Russian-Made Elbrus CPUs Fail Trials, 'A Completely Unacceptable Platform'
SberTech, a technology arm of Sber, Russia's biggest bank, has evaluated the Russian-made MCST Elbrus-8C processors in multiple workloads, but the results were utterly disappointing and the processors failed the test. The testers cited "Insufficient memory, slow memory, few cores, low frequency. Functional requirements not been met at all" as key reasons for the failure. However, there is hope, according to SberTech engineers.
[...] "The Elbrus-8C server is very weak compared to Intel Xeon 'Cascade Lake'," said Anton Zhbankov, a representative for SberTech, said at the Elbrus Partner Day conference (via ServerNews.ru) earlier this month. "Insufficient memory [256MB], slow memory, few cores, low frequency. Functional requirements not been met at all."
[...] In fact, SberTech's evaluation was the first in-depth testing of the Elbrus-8C platform in a banking application. The evaluators compared dual- and quad-socket Elbrus-8C machines (16 - 32 cores per box) to a dual-processor server based on Intel's Xeon Gold 6230 processor that the company currently uses. SberTech could not test the more powerful Elbrus-8CB as it is still not available despite being formally introduced.
[...] "One of the surprising things about the Elbrus-8C server was that it is a real product," said Zhbankov. "It was a real server that we were given. [...] It is an actual product that has its disadvantages, loads of disadvantages, but we can work with them."
[...] [While] SberTech's engineers expected the Elbrus-8C machine to perform much worse and be orders of magnitude slower than Intel's Xeon Gold 6230 machine from 2019, even a two to three times performance difference is significant enough for commercial companies not to deploy a platform since it makes no financial sense. "At the moment, Sberbank says no, we cannot deploy Elbrus machines into our ecosystem, but we are pleasantly surprised that it works at all," said Zhbankov.
The complaints were only partly about the CPU's relatively low performance, with problems concerning the build quality of the server(s) being highlighted.
1.3 GHz 8-core, 8-thread CPU does not beat 20-core, 40-thread Xeon. Surprise?
Previously: Programming Guide for Russia's "28nm" Elbrus-8CB CPU Published
Related Stories
Russia's Elbrus 8CB Microarchitecture: 8-core VLIW on TSMC 28nm
All of the world's major superpowers have a vested interest in building their own custom silicon processors. The vital ingredient to this allows the superpower to wean itself off of US-based processors, guarantee there are no supplemental backdoors, and if needed add their own. As we have seen with China, custom chip designs, x86-based joint ventures, or Arm derivatives seem to be the order of the day. So in comes Russia, with its custom Elbrus VLIW design that seems to have its roots in SPARC.
Russia has been creating processors called Elbrus for a number of years now. For those of us outside Russia, it has mostly been a big question mark as to what is actually under the hood – these chips are built for custom servers and office PCs, often at the direction of the Russian government and its requirements. We have had glimpses of the design, thanks to documents from Russian supercomputing events, however these are a few years old now. If you are not in Russia, you are unlikely to ever get your hands on one at any rate. However, it recently came to our attention of a new programming guide listed online for the latest Elbrus-8CB processor designs.
The latest Elbrus-8CB chip, as detailed in the new online programming guide published this week, built on TSMC's 28nm, is a 333 mm2 design featuring 8 cores at 1.5 GHz. Peak throughput according to the documents states 576 GFLOPs of double precision, with the chip offering four channels of DDR4-2400, good for 68.3 GB/s. The L1 and L2 caches are private, with a 64 kB L1-D cache, a 128 kB L1-I cache, and a 512 kB L2 cache. The L3 cache is shared between the cores, at 2 MB/core for a total of 16 MB. The processor also supports 4-way server multiprocessor combinations, although it does not say on what protocol or what bandwidth.
It is a compiler focused design, much like Intel's Itanium, in that most of the optimizations happen at the compiler level. Based on compiler first designs in the past, that typically does not make for a successful product. Documents from 2015 state that a continuing goal of the Elbrus design is x86 and x86-64 binary translation with only a 20% overhead, allowing full support for x86 code as well as x86 operating systems, including Windows 7 (this may have been updated since 2015).
Previously: Russian Homegrown Elbrus-4C CPU Released
(Score: 4, Informative) by sgleysti on Monday December 27 2021, @06:38PM (4 children)
Elbrus-8C seems to be the same thing as Elbrus-8S [wikipedia.org]. These have been in production since 2016, are on a 28nm TSMC process, and use DDR3-1600 RAM. Interesting thing is the Elbrus-8SV is supposed to have double the performance on the same 28nm process at only slightly increased clock (1.5GHz vs. 1.3GHz) but 16Mb L3 cache per core instead of per chip and DDR4-2400 RAM. Wikipedia says the 8SV started fabrication in 2020. Maybe they don't have motherboards for that one yet?
It looks like MCST is lining up a contract to make a 32 core variant on 7nm or better process node for HPC: https://www.tomshardware.com/news/russian-chipmaker-preps-to-develop-32-core-elbrus-processors [tomshardware.com]
I think more diversity in available CPUs is a good thing. Perhaps MCST will catch up to Intel, AMD, IBM, Fujitsu, etc. sooner than we think.
(Score: 1, Interesting) by Anonymous Coward on Monday December 27 2021, @06:56PM (2 children)
It's great that the Soviets are making chips again.
But it seems like a sunken cost fallacy to produce 'legacy' x86 compatible designs when they can't even produce hardware quicker than modest hardware they're trying to replace. And you're still at the risk of patent trolling on behalf of the US government's sanctions against the enemy (e.g. Huawei) if you ever try and replicate anything Intel/AMD have done in the past decade.
Don't get me wrong, I think VLIW code morphing is pretty cool in itself.
Looking to China and gaining self-sufficient IP freedom from the west, the smart money is on Risc-V.
(Score: 3, Interesting) by epitaxial on Monday December 27 2021, @07:40PM
Sounds like Transmeta all over again. That was hyped and failed so hard. Wonder how well AS/400 hardware would fare in this test? That stuff was designed around transactions and doing as many things as possible in hardware.
(Score: 2) by janrinok on Tuesday December 28 2021, @12:21PM
The 'soviets' ceased to exist at the end of the cold war when the 'Soviet' Union collapsed - in this instance you can call them Russians now.
I am not interested in knowing who people are or where they live. My interest starts and stops at our servers.
(Score: 2) by janrinok on Tuesday December 28 2021, @12:28PM
You are of course correct - Elbrus-8C == Elbrus-8S. It is simply a matter of transliteration of the Cyrillic alphabet into a western Latin alphabet. In which case, the '8SV' will also be seen as '8CB'.
I am not interested in knowing who people are or where they live. My interest starts and stops at our servers.
(Score: 0) by Anonymous Coward on Monday December 27 2021, @06:39PM (3 children)
That's a plus, I guess.
But 256MB? Who even makes memory modules that size these days?
(Score: 3, Informative) by pe1rxq on Monday December 27 2021, @07:32PM (2 children)
It is probably 256MB of cache memory.
This would be in line with the other Elbrus CPUs which had slightly smaller cache sizes. (E.g. 128MB L3 on the Elbrus 8SV)
(Score: 3, Insightful) by takyon on Monday December 27 2021, @08:29PM (1 child)
That would be a lot of L3, even if it was divided over 4 chips (quad-socket). Xeon Gold 6230 only has 27.5 MiB. Elbrus 8SV [wikipedia.org] has 16 MiB, not 128 MiB.
I fear the insufficient memory is referring to something more horrifying. I didn't see anything on the linked Russian page.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Tuesday December 28 2021, @12:55AM
These *are* VLIW, so maybe they need all that cache. https://www.crowdsupply.com/sra-centr8/icepeakitx-elbrus-8cb [crowdsupply.com] the newer version takes DDR4 it seems.
(Score: 5, Insightful) by Rosco P. Coltrane on Monday December 27 2021, @07:41PM (5 children)
All they need is comrade-in-chief Putin to write use-Russia-made-only into law and everybody will HAVE to use them processors.
(Score: 0) by Anonymous Coward on Monday December 27 2021, @08:30PM (1 child)
they are not there yet, but at least taking some first steps.
the trolling part refers to the pentium 4 prescott, and the other unrelated shortcuts intel took to get speed, that eventually lead to the various famous security bugs...
i wish EU had similar desires, as komrad putin, and had embraced the ARM or some other open arch "back in the days".
opensparc had entire complex cpu, if memory serves, more then a decade ago...
-zug
(Score: 3, Insightful) by fraxinus-tree on Monday December 27 2021, @10:25PM
EU doesn't have the "military above all, war imminent" attitude to everything like Russia has. EU is more like "trade with everyone, the politics will follow the money and the military is only to keep Russia calm". This is why EU is OK about CPUs being made in Taiwan as long as all the photolitography equipment is made in the Netherlands.
(Score: 4, Insightful) by fraxinus-tree on Monday December 27 2021, @10:32PM (2 children)
Putin pretty much would, if he could. Looks like they not only don't have the capacity, but they don't have the prospect of capacity to use them exclusively in any (however small) part of the government. The only purpose of these CPUs is the illusion of self-sufficiency.
(Score: 1, Informative) by Anonymous Coward on Tuesday December 28 2021, @01:36PM (1 child)
There is classic Russian scheme of doing things, imitate work, throw a show for general public and take money to connected officials pockets. I've seen this with GLONASS - Russians had head start, somehow manufactured few receiver modules and made it mandatory that all navigation devices sold in Russia must be compatible with GLONASS, and in few months time Swiss and Taiwanese manufacturers delivered their own GLONASS receivers that were order or two magnitudes better in power consumption. Well maybe those Russian ones could withstand nuclear strike and associated EMP, but for general consumer it is hard to sell a phone which lasts two hours on a charge:) And they will still use western semiconductors and in case of embargo will smuggle it, but I don't believe that there will viable semiconductor industry in Russia. In soviet times they copied western designs (usually poorly) and even for military purposes their quality assurance was so bad, that they turned to cherry picking working chips (with rates like 1 good chip in 1000 batch).
(Score: 0) by Anonymous Coward on Tuesday December 28 2021, @08:16PM
> There is classic
Russiancorrupt scheme of doing things...There us nothing Russian about it. If our corporate overlords could still do it here, they would. And they're trying.
(Score: 1, Interesting) by Anonymous Coward on Tuesday December 28 2021, @05:31PM
yes, yes, but we want to know if it can do math correctly? i mean a bank would emphasize that the calculations are correct before anything else?