New russian 8-core CPU | Hacker News Reader

New russian 8-core CPU

mcst.ru

68 pointsoxama55 comments

This is entirely hopeless - their development budget is 0.05% of that of Intel. So these will never leave drawing board to become usable product, and this is not a field rational to invest in - general purpose processors are olygopoly market with decades long learning curve worth too much. Even if someone puts hundreds of billions dollars and decades of time to become competitive, this will only diminish returns in the industry as whole so will be anyway bad investment: at best they could sink Intel into unprofitably small ROIs, and get similar ROIs themselves. So aside from pure politics ('Russia is the motherland of elephants') i see no purpose for anyone doing it.

The idea of boosting national pride might be the end goal in mind. They don't need commercial success. If they want to convey the idea to the nation that we can design and build on par and maybe even better than the rest of the world. Just feeling good about yourself has benefits.

Not even that: technological independence is the goal. Some national security system may come to depend on a specific chip — which Intel will discontinue, what then? Also, the government may simply want to hide the amount of CPUs it utilizes. Impossible with foreign suppliers.

Like sending a man on the moon, right? But that one was kinda cool, maybe this one might be, too? Anyway, don't we all agree that microprocessor space could use more diversity?

I guess it is purely political.

Russia wants to be independent from the US for computers, because they fear (rightfully) American chips to be bugged.

Having slow, expensive chips made in Russia is probably more interesting to certain agencies than fast and cheap American chips - or relying on typewriters :)

Yes. Bugging your own chips is cheaper than bugging Intel's.

China has a team developping a mips like architecture, with several produced iterations: https://en.wikipedia.org/wiki/Loongson

I don't know if Russia has the industry to mass produce it, but surely it has the skilled people to design it, and it may probably use foreign fabs for the remaining steps if necessary.

As for the software, with an established ISA, they may quickly leverage exising open source solutions. Of course, Windows and OSX support will be a problem (given that they are the major desktop players).

Politic or not, they know it's doable.

If they are concerned about bugged hardware (which is difficult to do and expensive), I don't think "bugged" software (which is much cheaper to replace) will be on their shopping list either...

True, what I meanbt is that the consumer software marketplace is originally built around those platforms. During the last several years, the market has been largely fragmented by tablets, but I have the impression they still represent a major segment. That said, industries and services may more easily accomodate the absence of these players.

If you're just worried about bugged hardware, why develop a new architecture? Can't you use your own clean room implementation of an existing architecture?

Yes, you're right, they could. But I suspect the decision is made of things like:

- If it proves to be actually good, they might actually sell some of those chips, too (it's developing their local high-tech sector. Even a tiny piece of the market is better than none at all, especially if their own agencies kickstart it).

- It highers the cost for foreign intelligence to write software for that platform: their arsenal of viruses and trojans are completely useless. Unless you acquire detailed specifications of the platform, find some test chips, develop new software and finally infect your target... Takes time and money.

- They get some (positive, for once) international press.

- They can parade how awesome mother Russia is to their own nationalist ego.

Someone else mentioned that the architecture is based of Sparc technology (Sun's line of CPUs, which is a mips derivative iirc). It's hardly a completely new architecture.

Sparc isn't a derivative of MIPS.

Thanks for correcting me: indeed, it's an open architecture which was used by Sun, Fujitsu and TI for their CPUs, and it's unrelated to MIPS.

However, how difficult would it be for a foreign fab to rig the design of a customer and include backdoors?

The Americans should probably fear bugged Russian processors too, right?

We are more afraid of counterfeit china-fabbed ASICs; even the US military has been burned by these. With more fabbing done in china, I guess the problems could get worse.

Well yeah, if they were to buy made-in-russia processors. Which is very unlikely.

This would not be hopeless to the people who own the chip's design and production facilities, if the government also passes a law saying all government-related work must be done on computers with this specific chip installed. If the government can give a company a complete, enforceable monopoly, this would make at least some sense. Still, a total waste of resources.

In the long run, not only are they (the russians) going to use it in place of any foreign CPUs but lots of states which currently have no options would prefer to buy processors and software from the russians rather than the americans (Brazil, Iran, Syria, Cuba and many other "not aligned" countries). Russians/Soviets have a long history of having parallel development of technology (including special purpose microprocesors) and no one really considered that as a "waste of resources".

But look at that list of customers - hardly rich nations. Total sales? Millions. Profit - around zero.

My opinion is, first, that this development is fueled for political and national security gains mainly and the commercial aspects of it are lower on the list. And second, if you try to measure the richness of those countries in dollars, they might be poor because they don't have dollars nor totally free access to markets where to obtain those dollars but still their natural resources, the work force and their political support (for the Russian interests) are valuable assets.

But none of those things will buy CPUs. So the political issue is the only real one.

Just because those nations aren't considered "rich", it doesn't mean they're penniless. Russia has a long history of profitable trade to that list of countries. They sold US$1.1 billion in goods to the Syrians alone in 2010. So the idea that they couldn't build an industry selling a good that those countries can't or won't buy from the West doesn't hold water.

Its a question of opportunity cost. Make a CPU to sell to Syria - you might make a few bucks (though I doubt it); could you have done better? Surely.

As for trade stats: the US SHOE industry is an order of magnitude larger than that - $20 billion annually {http://statisticbrain.com/footwear-industry-statistics/}. 1.1 billion seems like a big number until you consider: a chip manufacturing plant costs more than that.

Took Econ 101 and got a B-. Got it. This isn't a simplistic, classical microeconomic analysis.

First, just because a multi-billion dollar market exists for a product, that doesn't mean that there aren't highly profitable, multi-million dollar markets that exist in the same industry. This is especially true when the "market" is impacted by things things like trade policy asymmetry, sanctions, embargoes and national policy.

Second, there's no reason to think this "market" is motivated solely or even primarily by revenue/profit in the classic sense. If you add the "benefit" of poking the West in the eye, and the national pride angle, the alternative market decision isn't "I could make money doing other things with this money". The whole history of the Cold War proves this out.

There's a reason MiGs and AKs are common as cockroaches.

Third, it doesn't matter what a fab costs to build, unless you want to build one. This is an arbitrage issue; there's plenty of non-Western places that have first-world fab capacity that would be more than happy to crank out Russian IP all day long. More importantly, they have local, non-Western talent to do the heavy lifting associated with tailoring Russian IP to fine geometry fabrication.

Ultimately, this is a political, not economic decision. And it doesn't matter if Intel has a bigger/faster/stronger processor (I agree with you that it will). It matters that the Russians (just like the Chinese have done for a decade with Loongson) can legitimately say "we have a locally produced alternative to the Western hegemony". And let's not forget that the Chinese have taken what amounts to a licensed MIPS derivative with a fraction of Intels R&D budget and produced top-10 TFLOP supercomputers. Also see: ARM.

And I want to point out something that I don't see mentioned that much. If you read the proceedings of HotChips or one of the other conferences, not all, or even most, of those papers are being submitted by Westerners. While they work mostly for Western companies now, it's not like we're the only source of innovation in the world. You only have to look to ARM to see how a modest investment by wildly talented engineers can toss the market on it's ear over time.

Savings due to data secrecy - billions. USA not knowing their military plans - priceless.

A niche product operates under a different definition of "competitive". They seem to be going for trusted computing, with a focus on security, which isn't exactly something Intel CPUs are known for.

In addition, more CPU choices will always mean more flexibility for hardware startups. Just think about how hard it is to source a CPU in sufficient quantities. A hungry manufacturer eager to gobble up market share from others might be more willing to accommodate a startup's needs.

You are looking at it from an entirely different angle. You don't have to be Intel. Within 3-4 years you can have a neat processor comparable with current crop of mobile processors - assuming you have a good team & funding, no commercial pressure & can trust ARM.

That's the time frame it takes to setup & go to market with a mobile processor company in silicon valley.

Sure but if you decide to not compete with Intel and instead compete with Qualcomm, then the same story applies: 3-4 years from now Qualcomm will have the Triple Dragon Q-Zap 500+, built in Intel's foundries at ~12nm, and the Russians will have this.

China has already shown interest in using domestic processors simply because they are domestic. Russia is probably the same. I don't think Qualcomm could be considered competition unless they suddenly turned Russian.

You're missing the point. They aren't doing it for commercial success. You don't have to compete with Qualcomm or Intel or anyone. You just have to be able to produce processors that can handle the government issue about not trusting US made processors.

Like producing an ARM or Sparc or MIPS based design? Why would that be unimaginable? Qualcomm, AMD, Nvidia, Samsung and Co. do this. You're overestimating this and you're claiming that this processor would be new from the ground, which i don't see proof for. If it's a Sparc based design as others have claimed in this thread, it might become a slow, power-consuming processor, but it might work quite ok.

The ISA and even the logical architecture is easy. Laying it out and producing the chip is not, at all. You need a few billion $s investment in equipment AND people (you can poach the latter, the former is much harder to buy beyond outdated processes).

China is better off taking over Taiwan, who have much of both.

No need for the latest and greates from the R&D departments of Intel. Even if the CPU will be at the level of some older Pentium models, it will be good enough to run office suites and Linux in some government departments. More then enough.

Very much depends on the meaning of hopeless. State of the art power management and/or performance? Hopeless.

Taking this on outside of a consortium and using a novel "dynamic translation" architecture? Way not optimal.

Adequate performance for high-security requirements? Not so hopeless. Building and operating secure systems will costs thousands to tens of thousands per seat, so a few hundreds of dollars to know your CPUs are not back-doored is cheap in the greater scheme.

It could succumb to a wide range of reasons for failure. But making an CPU that doesn't have to be price/performance competitive is not out of reach for any of dozens of nations, if they feel a need to do it.

No offense, but have you ever seen a successful startup? Your (very mistaken) arguments apply to literally every industry, ever - plus, you even give, then discount, a viable path to market for them at the end of your comment!

There can be a successful startup in the processor arena. But it can never succeed by trying to do same thing better/cheaper. You want to offer replacement product on the other hand - very different but serving the same purpose.

You don't succeed in automotive industry by trying to directly compete with BMW, this is absolutely hopeless. Tesla has a real shot on success on the other hand.

Same here: next Intel won't be doing anything similar to Intel. But someone offering real quantum computing may get there. Someone making a chip with thousands of simple cores and somehow 'bringing Moore's law back' [1] may get there. But this one is just too similar. Intel's established, polished business processes on all levels can easily beat this.

1. http://www.paulgraham.com/ambitious.html

Sorry, I just find your argument without any merit. Contrary to your claim, Tesla does compete directly with BMW, for the same sales dollars, in the same part of the market, providing a solution to the same audience, and even segment, and in the same problem space. If we can't agree on even the definition of "compete", it's unlikely that we can further each other's understanding.

Also, the idea that 'if they succeed' (not a quote) then " this will only diminish returns in the industry as whole so will be anyway bad investment" (your comment) is just contrary to the experience of any other startup that succeeds.

It's just bad economics. I would question whether you've ever seen a successful startup - there is no shame in saying you haven't. But if you want to contribute meaningfully, then IMHO I would look into some.

Sorry if I'm blunt.

Just a note that despite a downvote on this response, my specific example is enormously vindicated 3 days later as Tesla just announced the Model 3, which sites headline as "Tesla Model 3 to challenge BMW 3 Series", writing: "Tesla founder and CEO Elon Musk has revealed plans for a new electric BMW 3 Series rival, including its name – Tesla Model 3."

https://www.google.com/search?q=model+3

So they should just accept dominance of the US in this industry and not even try to advance their own technology? Is this your suggestion?

Why should they do that? So that they can be pressured by possible sanctions in the future? (which wont work at the moment, since Russia is a strategic partner of China - but it might change in the next decade)

Running through Google translate [1] gives pretty sparse press release on the Elbrus-8. Interesting bits are 'binary translation of x86' and '25 instructions at one' so a 25 instruction super scalar pipeline and dynamic instruction translation, historically has been slow as the proverbial pig on non-native code. That arose from an incredible amount of memory churn (MemOPS) when running translated code slowing the processor down to the cycle time of DRAM (110 - 130nS or about a 10Mhz effective instruction rate[2].

I get the nationalist pride in wanting to build your own system but I think they could do wonders licensing the ARM A5x cores and starting from there rather than pushing so hard down the 'everything our self' road. Just look at the success many Chinese companies have had taking the ARM path (Allwinner is probably the most familiar example).

If there is someone reading who is part of this program or can advise them, I'd suggest they focus on building a 28nm FAB capability ala TSMC that they can use to make their own chips. Build a reliable semiconductor process and pipeline, and the use state money to subsidize the costs to pull foreign contracts into your FAB so that you can get a first look at people's ideas. If you are looking for strategic advantage that is a much better play than pushing out a computer architecture.

[1] Possibly translated link: https://translate.google.com/translate?sl=auto&tl=en&js=y&pr...

[2] For comparison this was the speed of the PC/AT 286 machine in "turbo" mode.

I doubt it is 25 operation super-scalar, it most likely is a 25-issue VLIW. Elbrus has been doing them for a long while[1].

I doubt it will be fast on translated code, but Transmeta could do x86-to-VLIW translation reasonably 10 years ago, so it isn't impossible. If they have similar software, it should be quite fast. Of course, the Crusoe and Efficeon are less wide (4 and 8 functional units iirc?) but I doubt x86 is the preferred instruction set for these, it is probably there to allow mixed-mode software (x86 OS + some native applications?)

Since it seems to be built for HPC, 3 cpu-to-cpu links, 4 DDR memory controllers, over 20MB of cache. This probably makes a bit of sense, you could run a regular x86 version of Linux for most applications and support binary driver blobs etc. But then compile certain computational software as native code.

The teams behind this processor probably have built HPC stuff since the seventies[2], I doubt they would gain anything by switching direction.

[1] http://www.realworldtech.com/elbrus/ [2] http://en.wikipedia.org/wiki/Elbrus_%28computer%29

It's ironical that the top rated comment on HN for the past few hours is going against the "hacker" culture just because it's a government that's doing it?

There's no need to for them to match Intel at any level. The current generation of tablets & mobiles is ample proof that an open source OS and with ARM based processor is a good alternative for technical productivity and may very well replace Intel/MS for generic office tasks in the future.

As a country in post Snowden/NSA era, it makes absolute sense if they don't want to rely on US companies.

Come on, lets be neutral to this (politically). We should look at this as a new era. So what, If the Russian design today is more expensive or slower than Intel or anyone else.

Tomorrow, we might all be using Russian-made processors. History requires only a moment to shift the balance created over our short lifespan.

Besides, a little bit of competition between US and Russia has sometimes helped the humanity as a whole. Take the "Space Race" for example. A bit of egotistical competition meant that Apollo 11 landed on moon, many years early than if there was no such competition.

This doesn't seem to be a new ISA, but a continuation of Elbrus E2K. RWT covered Elbrus E2K back in 2000.

http://www.realworldtech.com/elbrus/

Seems not to be "new" but an implementation of a Sparc V9 ?

Elbrus team worked on that architecture before collaboration with Sun - they can continue to improve it now. So it may have something new?

I remember two or so years ago, Russia endorsed ReactOS.

But if they are making their own architecture, I doubt programs will transition for them if they plan to still go that route.

This is the one based on ARMv8 ISA, right? It doesn't seem to mention it there.

I think that one may be from another company, though:

http://www.tomsitpro.com/articles/russia-arm-intel-amd-proce...

No, this one is different. T-Systems, who was making supercomputers with Intel chips, plans to make "its own" ARM-based chips. Elbrus is another, older, company that is using either SPARC or their own architecture. http://en.wikipedia.org/wiki/Elbrus_(computer) http://ru.wikipedia.org/wiki/%D0%AD%D0%BB%D1%8C%D0%B1%D1%80%...

Could be this one:

http://www.computerhistory.org/atchm/the-elbrus-2-a-soviet-e...

> Система программирования платформы поддерживает языки С, С++, Java, Фортран-77, Фортран-90.

its interesting they called out support for fortan 77 and fortran 90

It probably just means they ported gcc, but if it is supposed to mean more its probably just 'HPC is on our radar'

If the Russians are developing this chip for their infrastuccture and mission critical aspects, FORTRAN 77 and 90 support makes sense. Finite Element Analysis software (CATIA, SolidWorks and the like...) are Fortran or its ilk..

Same goes for HPC.

Reply on news.ycombinator.com