Russian Electronics: Semiconductor and Processors

kvs wrote:No. You are wrong.  The Intel and AMD x64 is wired at the CPU level for 64 bit floating point.   The Elbrus is clearly using
32 bit floating point and cannot do 64 bit natively.  It can do 24 floating point operations in a single cycle for 32 bits
and 12 for 64 bits.   I am quoting MCST itself so you can send them an letter telling them that they do not know what they
are talking about.   Originally it was claimed that the Elbrus would do 24 floating point operations per cycle for double
precision (64 bits).  

Your are also engaging in obfuscation with the vector MMX unit.   The Elbrus does not even have an MMX type multiple FPU
capability.   What relevance does 4 x 32 have for the Elbrus?   The problem is that it cannot do 1 x 64 in a single effective cycle
and requires two, thus the exact factor of two GFLOPs reduction.  

https://wiki.tuflow.com/index.php?title=Hardware_Benchmarking_Topic_Single_Precision_VS_Double_Precision

The primary difference between the SP and DP numbers in the above tests is due to main memory hits.  It costs more
to push 64 bit words between RAM and the caches than 32 bit words.   There are also hits from FDIV and FMUL total
cycle differences between single and double precision.  

The Elbrus can do 12 off 64bit multiplication in one cycle, or 24 off 32bit multiplication.

If you divide the 125 gflops with the frequency, number of cores then you get the wide of the  ELBRUS core FPU vector registers.

It is 768 bite wide (strange number)


On the AVX-512 the vector registers are 512 bit width , and capale to hold 8 off 64bit float.


4 float important because of you do 3d transformations (rotation, move, projective transformation, mirro, eigenvaule calcualtion or whatever) you need 4 dimensional vecotrs.


The 4 float/register is the most common representation, and it is like this since the AMD 3DNOW isntruction set (float32 expansion of the original integer MMX registers).

The intel version is the SSE, introduced with the Pentium III CPUs.

If you have speed issues with your program try to use the SSE registers ( 20+ years old).

There are C header files that can utilise this registers.

The Elbrus VLIW CPU, means the normal, x86 style FPU interpretation doesn't work.

If we consider thath the LEBRUS was developed specificly for vector computing then it would not be suprise if it can pus extreme ammount of float multiplication in high precision.

It is faster than a similar clocked intel CPU.

A processor for smartphones has been developed in Russia, but it is impossible to produce it in Russia
Sergey Karasev
March 4, 2021

The Russian Research and Production Center "Electronic Computing and Information Systems" (SPC "ELVIS") has developed several new processors designed for devices of different classes. However, it will be difficult to organize the production of chips on the territory of the Russian Federation.

The named enterprise was formed in 1990 on the basis of a structural subdivision of the research and production association "ELAS". SPC "ELVIS" develops microcircuits of the "system on a chip" type on the basis of its own design platform "Multicore".

According to the Vedomosti newspaper, ELVIS will announce three microprocessors in the near future. This is, in particular, a mobile chip for tablets and smartphones with a "trusted environment". Another product will focus on artificial intelligence systems. Finally, the third ultra-low power processor is designed with the Internet of Things in mind.

The mobile solution will make it possible to create devices based on a domestic component base. Probably, such smartphones and tablets will be targeted primarily at government customers.

However, there is a difficulty: there are no suitable factories in Russia for the release of new products. "Despite the fact that the ELVIS processor will be considered a Russian development, it will still have to be produced at the factories of Taiwanese TSMC," the publication says.

One way or another, by 2022 ELVIS expects to supply several hundred thousand processors to the market.

https://yandex.ru/turbo/3dnews.ru/s/1034084/v-rossii-razrabotali-protsessor-dlya-smartfonov-no-proizvodit-ego-v-rf-nevozmogno?publisher_logo_url=https%3A%2F%2Favatars.mds.yandex.net%2Fget-turbo%2F985264%2F2a000001651b42c8bec23a1fbba11e68c269%2Forig&promo=navbar&utm_referrer=https%3A%2F%2Fzen.yandex.com

And he factory in Taiwan fell from the sky? Build your one!

Good product -> Marketshare -> Factory

Those foundries are DAMN expensive!

I don't think setting up a factory/ production line for semiconductors is as easy as it sounds. China, which has a bigger manufacturing base and experience in semiconductors is struggling to produce 14 nm chips for its electronic products after US sanctions. From what I've read from Chinese forums, China has been barred from purchasing fabrication equipment and at the same time don't have a pool of experts who have experience working on 14 nm process and beyond. I think the latter is an aspect of the semiconductor industry that you can't simply bypass by just throwing in investments

And the equipment and the people that can do this shit came from the Moon or what? No, they build the equipment and developed it further and further. Like their experience in doing it. You have to start somewhere or you will never make it.

If the purpose of these chips is:

The mobile solution will make it possible to create devices based on a domestic component base.

Then how about completing the domestic component base.

When Ukraine stops supplying helicopter engines then Russia spent the money and improved the newest designs the Ukrainians were working on and made them better because they used all the newest design and manufacturing techniques and equipment... and they should do the same with everything.

If they aim for high quality best they can manage then that will mean the products should be good enough to sell world wide, which could create demand for Russian electronics and they could get back on that horse.

They could end up selling technology to China at cheaper prices than the west is offering.

The situation with Safran is key... the French asked Russia to produce their high tech expensive titanium components for their own engines they normally buy from US companies... this (making their own chips) is something Russia should get involved with with their skills I am sure they could improve progress in the field.

PhSt wrote:I don't think setting up a factory/ production line for semiconductors is as easy as it sounds. China, which has a bigger manufacturing base and experience in semiconductors is struggling to produce 14 nm chips for its electronic products after US sanctions. From what I've read from Chinese forums, China has been barred from purchasing fabrication equipment and at the same time don't have a pool of experts who have experience working on 14 nm process and beyond. I think the latter is an aspect of the semiconductor industry that you can't simply bypass by just throwing in investments

Isn't China the pioneer in semiconductor IC fabrication? They are often referred to as the electronic components factory of the world. Why would they struggle in this area?

Isn't China the pioneer in semiconductor IC fabrication? They are often referred to as the electronic components factory of the world. Why would they struggle in this area?

I dont think so, China still seems heavily reliant on imports of high-end chips for its companies like Huawei.

Bleak but salvageable’: Huawei has limited options as U.S. sanctions cut off supply to smartphone chips

Huawei is reportedly running out of its own high-end chips for smartphones due to U.S. sanctions — and the company may have limited options to secure future supply.

The inability to obtain cutting-edge chips will threaten Huawei’s newly-acquired status as number one smartphone maker in the world. It could also wipe out billions of dollars of sales for the Chinese technology giant.

https://www.cnbc.com/2020/08/12/huawei-options-as-us-sanctions-cut-its-supply-of-smartphone-chips.html

Huawei even contemplated getting rid of its high end smartphone products due to concerns that they will run out of chips. So if China is truly a semiconductor "Powerhouse" then why are they in this situation? The current leaders in this field atm is clearly Taiwan and maybe South Korea.

China understands that independence in the semiconductor industry is essential for its future economic development. Unlike, before, most consumer products nowadays involve some degree of technology and that makes these chips very important, from cars, to household appliances, phones, etc.

Taiwan prosecutes semiconductor recruiters accused of illegally poaching talent for Chinese company

Taiwanese authorities have raided the offices of two companies that allegedly broke the law in poaching local chip talent for a mainland Chinese artificial intelligence (AI) chip maker, accusing the headhunters of undermining the local semiconductor industry.

The raid hit offices of WiseCore Technology and IC Link, recruitment companies that have headhunted hundreds of chip experts over the past three years through a joint venture with a mainland Chinese chip maker, the prosecutors’ office in New Taipei district said on Tuesday.

https://www.scmp.com/tech/tech-war/article/3124845/taiwan-prosecutes-semiconductor-recruiters-accused-illegally-poaching

I really don't like the idea of Russia relying on TMSC for production, it's obviously a high tech field and not an easy one if china is struggling with it, honestly it feels like Russia needs another 10 years to close the gap .

I'm curious, do people with no experience and knowledge in tech always say things like "needs x amount of time..."? Why? There is no such thing as time frame for anything in development.

Russia replaced most tech that took years to make, in a matter of a couple of years (microbolometers for thermal imagers for example). Their processor design is fine too and competitive.

So where does this "needs 10 years" come from and even remotely mean? It doesn't actually mean anything and isn't translatable to something coherent.

It's called Fabrication. And the equipment needed is lithography. That is what they need and China is starting to produce such equipment while before hand it was either France or Denmark who made such equipment.

kvs wrote:

Rasisuki Nebia wrote:I really don't like the idea of Russia relying on TMSC for production, it's obviously a high tech field and not an easy one if china is struggling with it, honestly it feels like Russia needs another 10 years to close the gap .

Russia is slowly deploying domestic fabrication.   There is now a transition to 28 nm lithography.   This good enough for the Elbrus.  
I posted on this another thread, the "7 nm" FinFET from TSMC is actually 10 nm and what Intel calls 14 nm is look nearly identical
when you dissect the latest AMD and Intel CPUs.   Talk of 5 nm lithography qualifies as a scam.    There is a breakdown of the
material continuum around 3 nm and you have cluster complexes.   So no transistor can be manufactured as current leakage is
at 100%.  

The 28 nm transition is rather old news, here is an article in Russian from 2018:

http://smart-park.ru/index.php/news/media/398-28nm.html

There is no particular need to push everything to 14 nm unless the objective is large number of cores in the same package
or consumer product economies of scale.   For military uses 65 nm is already good enough.  

https://www.niisi.ru/devel.htm

http://www.kpda.ru/upload/iblock/56c/serdin20190424.pdf

I'm no expert so I'm going to defer to your knowledge. Is there a reason Ru-MIC will never need lithographic fabrication below 65nm, and if it was hypothetically economically viable would military tech not have improved performance from smaller lithography (up to 14nm)?

I would say the most important part for Russia now is to ramp up the domestic market with some protection measures, that will create the business case for the natural grow of the local production. Once that develops, they can start exporting also and gradually come close to matching Taiwan and other countries over time. I also agree the obsession with reaching the absolute smaller node makes no real sense for Russia right now.

kvs wrote:

Rasisuki Nebia wrote:I really don't like the idea of Russia relying on TMSC for production, it's obviously a high tech field and not an easy one if china is struggling with it, honestly it feels like Russia needs another 10 years to close the gap .

Russia is slowly deploying domestic fabrication.   There is now a transition to 28 nm lithography.   This good enough for the Elbrus.  
I posted on this another thread, the "7 nm" FinFET from TSMC is actually 10 nm and what Intel calls 14 nm is look nearly identical
when you dissect the latest AMD and Intel CPUs.   Talk of 5 nm lithography qualifies as a scam.    There is a breakdown of the
material continuum around 3 nm and you have cluster complexes.   So no transistor can be manufactured as current leakage is
at 100%.  

The 28 nm transition is rather old news, here is an article in Russian from 2018:

http://smart-park.ru/index.php/news/media/398-28nm.html

There is no particular need to push everything to 14 nm unless the objective is large number of cores in the same package
or consumer product economies of scale.   For military uses 65 nm is already good enough.  

https://www.niisi.ru/devel.htm

http://www.kpda.ru/upload/iblock/56c/serdin20190424.pdf

I really appreciate the response, hope i didn't come off as pessimistic to everyone but knowing the US might pull it's strings on TSMC and it's all for nothing, while the talk about smaller nm in military sphere in Russia is satisfactory, the Elbrus is from what I've read compatible with x86 and x86/64 , I don't see why Russia couldn't sell them to countries that are not so friendly with the west. Though hopefully in the long run, they start competing on the international market.

kvs wrote:A point I should have added is that China is now transitioning to 10 nm class manufacturing.   That is bleeding edge and actually puts
Taiwan in a bad place.   Thanks to the US orchestrated tension, China will move its production away from Taiwan onto its own soil.
I have seen China dissed as being not so good at IC manufacturing.  I think that is ignorant analysis.   Don't confuse the trashy
and dangerous consumer goods from China with ICs.   You can't hack ICs and all the consumer trash is not made by some giant
company that also makes ICs.  China is a big and complicated economy.   It has issues but it can't be dismissed as garbage.  

I can see similar facile commentary hurled at Russia for being "backward".

So if America pulls the sanctions card in regards to TSMC, it will face the import substitution response.   Its sanctions will just
deprive TSMC of customers but will not stop China and Russia from producing high resolution ICs.

Whenever the US places sanctions on a powerful country that can fight, I think it is a win-win situation for the powerful country that has sanctions placed on it.

If it were not for sanctions, Russia would not have been able to make, what it is making right now.

https://sdelanounas.ru/blogs/140367/

A new computer production facility based on Russian Elbrus processors has been launched in the Moscow region

65nm on 200mm wafers is enough for current Russian military requirements. But if they want to produce products to replace government civilian chips on servers and routers then it is not enough.

lancelot wrote:65nm on 200mm wafers is enough for current Russian military requirements. But if they want to produce products to replace government civilian chips on servers and routers then it is not enough.

Except Elbrus 8SM and newer (Elbrus 1SM, Etc) are 28nm and Elbrus 16 is also 14nm or so. But all TSMC built. Military ones are built locally but on 65nm.