Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
Technology

Russian E2k CPU at 135 SPECint95 / 350 SPECfp95 ??? 106

Posted by sengan from the can-someone-verify-this? dept.
jpatters tells us that Micro Processor Report is reporting (via MacInTouch) that a russian company (Elbrus International) claims to have a CPU design that achieves 135 SPECint95 and 350 SPECfp95. This compairs to Merced's scores of 45 and 70 respectively. It is claimed to run in a 0.18 micron process at 1.2Ghz consuming only 35 watts and 126 square millimeters of silicon. It includes a 256 Kbyte of on-chip L2 cache. It should also be both x86 and IA-64 compatible. Elbrus 2000 seems to exist (look at what Shevtsov is working on now), and seems to have had some history. Here is Shevtsov's FPU patent. S : I've tried to verify this story, but can't find the copy of MPR -- anybody else have it? Anyone care to speculate how it was done? Assynchronous logic? 256Kb L2 seems rather low though unless they're using a special point-to-point bus.
From an Anonymous Contributor

"I get MPR. I've got about 7 minutes before I have to catch a bus, but, from the MPR issue itself:

The processor uses EPIC. The Elbrus team has been together for 40 years, originally designing supercomputers for the Soviet defense establishments. "They've developed computers based on superscalar, shared-memory multiprocessing and EPIC techniques long before papers on those subjects appeared in the West". MPR claims that the lack of a good semiconductor Fab has been what was holding them back. MPR says that the claims would be unbelievable except for the credibility of the team.

The X86 and IA64 compatibility rely on binary compilation assisted by emulation hooks, similar to what Transmeta is apparantly doing. Supposedly Dave Ditzel spent several years while at Sun working with the Elbrus team.

The processor exists only as an executable Verilog database. However, the E2K design is based on the Elbrus-3 processor that was fabricated in 1991. The Elbrus-3 was built in an "ancient process", used 15 million transistors in about 3000 LSI and MSI chips, and delivered twice the performance of a Cray Y-MP."

Some more he sent later:

" It is actually quite a long article... 6 pages plus the cover, I'm about two thirds through it. The architecture is in fact pretty stunning, and very similar to the Merced and the SPARC in several ways. It has a 64K, 4-way instruction cache: one i-cache only. It has two identical, synchronously-loaded 8K L1 data caches, and a 256K, 2-way, 4-bank L2 data cache. In addition, it has a 4K array pre-fetch buffer for use in loop overlapping. There are two regions, each with an L1 data cache, a 256-entry register file, and three ALUs. The regions are symmetric except that only one region has a divide function.

A great deal in this processor is left to the compiler, a fact that is demonstrated by the single, 64K i-cache; this will only work if the compiler does its job. Much also depends on the compiler's ability to identify instructions that can be executed in parallel. With an optimal instruction load, the multi-ported caches can provide a potential operand bandwidth of 288 Gbytes/sec at a processor clock of 1.2GHz. Much effort is expended to avoid branching; extensive branch prediction support is provided, and in some cases it will actually just go ahead and execute both sides of a branch to avoid doing the branch at all; with so many parallel execution paths, the cost of doing so is much lower than what would be the cost of branching.

When loops are identified, an effort is made to overlap the loop execution, taking advantage the same mechanism as used for the sliding register windows. The 4K FIFO Array prefetch buffer helps to feed data to the overlapped loop. In loop mode, for perfectly optimal code, the processor can rates as high as 23 operations per cycle.

Much of the processor is designed in standard static CMOS gates, but some of the critical paths through the processor use self-reset gates, which do not have a clock but rather are triggered by the completion of cycles in previous gates. According to MPR, these are estimated by Elbrus to run 10-15% faster than static CMOS gates.

Just a couple more facts about Elbrus: The Elbrus-1 computer was a "...superscaler, RISC, processor with out-of-order execution, speculative execution, and register renaming..." This machine was designed and built... between 1973 and 1979!! They dumped superscaler designs becuase they were too complex for the payoff. The Elbrus-3, built between 1985 and 1991, used "an EPIC-based VILW CPU", implemented as a "16-processor shared memory system"

They started working on the E2K in 1994, and it is now at Verilog RTL stage, with compilers and binary-compilation software written. MPR expresses great doubt that a home will ever be found to build the processor, what with the Russian economy as bad as it is, and most capable semiconductor houses already in the midst of implementing their own designs or just not wanting to compete with Intel."
This discussion has been archived. No new comments can be posted.

Russian E2k CPU at 135 SPECint95 / 350 SPECfp95 ??? More

Russian E2k CPU at 135 SPECint95 / 350 SPECfp95 ???

Comments Filter:
  • Posted by BikE_PUnX:

    kid, how old are you?

    russia has one of the most advanced aerospace programs in the world. remember, they managed to go from nothing to putting a man in orbit in less that 50 years. just because they have problems with their currencey (ans yes, that is what their problems mainly are -- everyone has enough food and clothes and heat to survive), do not assume everything else has gone to crap also.
    ---
    "If you can't fix it with duct tape, it's fucked."

  • Yes, the west laughed about those vacume tubes in military hardware for years. Then we thought about EMP (electro-magnetic pulse). IIRC everyone got really quiet for a few years after that.

    The belief that Russians cannot out innovate an American company is no different that the U.S. government's apparent position that only Americans can create strong crypto technology.

    The USSR got a man in space first. The USA put a man on the moon first, USSR landed a probe on Venus first, US made it to Mars first, USSR holds the record for space station service time, etc, etc... The space race is really a tie in the long run.

  • Want some better links?


    http://www.ipmce.su/ [ipmce.su]


    http://www.el2000.ru/ [el2000.ru]

    K, I found them, now someone translate for me? :-)

  • Sorry I didn't write it all down, it's long.


    Basically, I think it's BS. It's not real, it's vaporware, not hardware. There is no existing stuff, it's "planned" for thier "project." They are partners with Sun, BUT.... The project is not even lead by a Russian, it's lead by an Armenian(sp?) guy, and Russians have a trust issue over that, because it might not even benifit thier country (which could use a technological boost).

    Anyway, that's the spin I get from http://www.el2000.ru/news/presentation.html

  • That raises a red flag for me-- when did Intel make the IA64 instruction set available to other processor designers?
  • Sun microsystems lists them as a partner [www.sun.ru]. They used to be the Moscow Center of SPARC Technology.
  • What the hell you mean "where it's not needed"?

    It's needed everywhere.

  • I've always knew that military didn't use the crap stolen by KGB from West (i.e. OS/360, VAX clones)...

    Now I at least know the name :-)

    If the soviet government wasn't so paranoid about their military technologies we could really have another Silicon Valley now..
  • I was unaware of any computer hardware coming from Russia.


    --
  • "Pentium" is just a marketing name.


    --
  • RISC chips have traditionally larger caches to be able to cope with the larger number of instructions. Newer Alphas have 4+ meg of L2 cache.


    --
  • I've known for some time that the Russians had the same level of scientific and engineering that the rest of the world does (I wouldn't be surprised if China is right on up there with us- Nature reveals her secrets to anyone who asks the right questions...) but that the USSR happened to be hampered in a few critical resource areas and as such, some things suffered (like semiconductor tech)- while other things flourished like optics (they make some of the finest in the world!) and tube technology (don't laugh- tubes are intrinsically immune to EMP attacks that'd take out solid state devices and they are the ONLY route to high power levels.).

    I'm surprised that the people behind this design haven't contacted the people that got behind Svetlana Electron Devices and got something going with them. (Svetlana is one of the largest suppliers of vacuum tubes in the world)
  • How can they say what the SPECint95 and SPECfp95 ratings will be when they don't even have a prototype? And what secret knowledge have they scryed from Intel and HP that they know enough of the inner-workings of the Merced chips to tell what those specs will be?

    We might as well read the Farmer's Almanac to determine this "supposed" chip's ratings. Their
    design may be innovative or it may be non-existant, but we won't know until something
    is done with it. Blech.
    --
    Aaron Gaudio
    "The fool finds ignorance all around him.
  • I was unaware of
    any computer hardware coming from Russia.

    See this page [mailcom.com] on the BESM-6 computer.

  • Wasn't that just a question of economics rather than technology? Given that they couldn't afford to feed themselves, how do you suppose they were going to afford all that kerosene?
  • Well, this is not a hoax. The chip is based on a
    post-VLIW design. Apparently it is
    5 seconds close to be baked in the silicon.
    Some details are available at the Elbrus
    web site: www.el2000.ru [el2000.ru] (in russian).
    This company was building computers since 1959.
  • There have been too much self-indignation on how Asia having to rely on Silicon Valley on everything....

    Now that the Russians are proving that they can produce something this amazing even without a good semiconductor Fab.

    Those of you in Japan, Taiwan, Korea, China and India better take heed --- you've got the Fabs, why don't you work with the Russians and produce this amazing chip?

    Are you listening, Asia?

    This is the BIG CHANCE for you now !!!

    JUST DO IT !!
  • There have been too much self-indignation on how Asia having to rely on Silicon Valley on everything....

    Now that the Russians are proving that they can produce something this amazing even without a good semiconductor Fab.

    Those of you in Japan, Taiwan, Korea, China and India better take heed --- you've got the Fabs, why don't you work with the Russians and produce this amazing chip?

    Are you listening, Asia?

    This is the BIG CHANCE for you now !!!

    JUST DO IT !!
  • From the "about page":

    They are working with Transmeta on "commercial projects, related to creating apparatus and programming environments for most modern calculating architechtures" (or something to this effect)

    Interesting...
  • Actually it said it would be about 5 times faster than intel's merced, which is at least 2 years off.
  • Entirely possible they've been doing work for Western companies under contract, and have all the software etc. Isn't that hard to envisage.
  • Moore's Law actually only indirectly affects CPU speed.

    Moore's Law states that transistor density doubles every 18 months. Finding how to use all those transistors usually results in far less than a 2x speeded.
  • Strikes me that it's very probable that all us North American folk have been fed a line of bullshit by the media and governments all these years. Little wonder it's shocking news to hear that 'the Russkies' have advanced technology -- we've been damn near programmed to think they're a bunch of ignorant rat-bastard Commies who deserve to starve for not kow-towing our brand of government. After hearing this bit of CPU news, I'm now curious to hear about what other bits of advanced technology they've invented/been working on/kicked our ass on...
  • Fountain pen.

    Capillary action. No need for gravity. Ne pas?
  • wouldn't it be nice to get that E2000 design under GPL and got a Free CPU that kickes ass in a same way as Linux does? I would feel much more like a free person then. But... it's not likely having Elbrus key people military industrial complex heritage. So, as long as it's not Free it's of not much interest really. Since I would agree that chances of Elbrus team pulling it in Russia is quite slim. If they come to the West for that they would be used and abused. Not by our community of course :-) ... And we need to put more effort into Freedom CPU project [tux.org].
  • I think they are more than a Sun partner. I remember Sun buying them in perestroika days. I also remember HP being quite upset since they were going to buy it, but Sun somehow beat them to it. That is what I remember anyway.

    Sinan

  • Hm.. (Score:1)

    by Axe ( 11122 )
    Should ask Babayan's (Elbrus inventor) daughter, Osana, about that. Last what I heard from her, her dad was not very happy with how his team was working. Luck of funds etc.
    BTW - she leaves in Palo Alto... Rather natural.. Called brain drain.. (Say hello to myself)
    :)
    Hope she does not read this... :)
    In case she does - hi, how about Kirkwood this weekend.. :)


  • Ok, It's fine, it's dandy, it's not even seen the silicon yet.... this thing is 100% vapor, nothing more than a set of design goals really.

    I'd like to see some hardware manufacturers quit talking about the far and wide and start talking about the here and now. Give us new high performance hardware that will be on the market soon, it is nice to see what's coming up, but c'mon, nice concept, lets see if it works when we build it... That just doesn't work for me.

  • Is this anything like their Buran, which was supposed to kick our Space Shuttle's butt, but never got beyond the testing phase - all the while ours were launching THEIR satelites??

    I'll believe Russian cutting egde technology when I see it. Until then, Yuri Gagarin was their greatest scientific achievement of the century, and that's old news. They certainly DO NOT lack the brain power, their scientists are brilliant. But with little resources, you can't bring ideas to fruition. This is an attempt at investment funding.

    Sorry to the Russian readers, but if your neighbors are burning furniture for heat, you're not 'cutting edge'. Yes, we have homeless too, but not for lack of resources.

    Theory is nice, publish it. Practice is better.
  • But at least I learned something in the precess.

    Cheers!
  • Come on. I'd like to think that Americans have grown up a little bit since the cold war. BTW, communism is actually a VERY good concept. Support everyone, regardless of their abilities to work. Welfare, Social Security, Unemployment are all basically communist programs. Communism is great in theory, it's just fallen apart in every real world trial. It's really a shame. I personally would like nothing more than a world in which money doesn't exist.
  • is that an eastern company will pick up and make this chip if western companies don't. Someone like Hitachi. I think they'd love to get the jump on something so high-end. Well, if the chip is actually producable, and the results are real.
  • Intersting considering russian silicon technology wasn't very competitive. I guess necessity is the mother of invention. Put it togeather with a modern fab and you have something truly world class. Aint hybrid vigor wonderful.
  • The russians also once claimed to have control over cold fusion technology.

    As any true skeptic would say, I'll believe it when I see it.
  • It was something like this:

    Rus. politicians were freaked because of an imaginary technological gap between Shuttle and soviet rockets, so they ordered to immediately make 'our own'. Many design solutions were copied from Shuttle and...
    good 'ol rockets are just as good and cost twice less. Also, Mir payed off while Shuttles haven't.

    I never heard of russians doing anything good in consumer electronics. I remember when I was in Russia, 99% of computers were made from cheapo asian parts while 1% was compaqs, ibms and such. When you talk about defence technology, Russia was severely lacking development until '63, when Stalin died. Before that, Cybernetics was not considered 'real science'. Still, this chip is possible, if you consider that defence technology is being transferred to consumer market ever since the USSR breakup. This definitely looks more plausible than recent hypercomputer sensation.
  • I agree..

    The article that started this thread was saying that: Buran was a failure, even though it was copied from our shuttle but they couldn't pull it off, hence they can't design the chip on their own.

    My point was that Buran's failure is due to the fact it was mindlessly copied from shuttle.

    I don't know who's space program is superior.. I think even experts would have trouble proving that one side is better, due to the fact that space programs are very complex and multi-faceted and an expert who is sufficiently aware of one field in space program wouldn't know much about other fields.

    I definitely wouldn't even try to say which program is superior.. its impossible to say just by following news or reading a few books targeted at general population.
  • This seems like a classic case of vaporware. They have no Western sponsors and are admittedly several years from production. It never ceases to amaze me how often people will fall for this trick: "Our CPU will be twice as good as Intel's best current offerings! Look for it in less than two years!" Of course, given Intel's track record and Moore's law, Intel will have something twice as good as their own current best in 18 months. Meanwhile, though, their competitor gets lots of positive press.

    Anyway, the Pentium was actually declassified military technology that Intel obtained in the early 70s; they've been releasing it piecemeal ever since.
  • Pardon my ignorance, but what Intel chip has ever had more? The Pentium Pro had >256K L2 cache, but it was on a separate chip (though it was in the same package). The latest Celery^Hon has 128K. Are there non-Intel processors that have more than 256K L2 cache included on the chip?
  • Err, the PPro is and was available with 512K and 1MB L2 caches, while the P-II has 512K, and the Xeon is availabe with multiple MBs of cache.
    Err, the PPro had ZERO L2 cache on the IC inside the pin package. If you open up the pin package you found one square of silicon for the CPU and another square of silicon for the cache. P-II has L2 cache in a separate pin package, though it's inside the cartride. Same is true for Xeon.

    Meantime, some PA-RISC, SGI, and Alpha systems have more L2 than some of us have system RAM.
    So are those on the same chip as the processor, or a separate chip?

    My original question stands unanswered.

  • Not quite correct. In the beginning, CPUs communicated directly with the RAM. When CPUs got faster, L1 cache was implemented. It was NOT on the CPU chip. It was on separate chips, elsewhere on the motherboard. With the 486, 8K of L1 cache was on the CPU chip. Many motherboards also added L2 cache. With the P-II, the L2 cache was taken off the motherboard and put on a separate board that also held the CPU. The Celeron has both L1 and L2 cache on the CPU chip. This is the situation that would appear to be the case with the Russian chip - both L1 and L2 cache on the CPU chip...

    ...or it could be just a typo.
  • Yeah, but you've got to reduce the chip size by a factor of four to reduce your linear measurements by a factor of two. Theoretically, that cuts the time the signals spend moving around by a factor of two. But you've also got to worry about the latency within each transistor, so transistor density alone isn't the only factor.
  • Yup. That's why my original statement was "given Intel's track record and Moore's law"

    Of course, one might argue that Intel's track record is no predictor of future performance, but that's true of all empirical evidence, including Moore's law.
  • I found their website: http://www.el2000.ru [el2000.ru]
    It's all in Russian. I attempted to figure it out using my very basic school russian. They seem to be a Sun distributor for Russia.
    I think the new chip is described here [el2000.ru]. At least I think it is.
    I can understand individual words but not much more than that. So if some one knows Russian. Please translate.
    -Pelle
  • Very abbreviated translation of http://www.el2000.ru/company/comp.html

    Elbrus is a spin-off company (actually tree companies) of Institute for Precise Mechanics and Computing Equipment (IPMCE), who designed supercomputers used for Soviet missile control, missile defence, nuclear research and other applications. R&D people are mostly from IPMCE.
    ...
    From the history of IPMCE

    1957-59 : M-40 - a computer on vacuum tubes, used in first successful missile-defence test.

    1964-69 : 5E92B - 2-processor system on ICs, main component in first missile-defense system
    of Moscow.

    1973-79 : ELBRUS-1 - first SMP system on medium-integration ISc, 10-processor system.

    1977-84 : ELBRUS-2 - SMP supercomputer, based on ECL (?) high-integration ICs, 10 processors,
    used in Space Flight Control Center, in nuclear research centers, for military applications.

    1985-94 : ELBRUS-3 - LSI, ECL high-integration ICs, 16 processors, 2 times performance of CRAY-YM. Wasn't put in production for financial reasons.

    Some of IPMCE design achievements

    1955 - high-speed arithmetics
    1964 - fault-tolerant non-stop architecture with full hardware control (?)
    1979 - SMP system with shared memory, 10 CPU
    1979 - protected programming technology, hardware support for data types
    1986 - an architecture with explicit parallelism
    1986 - binary compilation technology
    ...
    Novosibirsk and StPetersburg offices were formed in 1992, and work mostly on compilers and Java technology.
    ...
    MCST (Moscow Center for SPARC Tecnologies) - a part of ELBRUS group, 400+ employees
    ...
    Recent designs :

    - a SPARC-compatible CPU for military applications
    - universal high-performance system ELBRUS-90micro, works with 100+ different network and I/O-adapters.

    Since 1992 MCST works with Sun Miscrosystem: design and support of compilers, operating systems, perspective workstation design, Java technologies.

    Long-time partnership with Compass Design Automation B.V. (a VLSI Technology, Inc. company), a VLSI CAD specialist.
    ...
    Currently MCST works on new E2K CPU design, using postRISC architecture with explicit parallelism and binary compilation technology, providing effective and reliable execution of Intelx86 binaries, with estimated performance of up to 2 times higher than Intel's P7 (Merced).
    ...
    Key elements of the project are CPU architecture and a complementary "superoptimizing" compiler, which permits fullest usage of hardware parallelism.

    Another important direction is improvements of binary compilation technology, which provides 100% reliability and high efficiency of binary compatibility modes for x86 and SPARC architectures.
    ...
    MCST also works on number of other joint projects in hardware and software areas, with Transmeta and other companies.

    ************************************************
    So those guys certainly have know-how, experience and gray matter to design kick-ass computers.
    Manufacturing is another story...

  • Sorry, but I've never worked for Transmeta.
    Oppositely, some guys who are now at Transmeta worked closely with us some time ago.

    Very clever guys they were and are! Luck to them!

  • Yes on necesity but it shouldn't be too surprising given the number of "minds" that create most of the profound technologies aren't that many and they tend to be a bit scattered across the globe.

    I look forward to watching this one unfold.

Slashdot Top Deals

"Your stupidity, Allen, is simply not up to par." -- Dave Mack (mack@inco.UUCP) "Yours is." -- Allen Gwinn (allen@sulaco.sigma.com), in alt.flame

Close