than 10 thousand data points on identical systems and done by the same tester (me), so it seems a natural path to extend this comparison for 386 systems. I still have a lot of tweaking and configuring to do before I can even begin the benchmarking process on the 386 board.
Is there any way to receive an email notification when certain threads are responded to? I have to keep the tab open and periodically refresh the page to check for responses.
DIR = Device Identification Register. It identifies the processor model, but it’s also used by BIOSes to detect the 2x/3x multiplier of DX2/DX4 chips. The DIR registers for Cyrix CPUs were documented by Cyrix, the docs are available among others on chipdb.org (ping me if you can’t find it). It’s in a Cyrix CPUID document.
I don’t know if TI added further tweaks to the DX2/DX4 chips, docs seem unavailable. If I had to guess I’d say yes, otherwise what’s the point of different identification, but you never know. I doubt they worried about 386 board upgrades because these CPUs were really made too late for that, the DX4 came out in Sep ’95. They were much more interested in having a pinout compatible with Intel for easy 486 board upgrades. As far as I know these CPUs all assumed a chipset with proper 486 support, no hacks required.
I have one or two ST-branded 486 chips so I could check those out.
I think a 386 project would be worthwhile because the upgrades could make a huge difference. And that’s when just considering PGA-132 chips like the Cyrix DRx2 or TI486SLC2. When you take interposers into account it gets even more interesting.
I’m honestly not sure how to get e-mail notifications, but if you want to take this conversation directly to e-mail, I’d be OK with that 🙂
Sidetrack:
notifications about comments in general are available here:
http://www.os2museum.com/wp/comments/feed/?
However there is AFAIK no way to get more than a page, i.e. on this site you have to reload that page atleast every other day or even every day to not miss something.
It would be nice if it were possible to tweek this feed thingie somehow.