The Soviet 8080 Processor – The 580 | Hacker News Reader

The Soviet 8080 Processor – The 580

cpushack.com

72 pointsprotomyth26 comments

There's a strange romanticization of the BESM-6 in the West. I cringe when I read the "it took us 30 mins and them 1 min" anecdote - given that there are no comparables here beyond the vague result.* There's another similar one comparing BESM-6 to CDC6600 with the claim that if the former ran at the same clock speed as the latter, it'd be comparable. If a horse ran at Mach 2, it'd be comparable with an F-15.

*Who knows what crud code we were running on some IBM thing in what language, using what algorithm, to what level of precision? Or what the BESM-6 was running?

It's not necessarily the West romanticizing it in this case:

> This article is largely from guest author Vladimir Yakovlev, translated from Russian, and edited/expanded by me.

I was thinking of the Science Museum (in London) exhibit along with quotes from the likes of Doron Swade. The article is just fairly quoting this stuff.

Ironically, decades later this clone of the 8080 was itself reverse-engineered by someone in Ukraine: https://zeptobars.com/en/read/KR580VM80A-intel-i8080-verilog...

This is a great cautionary tale about why not to copy. Imagine if they'd taken their approaches from SECM and put them on an IC instead of copying the 8080... they probably would have had a superscalar chip when the West was just shipping 8-bit in-order cores.

Yeah, I also read that the order to copy CPUs developed by the West has greatly demoralized Soviet engineers working in this field. You should learn from your competitors and build on top of that.

Dijkstra wrote somewhere that the Cold War was effectively over when the Soviet Union decided to copy the IBM 360.

don't know, developing the s360 line cost ibm 5 billion $ [1] (that was quite a bit of a moon shot, given that the moon program has cost 25 billion $ [2]). Maybe it would have cost them less if the s360 instruction set had a stack, but who knows. In any event this price tag pursuaded them to clone the whole thing.

[1] https://spectrum.ieee.org/tech-history/silicon-revolution/bu... [2] https://money.com/apollo-11-moon-mission-cost/

US companies also copied the IBM 360. RCA did, perhaps others.

US companies also copied the IBM 370. I used a far superior clone computer called the Itel AS/5. Much of the hard work was actually trivial because:

ALD's were Automated Logic Diagrams. IBM sold these. They were complete schematics for the S-370. EXSYSCO basically duplicated the logic of the S-370. http://www.silogic.com/NAS/NAS.html

These schematics were freely available in the machine room. I remember perusing them at night when things were slow. Microcode listings were also freely available. (I believe a large part of IBM's openness at the time was because of antitrust litigation).

I also remember seeing the Itel engineers using the IBM schematics when debugging problems.

Why? The Cold War more or less started with soviet spies stealing the western allies nuclear technology. It’s not like the USSR was standing on their own 2 feet entirely independent of western technology at any point during its existence.

Because then your engineers won't have learned the lessons from building their own machine. They'll be even further behind when the next generation comes out. At that point the only option is to keep stealing the designs which means you'll always be behind. It's like cheating on your homework, it works at first but then you fail the final.

I don’t disagree with you, but my point is that was more or less the status quo in the USSR.

I don't think that's really true because 1) currently developed states like the US got that way in large part by IP theft 2) if this were really true you'd expect it to be a detriment to, e.g., learn well-known algorithms in school instead of deriving them from first principles.

Doesn't seem right. The US was hardly doomed by the great lengths it went to to copy British textile mills!

On the contrary, this is a great tale of why reverse-engineering standard technology makes sense (if you can't buy it at least, which the East couldn't because of the COCOM embargo).

If the Eastern hemisphere had developed their own computer technology, everything I had learned as kid would've been worthless after the Wall fell. But since East Germany was Z80 (and towards the end, 8086) land, I could more or less seamlessly continue with my self-education ;)

I've heard that East German engineers liked the Zilog Z8000 a lot more than the Intel x86 architecture, but pragmatically it made a lot of sense to adopt the same standards that had won in the West already.

Meh. I grew up with x86 and it was trivial for me to get into ARM and even using ARM NEON, etc. Computers are computers and understanding of them is largely transferrable.

Well yeah, but I still had a head start because I already knew that PCs were a crap architecture and I could focus on the Amiga instead as soon as I could get my hands on one ;)

It goes much beyond the CPU architecture though, more important (and expensive) is the entire software ecosystem that needs to be built around such "non-standard" CPUs. From an economical pov it doesn't make sense to create a "computing island".

Another reason to not copy what the west was doing was the deliberate crippling of early systems. It's difficult to find this information on the web, but I recall that the early IBM PC/AT machines ran at 6MHz and contained BIOS code that prevented them from booting if they were clocked any faster. (Later versions ran at 8MHz.)

The reason was to prevent people from buying IBM PC/AT machines instead of IBM RISC 6000 workstations.

I think early 286 chips didn't go up to 8MHz. The IBM RT PC (the ancestor of the 6000) was only introduced 2 years after the AT:

https://en.wikipedia.org/wiki/IBM_RT_PC

Hmm. The Tandy 2000, Radio Shack's ill-fated first foray into the world of MS-DOS in 1983, used an 80186 (yes, 186) clocked at 8 MHz, so it would seem odd if the 80286 couldn't have at least matched that. From what I can find at a cursory search, the 80286 first shipped at 5 MHz, 6 MHz and 8 MHz versions, although the AT apparently used the 6 MHz version when it first shipped in 1984, not bumping up to 8 MHz until 1986.

While there was plenty of overspec in the AT design, it was built around a CPU rated by the vendor for 6MHz operation. You don't want the support headaches when some local reseller cracks the case, swaps in faster crystal, and things mostly work. There were aftermarket speedup circuits which you could turn on after the POST check completed; I recall one with a knob offering like 8 different speeds.

I always wondered why my friend's dad's computer had a TURBO button on it, thanks!

The west had superscalar computers around a decade later but superscalar designs didn't make their way into ICs until the 90s for the simple reason that chip fabrication technologies simply weren't there for making high-density ICs. The complexity and cost of the SECM was likely one of the major reasons it was shelved in favor of more straightforward in-order processors. Your alternate history would be predicated on several decades of semiconductor fab advancement happening in just one or two within the Soviet Union, which, regardless of where you stand in terms of ideology, simply isn't a realistic assessment of the facts.

Don't forget that Tomasulo invented his algorithm in the 60s at IBM - e.g. vaguely modern out of order execution

Site not accessible.

The 8080. If I remember well, there was no instruction to adjust after BCD substraction (as with the Z80) So you had to do 9's complement the second operand and do an addition.

https://archive.is/lJirU

Reply on news.ycombinator.com