5 ms·
IBM Blue Lightning: World’s Fastest 386? (2014)
- johnklos 3y agoThe x86 world certainly had a lot of interesting outliers that other architectures, like m68k, never did. Sometimes I wonder how a 100 MHz m68030 with tons of cache would compare with an '040 or '060... The Blue Lightning supports 80486 instructions, so it can run modern NetBSD. That'd make for some interesting observations.
- sitzkrieg 3y agovery true, i also wonder if in some weird twist of fate some outlier like cyrix became the dominant x86 manufacturer and designers or something wack like that
- userbinator 3y agoIBM choosing x86 for the PC meant that development focused there, and back then it seemed IP/patents was less of an issue than it is today, so there was a lot more competition. There was at one time over a dozen independently designed x86 cores available as purchasable products: https://en.wikipedia.org/wiki/List_of_x86_manufacturers https://en.wikipedia.org/wiki/List_of_x86_manufacturers
- WarOnPrivacy 3y ago> IP/patents was less of an issue than it is today, so there was a lot more competition. There's a powerful lesson in here somewhere.
- scns 3y agoRemy Martin are, per decree from 1936, the only distillery making Cognac in the Champagne. Good aroma but way to sharp for 40% alcohol for my taste. Zero Competition.
- numpad0 3y agoYes. My interpretation is there had been a problem of population and economic growth on Earth: The world population had doubled by 2020 from 1980, while US population grew only 1.5x. US GDP had grown 10x, world 3x. So the world had shrank at least 1/2, and the US both had to earn more influence and had gained more leverages during these years. It would not be surprising if seemingly overly litigious and possessive culture both in the US and world at large were mere symptoms, than cause. It's time to, either have a global thermonuclear war for a proud rewind(please no, I could be sitting far enough to die liquefying over weeks when suns appear over), or to boldly go where no one has gone before.
- scarface_74 3y agoNot really. Intel licensed the x86 architecture
- WarOnPrivacy 3y agoIP in more hands and not locked away is the thing that led to more competition. In regards to advancing competition, licensing's role was to be an obstacle that needed overcoming.
- jpollock 3y agoThey weren't any less litigious, Intel sold licenses to the x86 architecture. There were several fabs that had licenses, and there were lawsuits around whether or not they were transferable. https://jolt.law.harvard.edu/digest/intel-and-the-x86-architecture-a-legal-perspective https://jolt.law.harvard.edu/digest/intel-and-the-x86-archit...
- formerly_proven 3y ago> back then it seemed IP/patents was less of an issue than it is today, so there was a lot more competition Maybe it was rather because IBM required second sources, so Intel wouldn't have gotten the x86 win without licensing it to other manufacturers.
- bombcar 3y agoIt was also surprisingly easy [1] to switch entire CPU architectures back in the "olden days" and it wasn't a given that x86 would even be a long-term architecture the way it ended up. [1] for a value of easy that was something between "we made all the software and so recompiling on a new processor isn't customer painful" to "the thing is programmed entirely manually anyway so who cares".
- flenserboy 3y agoI'd be curious to see the timeline where IBM went with the 68000 instead, with a similar resulting proliferation of chipmakers and chips. Making Motorola compete, as well as having to push the envelope, could have led to some special designs.
- anthk 3y agoWell, a 486 with an FPU @70mhz would stomp it.
- shrubble 3y agoThe 40MHz 68040 would be a good one to use for comparison. This chip had block move instructions as did the 486 which greatly increases performance. Supposedly Motorola felt uncomfortable putting a heat sink on the CPU and thus they didn't increase the clock speed further...
- mysterydip 3y agoI know it wouldn't be a 1:1 comparison with an actul chip, but could this be done on an FPGA?
- LargoLasskhyfv 3y agohttp://apollo-core.com http://apollo-core.com running on not so fast and large FPGAs should give an impression. Now imagine that as ASIC on modern process nodes, coupled with HBM.
- bigmattystyles 3y agoI was a kid back then but had some close family members who worked with these and I just had a flashback and remembered the joke was that they could spark, living up to the blue lightning name. Is that apocryphal? I have no way of checking the validity of that joke.
- throwanem 3y agoI'm pretty sure they were pulling your leg.
- bigmattystyles 3y agoJust got one of them on the phone and they had no idea what I was talking about - so most likely a throwaway joke I believed and committed to memory all those years ago. :-)
- cbsmith 3y agoI remember these. I used to sell computers back in the day, and the DLC3's were notoriously "different" and therefore "buggy" with certain software. Not as bad as the later Blue Lightning chips that were based on Cyrix designs, but still...
- manual89 3y agoBack in the day I came across a PC port of Mortal Kombat II. The box had a label specifically stating that it would not run on the AMD CPUs of its day...
- muxator 3y agoWhat was the power consumption of these chips? More interesting yet it would be to know the Power/speed ratio compared to modern chips.
- jl6 3y agoInteresting question - it’s not trivial to find a metric that’s directly comparable between a 486-era chip and a modern x86 chip. MHz isn’t the whole story, as modern chips have multiple cores and greatly improved instructions-per-cycle. Synthetic benchmarks like Whetstone might provide a point of comparison, but it’s hard to be sure the different benchmarkers were running the same version. Wikipedia says the 386SLC from the article drew 2.5W at 25Mhz. A Ryzen mobile CPU draws something like 15W. My gut says the Ryzen is a lot more than 6x faster.
- qubex 3y ago> MHz isn’t the whole story MHz has never been any part of the story.
- pavlov 3y agoThe 386SLC had 815,000 transistors manufactured on a 1.2-micron process (1,200 nm). The Ryzen 7040 has 25 billion transistors manufactured on a 4nm process. That’s 30,000x more transistors on a much more power efficient process. There’s no real-world task where the Ryzen won’t perform many orders of magnitude better. Anything involving floating point must be millions of times faster because the 386SLC didn’t have an FPU at all whereas the Ryzen has wide SIMD and an integrated GPU.
- actionfromafar 3y agoWouldn’t surprise me if the 386 could respond to interrupts faster.
- jacquesm 3y agoI cycles, yes, in absolute time, no.
- api 3y agoPerspective: the 95MB/sec memory speed quoted here is roughly the same as gigabit Ethernet between stuff in my house. This speed can also be achieved by 802.11ax if you are reasonably close to the access point.
- cmicali 3y agoI bought a machine with a 75Mhz blue lightning "486" as an upgrade to a 386SX/20. It was sold as the budget 486 at the time, and I learned the lesson the hard way that sometimes budget really does mean a lot worse! Performance wasn't great and compatibility was an issue more than you'd expect.
- RecycledEle 3y agoI got an IBM ThinkPad 500 sub-notebook as a high school graduation present in 1994. There is no documentation of it today. A different laptop is recorded as the ThinkPad 500. Mine had a 486slc-2 50 processor and 4 MB RAM that I later upgraded to 12 MB. I think they were closeouts IBM decided not to mass produce.
- sacnoradhq 3y agoThe fastest 386 was the 486SX. ;) Interestingly, the 487SX was a complete 486DX that disabled the original 486SX. https://micro.magnet.fsu.edu/optics/olympusmicd/galleries/chips/intel487sxa.html https://micro.magnet.fsu.edu/optics/olympusmicd/galleries/ch... EDIT: IBM's LC chips were targeted at laptops initially but then found their way into embedded systems. Cyrex (used in the BL line) was often slightly ahead in performance and performance/price, but compatibility wasn't always there.
- smileykenziekv 3y agoThis is mind-blowing, I can't help but feel a sense of nostalgia for this era of computing. The technical insights into its design and how it pushed the boundaries of performance for its time leaves me in awe of the engineering prowess at play.