Orphan CPU chips

[vwestlife]

IBM Blue Lightning (2nd version) DX2... Cyrix's 486DX2, manufactured by IBM.

The first version of the Blue Lightning was based on the Cyrix 486DLC (really just a jazzed-up 386DX) and was the first clock-tripled x86 chip and the first x86 chip to reach 100 MHz (it beat Intel's 486DX4-100 to the market).

AMD 5k86... AMD's first, unsuccessful attempt at a Pentium competitor. Later renamed the K5 but still without any success. Meanwhile, AMD's 486DX4-based 5x86 was selling like hotcakes, and they quietly scrapped both the 5k86/K5 and their own in-house plans for a K5 successor (see below).

NexGen Nx586... RISC-based competitor to the Pentium. Most versions did not have a built-in math co-processor, instead using the outboard Nx587 chip. Last version ("PF" code) did have an onboard coprocessor. Quickly bought out by AMD and used as the basis for their popular and successful K6.

Cyrix MediaGX... Cyrix's attempt at combining sound and video into the CPU. Used in some low-end laptops but otherwise a flop.

Others:
Intel 80186/80188
NEC V40 & higher
Intel RapidCAD
CDT WinChip

 

[Anonymous Coward]

A couple of comments:

-I think it's not quite right to call the IBM (Cyrix based) Blue Lightning a "version" of IBM's Intel based SLC/DLC family. They are not related in any way except for the name which IBM kept for marketing reasons. IBM actually spent some serious time on their DLC/SLC family. The 486DX/2 Lightning is just a Cyrix chip with a new label.

-The am5x86 did not kill the K5. The fact is that the K5 was about 2 years late to market, and when it finally arrived it underperformed Pentium chips by a wide margin. I believe the 2nd version of the K5 was much better, but by this time the K6 and Pentium II were already out. K5 only on the market for just over a year and was done in by the K6, not the 5x86.

-It's pretty rare to find a nexgen motherboard with an FPU socket. The only ones I am aware of are pre-production versions. Infact, I don't believe the nx587 was ever available to the general public.

-I don't know if I'd call the MediaGX a "flop". It certainly didn't compete with the mainstream chips, but then again it was never designed to. I believe the MediaGX was pretty damn popular in 3rd world countries like China. Infact, the MediaGX is still produced to this very day. I believe National Semiconductor bought it, changed the name to Geode and may have possibly sold it to VIA. Either way, I'm pretty sure it's still around.

 

[Jorg]

I'd have to see what version of the blue lightning I have, but my favorite socket 3 cpu is this one:

Unfortunately I've got only one motherboard that fully supports it, and that seems have gone flakey.

 

[shawn510]

I wonder, are the Via C3 processors in any way related to the Cyrix chips (MediaGX, 6x86, M3)?
I had a 6x86 and later a mII (overclocked by setting the bus to 75 like the old 6x86). They were cheap and did okay until you wanted to play the fancy new games that used the FPU heavly.

I remember some sort of webcomic back in those days had one where a group of characters out af a game were gathered together, except one was very pixalated and not too popular. The caption was "It's hard being the guy with the Cyrix".

Yes, it was.

 

[Anonymous Coward]

I'm pretty sure all of the VIA C3 is based on the IDT Centaur. I think they pretty much gave up on the MII/MIII core.

One thing i really don't understand about Cyrix is why since 5x86 they had such a lousy reputation for FPU performance. FPUs is of course what put them on the map in the first place. From what I understand after developing the 83d87 core they just never bothered to update it. I've been told that their 486DX, 5x86 and 6x86 all used the original 83d87 design.

@Jorg

If you have a Blue Lightning CPU and it's not soldered onto a pcb, then it's going to be Cyrix based. IBM wasn't allowed to sell the Blue Lightning separately because of an agreement with Intel...I think that's their main reason for switching to Cyrix.

 

[shawn510]

I think maybe that the sudden demand for FPU performance by average users caught them by surprise. Until quake and it's progeny came along, a lot of home users didn't know/care what a FPU was. Integer was where it was at.

Cyrix was the king of coprocessors, though. I had a am386sx-40 with the cyrix coprocessor handed down by my brother. It was sweet. It came with a fractal demo program that required the FPU.

 

[vwestlife]

-I think it's not quite right to call the IBM (Cyrix based) Blue Lightning a "version" of IBM's Intel based SLC/DLC family. They are not related in any way except for the name which IBM kept for marketing reasons. IBM actually spent some serious time on their DLC/SLC family. The 486DX/2 Lightning is just a Cyrix chip with a new label.

There were two entirely different series of Blue Lightning processors. The first used the 386DX pinout and was a clock-tripled 486DLC with a large L1 cache (16 KB, versus Cyrix's 1 KB), unofficially called the 486BL3, in 75 and 100 MHz versions. I believe IBM may have also used the Blue Lightning name on their clock-tripled 486SLC3 (also with 16 KB L1 cache).

Later, IBM took Cyrix's 486DX2 and put their Blue Lightning name on it. Unlike the original BL series, there was no engineering done by IBM; it was just renamed. IBM also manufactured and sold the Cyrix 5x86 under their own name.

-The am5x86 did not kill the K5. The fact is that the K5 was about 2 years late to market, and when it finally arrived it underperformed Pentium chips by a wide margin. I believe the 2nd version of the K5 was much better, but by this time the K6 and Pentium II were already out. K5 only on the market for just over a year and was done in by the K6, not the 5x86.

Cyrix could've been in the same boat with their M1. Just like the K5, it was their attempt at cloning the original Pentium. But when the full-blown M1 fell behind schedule, Cyrix scaled it down to work with a 486 pinout and motherboard and released it as the Cyrix 5x86 (code name M1sc).

The Cyrix 5x86 and AMD 5x86 were entirely different chips. The Cyrix 5x86 was a Pentium-class CPU scaled down to work with a 486 motherboard. The AMD 5x86 was simply a clock-quadrupled 486DX souped up with a large L1 cache. As a result, with correct BIOS support for both chips, a stock Cyrix 5x86-120 could easily beat even an overclocked AMD 5x86-160 in benchmarks and real-world performance, especially when you use a utility program to unlock the special M1-class features that Cyrix purposely disabled by default in order to improve stability. (The full-scale M1 itself was later released as the 6x86.) Unfortunately, the Cyrix 5x86 was more expensive, was more picky about which motherboards and BIOSes it would work with, and suffered from production shortages, especially of the 120/133 MHz version.

 

[vwestlife]

I wonder, are the Via C3 processors in any way related to the Cyrix chips (MediaGX, 6x86, M3)?

The VIA C3 is based on the IDT WinChip series. Initially it was going to be yet another iteration of the Cyrix 6x86 series, but the results were unimpressive and VIA (who had bought out Cyrix) instead bought out and used IDT's plans for a new version of the WinChip. Confusingly, VIA first released it as the "Cyrix III", even after the attempt to base it on a true Cyrix design had already been abandoned.

Now how about Intel's biggest goof-up before the Pentium math error? The first 80386-16 (released before the 386SX, so there was no "DX" suffix) could have a fatal bug when running 32-bit software, so those chips which exhibited the bug were marked "16-bit software only", while those which didn't have the bug and were OK to use with 32-bit software were marked with a double-sigma logo.

How about a 386-12?

386EX... I'm not sure how it compares to the DX and SX.

80C188-20...

386SX-40 with Windows logo...

AMD 486SX2-66...

AMD 486DX4-90...!?

Early AMD 5x86 also marked as the 486DX5 and with the Windows 3.1 logo.

Rare but true 150 MHz and 160 MHz versions of the AMD 5x86. For marketing reasons, most of these 150 and 166 chips were only marked as 133 MHz. Users quickly found the extra speed available with perfect stability, because it wasn't really "overclocking"; it was just using the chip at its originally designed speed. The temperature code indicated the chip's true design speed: ADW = 133, ADY = 150, ADZ = 166 (but apparently some ADWs were also rated for 150 MHz).

Cyrix 5x86-133...

IBM Blue Lightning DX2. Rebadged Cyrix.

 

[carlsson]

I also got an AMD 486SX/2-66. I had no idea a such beast existed until I found it in my dad's Compaq Prolinea, originally a 486SX-25 once upgraded. I installed an Intel 486DX/2-66 instead but I don't know if it made any speed improvement.

 

[vwestlife]

I also got an AMD 486SX/2-66. I had no idea a such beast existed until I found it in my dad's Compaq Prolinea, originally a 486SX-25 once upgraded. I installed an Intel 486DX/2-66 instead but I don't know if it made any speed improvement.

It won't unless you're using the math co-processor (FPU). That's the only difference between the two. In fact, many Intel 486SX chips were actually just a 486DX with the built-in FPU disabled. It would cost Intel more to leave out the FPU than it did to just clip the one pin that powered it.

The biggest joke was the 486SX-16. It was more expensive than an AMD 386SX-40 and ran more slowly. You basically just paid extra to get the 486 name.

Here's another oddity: the 486DX-50 (not DX2). It was blazingly fast at the time, but the 50 MHz bus speed was simply too high for most VESA Local Bus and PCI cards to handly properly -- just like the old 286 and Turbo XT motherboards that tried to run the ISA bus at 12 MHz. (Eventually during the Pentium era, the bus speeds were standardized at 33 MHz for PCI and 8 MHz for ISA.)

 

[Druid6900]

Yeah, I had one of the DX-50 chips when I upgraded from my original 386-33 and I made my supplier I bought the VIP system from unpack cards and test them at 50MHz until he found the cards that I wanted and they worked right at top speed.

Then that became the standard system for the MoBo/Chip combination and we sold a LOT of them.

 

[vwestlife]

I almost forgot about this one... the Weitek 3167 math co-processor -- a faster competitor to Intel's 80387. It was a famous name, but very few people actually had one, at least outside of applications like CAD.

Also, a "4C87DX"-40... but in the package of a 387SX. Perhaps designed for 486SLC-type systems, but the designation is just odd.

And a Harris 80C286-25 -- the fastest 286 ever...

 

[Anonymous Coward]

Here's a couple more things I know about some of these chips:

-486SX2s came out pretty late. I think 1995ish. Both Intel and AMD made them. Uncommon, but not rare.

-I don't know the exact reason for the IIT 387DX FPU in PLCC package, but it was probably a cost saving procedure. I have seen several 386DX motherboards with the PLCC FPU socket. I think I've also seen a few 387SX chips available in PGA.

-Harris planned a 33MHz version of the 80c286 but canned it at the last minute

-The 12MHz 386 was released because intel had production problems and couldn't meet the release date for the 16MHz part.

-There are two versions of the intel 50MHz 486DX. The early version had overheating problems and was recalled. The problem was fixed by reducing the die size. Also, 50MHz parts aren't that rare. People with EISA systems liked them because EISA wasn't a local bus. I think these were more popular with academic institutions. Personally I don't like using the 50MHz parts because of the need for extra memory wait states.

-The 386EX has 26-bit memory addressing so that it can use up to 64mb of DRAM. This one was meant for embedded systems and wasn't for end users.

-I believe the 486SX-16s were OEM only parts...I think specifically for DELL. I believe the same was true for the am486DX/4s, but instead for Compaq.

-Blue Lightning was the ibm 486DLC2/3. The SLCs were never called Blue Lightning.

 

[carlsson]

Frankly, I'm clueless to which if any applications in a standard Windows 95 installation has use of the FPU. Excel maybe but I have no idea otherwise. I'd gladly sell or even give the CPU away to anyone who collects this stuff - I don't.

 

[vwestlife]

-I don't know the exact reason for the IIT 387DX FPU in PLCC package, but it was probably a cost saving procedure. I have seen several 386DX motherboards with the PLCC FPU socket. I think I've also seen a few 387SX chips available in PGA.

My last AMD 386DX-40 motherboard had a soldered-on CPU (like a 386SX), therefore making it impossible to upgrade to a 486DLC, but it at least had a regular 387DX socket. The whole motherboard was only as long as the 16-bit ISA slots!

-Harris planned a 33MHz version of the 80c286 but canned it at the last minute

That would've smoked quite a few low-end 386s in the days before true 386 power was needed for Windows.

-The 12MHz 386 was released because intel had production problems and couldn't meet the release date for the 16MHz part.

Same thing with the 60 MHz Pentium, versus 66 MHz... and they still screwed up the FPU! The 60/66 MHz Pentiums are true orphan CPUs -- basically beta testing done by consumers, to get the bugs out before the "real" 75 MHz and up Pentium was released. Speaking of which, most 75 MHz Pentiums were actually relabeled P90 chips, done so cheap brands like Packard Bell could sell Pentium machines at a lower price point! Eventually the same thing was true of all the "one step below" Pentiums: 120 (vs. 133), 150 (vs. 166), etc....

 

[Jorg]

@Jorg

If you have a Blue Lightning CPU and it's not soldered onto a pcb, then it's going to be Cyrix based. IBM wasn't allowed to sell the Blue Lightning separately because of an agreement with Intel...I think that's their main reason for switching to Cyrix.

Nope, its not soldered, and got a big IBM logo on it. Its 66 mhz but I would have to look for the exact model.
I remember it had a non-standard voltage or something.

 

[IBMMuseum]

My last AMD 386DX-40 motherboard had a soldered-on CPU (like a 386SX), therefore making it impossible to upgrade to a 486DLC, but it at least had a regular 387DX socket...

This would be a "PQFP" (Plastic Quad Flat Package). Usually even those motherboards did still have a PGA socket for a CPU upgrade. There were also Cyrix PQFP "FasMath" FPUs for the 386DX motherboards.

...Same thing with the 60 MHz Pentium, versus 66 MHz... and they still screwed up the FPU! The 60/66 MHz Pentiums are true orphan CPUs -- basically beta testing done by consumers, to get the bugs out before the "real" 75 MHz and up Pentium was released...

But there was a few early "Socket 5" Pentiums (75MHz+) that also had the FDIV bug. Yes, the "Socket 4" Pentiums (60 and 66MHz) did have some other small bugs, but the Pentium wasn't a first occurance for those sorts of things. In fact the 386DX CPUs had several bugs, some not even purged until the last core version, after they were being heavily supplanted by the 486 CPUs.

 

[vwestlife]

-Harris planned a 33MHz version of the 80c286 but canned it at the last minute

A bit more background info: The 386 was the first CPU Intel refused to let other companies manufacture under license. So when the 386 was released as an Intel exclusive, companies like AMD and Harris responded by offering cheaper, faster 286s. AMD fought Intel in court for the right to design and manufacture their own 386 chips (which they finally won in 1991), while Harris simply elected to stop making x86 CPUs after the 286.

 

[animekenji]

Actually, the nx586 was not the basis of AMD's K6. It was Nexgen's planned successor to the nx586, tentatively titled nx686, that was to be the basis of the K6. There were also no nx587 chips or motherboards with nx587 sockets released to the general public to my knowledge. I've been searching for both for years. I've only just managed to source a Nexgen PCI motherboard recently and only in the last year or so that I was able to locate my 2 VL Bus motherboards. The only nx586 with a built in FPU that I have EVER seen for sale is the one on ebay right now for $850 and barring a substantial windfall coming my way, I won't be buying that one. I saw one nx586 120mhz for sale, but the pins were smashed on it so I wasn't going to risk buying it, going blind trying to straighten all the pins without breaking them, and then finding out that it doesn't work anyway. I left that BS for someone else to enjoy.

 

[Stone]

Here's another oddity: the 486DX-50 (not DX2). It was blazingly fast at the time, but the 50 MHz bus speed was simply too high for most VESA Local Bus and PCI cards to handly properly -- just like the old 286 and Turbo XT motherboards that tried to run the ISA bus at 12 MHz. (Eventually during the Pentium era, the bus speeds were standardized at 33 MHz for PCI and 8 MHz for ISA.)

I still have two of these chips if anybody's interested. They're on Dell processor boards and have heatsinks on them.