The Blue Lightning CPU is an interesting beast. There is not a whole lot of information about what the processor really is, but it can be pieced together from various scraps of information. Around 1990, IBM needed low-power 32-bit processors with good performance for its portable systems, but no one offered such CPUs yet. IBM licensed the 386SX core from Intel and turned it into the IBM 386SLC processor (SLC reportedly stood for “Super Little Chip”).
Later on, IBM updated the processor to support 486 instructions. It is worth noting that there were still the SLC variants available—nominally a 486, but with a 16-bit bus.
The licensing conditions reportedly prevented IBM from selling the SLC processors on the free market. They were only available in IBM-built systems and always(?) as QFP soldered on a board.
A very unique processor from the days Intel licensed others to make x86 chips, even allowing them to improve upon them. Those days are long gone, with only AMD and VIA remaining as companies with an x86 license.
Zhaoxin a few years back took over the licence from VIA (joint venture between VIA and Shanghai). In practical terms, only AMD and Intel are competitive.
I was under the impression that VIA’s license lapsed circa 2018 and that it was specifically non-transferrable, though. (Basically, they won it in 2003 for 10 years, and then a 2010 FTC settlement as a result of anticompetitive pricing practices gave a 5 year extension.)
Of course, Centaur (the VIA division that was doing the CPU designs, and AFAIK the most recent Zhaoxin design was really a Centaur design that VIA never shipped themselves, with the AI accelerator IP deleted, as much as Zhaoxin wants to claim that it’s a Chinese design (it’s not)) has no need for an x86 license any more – VIA sold them to Intel.
And, really, as Zhaoxin’s products aren’t meant for markets outside of China, they don’t really need a license from a US company.
It’s also worth pointing out that all 32bit x86 patents have expired.
The initial x86-64 patents should also be expired or nearly expired. So no license needed globally for that.
I think we also need to take into account that even though patents may expire, there is tons of associated copyright and IP that has a significantly longer “staying power.”
Copyright is irrelevant… and IP itself has no protections other than secrecy without patents and copyright both of which no longer protect any AMD or Intel instruction set before 2003.
Intel licensed the x86 to many second sources during the 70s and 80s. But it very rarely allowed them to “improve” upon them. Just look how they sued NEC over the V20.
Incidentally, the second sourcing of the 1st 3 generations of x86 is what made it the “open” architecture of that period. And it built a tremendous ecosystem around it. Intel started to put that cat back in the bag with the 386, which they did not intend to license to 2nd sources.
SiS/DMP has a valid i386 license from when they bought RiSE Technologies with their MP5. The latest Vortex86DX2 CPU is a 800mhz 386 cpu and would be by far the fastest 386 ever made. The Vortex86DX3 is a dual core i686 compatible chip with dual core 1000mhz cpu.
They are quite snappy for small embedded systems and supports much faster ram than any other 386cpu could even dream of back in the day.
AFAIK, Rise didn’t actually have an x86 license, they just hoped to get enough patents of their own to countersue Intel. (Which was also Centaur/VIA’s strategy, and it worked for them – I think they actually sued Intel first, and got an x86 license out of it.) Nobody sued them because the Rise mP6 wasn’t successful, AFAIK (but SiS did buy the IP for set top boxes). I think DMP’s strategy is simply to use technology older than 21 years old, which neatly avoids any patent protection.
A few other x86 vendors (most notably Cyrix and NexGen) piggybacked on their fab partner’s license (although I don’t actually know where TI or SGS-Thomson got theirs, but IBM definitely had one). (Cyrix never got sued, sold their older IP to NatSemi (who then sold it to AMD), got bought by VIA, and then essentially written off because the actual Cyrix III design was worse than the Centaur C5, which picked up the Cyrix III name, and then was shortened to C3. NexGen was bought by AMD, the Nx686 was adapted to the Pentium bus and became the AMD K6, and then the NexGen team (together with the former DEC Alpha team, that AMD somehow ended up with) went on to make the AMD K7.)
And then, Transmeta used both the VIA and Cyrix/NexGen strategies – defensive patents used offensively, and piggybacking on IBM’s x86 license. It didn’t work as well for Transmeta, the settlement included discontinuing all x86 CPUs (but it was after Transmeta’s designs were no longer competitive anyway), although Nvidia ended up with the IP (and then Nvidia didn’t actually make x86 CPUs with it even though that’s why they bought it, instead using an ARM license to make ARM CPUs).
From the time I was @ NVIDIA, I remember vaguely that most of the IP of interest from transmeta was regarding the power/frequency scaling and dynamic instruction translation. Some of the ARM cores that NVIDIA has made used them extensively.
32bit CPU with a 16bit bus? Sounds like the SX series.
The 386SX had a 16 bit bus ( referenced in the article ) but the 486SX was fully 32 bit. The 486SX had no FPU which is what made it the budget version of the 486DX.
The 486SX was a 486DX with a fused FPU. You could actually buy a FPU for the 486SX as an aftermarket plugin. It was a 486DX with the CPU fused. I assume This was just binning bad CPUs and FPUs.
You are correct. Quite a nice Product Management solution to manufacturing waste really. They managed to create 3 products from one design while reducing the defective rate across all of them. Nice work.
I’d have to look again, but I believe the 386 and 486 upgrade boards I have in my two IBM PS/2s are both ‘SLC’ chips. Also of course would be running on a 16bit bus.
It’s worth noting that Intel didn’t do this out of the goodness of their hearts, they did it because IBM demanded a “second source” for every part in the original IBM PC.
Other companies such as Cyrix/Via managed to acquire an x86 license via other means, see the comments above mine for details.
But the thing is, those early x86 patents have most likely expired by now. In fact, every instruction up to the Pentium III has probably expired by now, and the x86-64 patents are probably expired or expiring very soon.
At this point, it’s more a case of operating systems such as Windows 11 preventing someone from making a minimum viable x86 CPU than anything else (by forcing the implementation of new patented instructions and keeping the Intel-AMD duopoly alive). And this is Microsoft’s big strategic mistake, because they are essentially preventing a company like Qualcomm from competing in x86 and bringing prices down. And no, Windows users won’t take a transition to another ISA lightly. Windows NT is essentially the backwards compatibility OS and is a big part of its value as an OS, so Microsoft is really shooting themselves in the foot by not letting some baseline x86 instruction set become a royalty-free standard.
x86-64 is AMD64 which was invented by AMD, not Intel. Intel was still pushing Itanium when AMD64 came out. Intel 64 is what Intel called their copy of AMD64 and we generally use x86-64 to refer to them both.
amd=amd64, intel=emt64, itanium=ia64.
Intel calls it now “intel 64”
They suck at naming schemes…
amd64=x86-64 from AMD ( they invented it )
emt64=Intel 64=X86-64 from Intel ( they copied it and rebranded it multiple times )
ia64=Itanium ( another, incompatible, 64 bit arch from Intel – the market rejected it )
Probably not so much IBM itself as it’s government contracts required it.
Intel was second sourcing x86 before the IBM contract. That was one of the reasons why they got the contract for the PC over the 68K, which apparently Motorola wasn’t willing to second source.
2nd sourcing was very common in the 70s and 80s. The microprocessor was used initially mostly in military applications, with the US govt requiring multiple soures of fabrication for he same part. Also, during the 70s, the processes for the dynamic CMOS logic had much worse yield/capacity than static/RAM. Which is why vendors like IBM required the multiple sourcing, as they expected Intel not to be able to meet the volume they needed.
Prices for x86 are already low enough, so I don’t see microsoft stressing much about further competition between Intel/AMD. Specially since Windows on ARM devices may be putting the fire under the belly of those two.
The 8086 and 8088 where NMOS based and not yet CMOS.
You raise an interesting question. How viable would it be to produce an “off patent” x86-64 compatible CPU these days? AMD64 itself is more than 20 years old as are the SSE, SSE2, and SSE3 instructions. It is hard to imagine selling an x86 CPU without virtualization support these days. Intel first released that in 2005 and so still patented but not for long.
I guess the question is what market this new x86-64 entrant would be targeting though. It is hard to imagine entering the desktop or laptop market to compete with Intel or AMD. Even companies like Qualcomm would struggle to do that I think. In other markets, ARM or even RISC-V seem to have enough advantages that x86-64 compatibility may not be enough of a draw.
The SBC market could be an option. That market would certainly benefit from the “standardization” and especially the mainline Linux compatibility that UEFI and x86-64 would bring. The market was crying out for Raspberry Pi alternatives and many emerged but they mostly suck due to a lack of ongoing software support.
It is perhaps a surprise at least that there is not a well-known Chinese knock-off of X86-64 for domestic Chinese and low-end motherboards and laptops.
although parts of the ISA may be now out of patent timelines. Stuff like the microcode etc is not.
Plus ARM and RISC-V are just far more attractive from a licensing stand point.
But in the end it all boils down to the fact that both Intel and AMD have managed to scale x86 across a wide enough spectrum of price points, that it is just not really that attractive to compete against them given the investment it would require. So there is little business incentive nowadays. Back in the 80s and 90s there were clear gaps that 3rd parties could fill with their own alternative x86 designs.
Man, everything computing was just far more exciting in the 80/90s. I remember when Cyrix and AMD busted on the scene and shook up the processing space. BeOS, Linux, alternative OS’s just coming out of nowhere. I had a crappy 386SX off brand computer and just dreamed of upgrading to the latest Cyrix 686.
There is plenty of excitement right now, e.g. there are few times in history when we have had access to so many uArchitectures in the consumer space, at ridiculous price points. Never mind the amount of compute density we can get out of a relatively cheap GPU, you can have TFlops of compute in your PC now.
I think we tend to associate the time we were more engaged with a field, as it being the most exciting of times. For the younger generations, they experienced a massive wave of development in the mobile space in a bit over a decade, for example. Which most old foggies have completely missed, because for them all cell phones are the same. Just as all them PCs were the same for the previous generation forged in the most exciting of times; the minicomputers!