Russian Electronics: Semiconductor and Processors #2

Do not expect Russia to manufacture anything of significance with 65nm. From what I remember Mikron has a prototype process. But they figured out because of surface defects on 200mm wafers not designed for doing anything over 90nm plus other things, the defect rate on the chips manufactured at 65nm with 200mm KrF tools is kind of high, so at most it will be useful to make highly regular chips like memory chips. I think the only thing that Mikron did manufacture at 65nm was an SRAM.

Mikron has significant capacity at 180nm and 90nm. Both older fabs with 150mm wafers which can do 180nm, and the new fab with 200mm wafers which can do 90nm.

There is also Nm-Tech, it has tools which should be able to do 90nm on 200mm wafers but last I heard they were still stuck manufacturing at 250nm.

Syntacore seems to be working on a new RISC-V core design called the SCR9.
This is a Russian company owned by Yadro which designs RISC-V soft-cores that they sell as RTL to chip designers and manufacturers.
This is kind of similar to how ARM (the company) operates when they design and sell ARM core designs.

SCR9 is a 64-bit quad-issue out-of-order processor with vector instructions. The processor cluster can have up to 16 cores and up to 16 MB of L3 cache.
This processor should be twice more powerful in terms of integer performance than their older SCR7 processor which was dual-issue out-of-order.
It also has vector instructions. The vector unit is capable of operations with 8-bit to 64-bit values with a vector unit up to 512-bit wide (!)
The vector unit can do two vector operations per cycle. This is quite powerful. Would make it one of the most powerful RISC-V core designs around.





This slide is showing cases of clients using their prior SCR7 dual-issue out-of-order core design.


Notice the 56 core (!) SCR7 processor design at 7nm which they claim taped out in 2021 for an EMEA client. That is pretty significant.

They also claim an EMEA client taped out a quad-core SCR7 processor at 12nm just last year.

The 200 mm wafer defect story smells. Russia is one of the few countries on the planet that can manufacture high purity and consistency
crystals on the planet. This includes industrial diamonds. The notion that it couldn't produce high grade 200 mm silicon wafers is BS.

Every story about Russia is tendentious and tainted with caricatures predicated on the mass hate propaganda launched at it from the
west. The west would not waste so much effort if Russia was the banana republic it claims. In this 200 mm story the subtext is the
"decrepit" nature of Russian manufacturing where everything is inferior. So Russia is supposedly making 200 mm wafers with backyard
ovens and hacksaws.

The 28 nm lithography effort in Russia is real. But Russia is not trumpeting it. Sort of like Russia did not trumpet its economic potential
over the last 20 years and successfully fooled the west. Even China kept strategically quiet about its lithographic capability giving us
the Kirin surprise.

Any talk about 90 nm and 65 nm is based on obsolete information. The 65 nm Mikron manufacturing story is a decade old. The 28 nm
EUV lithography breakout will be probably like with China. Once the 28 nm capability is available the 14 nm capability is a dial setting away
and all the effort is in the chip physical architecture (SOI, copper, FinFet, etc.).

kvs wrote:The 200 mm wafer defect story smells.   Russia is one of the few countries on the planet that can manufacture high purity and consistency
crystals on the planet.  This includes industrial diamonds.   The notion that it couldn't produce high grade 200 mm silicon wafers is BS.  
...
Any talk about 90 nm and 65 nm is based on obsolete information.   The 65 nm Mikron manufacturing story is a decade old.   The 28 nm
EUV lithography breakout will be probably like with China.   Once the 28 nm capability is available the 14 nm capability is a dial setting away
and all the effort is in the chip physical architecture (SOI, copper, FinFet, etc.).

According to several Russian news stories, in Russia they only produced 150 mm wafer substrates from silicon crystals. The 200 mm wafer substrates were imported. Right now there is a program to make 200 mm wafer substrates to stop relying on imports. As for why they did not do this sooner, it was probably because only a couple companies like Mikron and Nm-Tech used those wafers.

No one really does 65 nm with 200mm wafers other than Mikron. The rest of the world switched to 300 mm wafers around the time that processes smaller than 90 nm came out. So the 200mm wafers available in the international market aren't defect free enough to manufacture under 90 nm reliably.

Like I posted here before they are building a small 28 nm test fab with 300 mm wafers right now just north of the Nm-Tech fab in Zelenograd. Who knows when it will be operational. But I would not be surprised if they got it to work somehow. They might get the tools for it via the black market just to ensure 28 nm production before the Russian lithography machine is available.

https://www.eetimes.com/russias-angstrem-modernizes-wafer-fab/

Angstrem has a 200 mm wafer production line from around 2010.

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

https://mvc-nn.ru/en/wafers

ICSM makes 200 mm SOI wafers.

"Small batch" does no mean inferiority, unlike western wankers love to peddle. The product still has to be made. Scaling production is another
problem mostly about money.

There is a surge of 200 mm wafer plants underway in the global space because of industrial demand. The value of 300 mm wafers is
volume of chip production and not chip resolution. To get higher yields the defect rate in the wafers needs to be no worse than the
200 mm wafers per unit area. This requires higher manufacturing tolerances since defect rates are proportional to surface area.

https://en.wikipedia.org/wiki/Wafer_(electronics)

https://www.hindawi.com/journals/amse/2022/1829792/

kvs wrote:https://www.eetimes.com/russias-angstrem-modernizes-wafer-fab/
Angstrem has a 200 mm wafer production line from around 2010.

That is a foundry i.e. a fab where they do lithography and etch wafers to make ICs. They "print" chips on blank wafers. That EETimes article is talking about what became the Angstrem-T fab and is Nm-Tech today. They neither make silicon crystal nor cut it into blank wafers.

kvs wrote:ICSM makes 200 mm SOI wafers.

These SOI wafers are made by processing regular imported 200mm Si wafers with a special process and then grinding them down. You still need to get the Si wafers somehow. i.e. grow the silicon crystal and cut it into wafers. From what I get from the articles in the Russian press, right now Russia still does not grow its own 200mm silicon crystals. But only 150mm crystals with less diameter. Anyway, I think this "problem" is kind of overblown. China has massive capacity to make 200mm and even 300mm wafers with several Chinese companies making them. Blank wafer supply from Europe was cut, and replaced with Chinese suppliers. Big whoop.

kvs wrote:There is a surge of 200 mm wafer plants underway in the global space because of industrial demand. The value of 300 mm wafers is
volume of chip production and not chip resolution. To get higher yields the defect rate in the wafers needs to be no worse than the
200 mm wafers per unit area. This requires higher manufacturing tolerances since defect rates are proportional to surface area.

It is not that simple. Higher resolution tools are typically more expensive. The cost of the tools and fabs grows exponentially with each generation with a power law known as Rock's law.
https://en.wikipedia.org/wiki/Moore%27s_second_law

To keep the cost of individual chips down, they increase the size of the wafers every several generations so that each tool can process more area simultaneously. The industry has already gone from 1-inch (25mm), to 2-inch, 3-inch, 4-inch, 5-inch, 6-inch, 8-inch (200mm), and now it is at 12 inch (300mm) diameter.

The ArF, ArFi, and EUV lithography tools are made to process 300mm wafers from the get go. And modern KrF and i-line tools can process either 300mm or 200mm wafers. If you can process 300mm wafers, it does not make economic sense to use the smaller wafers. Because you need to amortize the cost of the tools by producing chips. The more chips you produce the better. A 300mm wafer has 2.25x more area than a 200mm wafer. So you can produce 2.25x more chips in the same amount of time with one. Making them 2.25x cheaper. The cost of the blank wafer and chemicals is kind of marginalized by the cost of buying the tools.

The smaller wafers aren't made to the same tolerances since there is no demand for working at high resolution with small wafers. So no one makes them with those tolerances. So let's say they are neither cut, grinded, nor polished to the same specifications.

Rodion_Romanovic wrote:So apparently there is also this plant in Moscow, Crocus nano, which produces 300 mm wafers

Yes. This was founded by Rusnano initially to make MRAM. But that fab can only print 2 metalization layers in a special process specific to MRAM. Then the wafers are sent to an independent fab in China to print the rest of the layers. This Crocus fab has too few tools to run a full logic process. For reference a 65nm logic process has like 11 metal layers.

It's quite funny, however, that the problems of Russian silicon products are discussed here first of all. At the same time, there is practically no news that Russia, for example, occupies 40% of the world market for artificial sapphires, which is also a necessary part of the electronics manufacturing process. This is not counting the market for noble gases of high purity.

We already talked about that here several pages ago. Sapphire crystal is used for specialty processes. It is used, among other things, to make GaN LEDs with GaN on Sapphire process. It is not typically used to make regular logic or memory chips. Those use silicon wafers.

There are alternatives to using GaN on Sapphire. There is GaN on Silicon and GaN on Silicon Carbide as well.

But yes, the fact that Russia can already manufacture 150mm silicon crystals, and 200mm sapphire crystals, shows they could manufacture 200mm silicon crystals if they wanted to. And they are working on it.

As for the noble gases, the main producers were Russia, Ukraine, and China. With the war and the sanctions Taiwanese TSMC and the West have been buying gases from China.

It is kind of pointless to be extremely worried about the wafer and chemical imports though. The West stopped selling them, so Russia switched to Chinese suppliers. There are plenty of those.

https://www.typhoon.su/


Some interesting news..

Unfortunately I dont know what most of the stuff in here is but it tells me that there is some serious work om semiconductors in Russia..

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

Yeah thats not helping..

Starting at 12:12 you can see Russian machine tools for cleaning 100-150mm wafers and test equipment for checking for defects on these kinds of wafers. The wafer cleaning is done with plasma.
These guys have developed a vacuum chamber for the machine tools and enclosures for wafers to prevent their contamination while handling them.
These machine tools are used by educational institutions and small scale laboratories in Russia.

Looking at their website they also either have or are developing plasma etching tools, and layer deposition equipment:
https://www.gn-tech.ru/en/produktsiya/

The photon wavelength is what governs the translucency outside of absorption windows for different molecules. Short wavelength light scatters
more off any air molecules than long wavelength light such as infra-red. That is why the sky is blue. High energy photons including gamma rays
can plow through the air and breaking up molecules in their path (and ionizing them) but all the interaction means that they are attenuated faster.

If air was just O2, N2 and Argon, then IR photons would go through it without loss (except at ultra high flux densities). But air has CO2 and H2O
which are strong IR absorbers without any large spectral gaps for this absorption.

https://nvlpubs.nist.gov/nistpubs/jres/40/jresv40n5p393_A1b.pdf

X-rays are just too high energy to be passively transmitted through air, so the attenuation is substantial. This is a good thing for life on Earth,
otherwise it would have been killed off by regular X-ray bursts from the Sun during coronal mass ejections in the direction of Earth.

All things considered some sort of longer wavelength laser would be the best option for attacking UAVs. There will be energy loss, but such loss
gets worse at shorter wavelengths. Teller's nuke-driven X-ray laser, that produces a vast amount of photons allowing even a small fraction that
makes it through the atmospheric column to do damage, is overkill that cannot be downsized for limited applications.

Plans for maskless X-ray lithography in Russia are not hot air. There has been enough primary research work to scope out the potential. The
knee-jerk skeptics are assuming Russia is starting from nothing and flying blind.

Of course, realizing research into industrial production is a whole problem onto itself. But we are not dealing with blubbering idiots and no physics is
preventing such machines from being built. Money and economic efficiency are the transient constraints.

More here:
https://smoothiex12.blogspot.com/2023/12/a-very-timely.html

No, the sky colour has nothing to do with absorption by the atmosphere. Blue and high frequency light is preferentially scattered
whereas the longer wavelength light is transmitted. You are not seeing the selective transmission of high frequency light, that
does not make any physical sense. Lower frequency light is absorbed less.

https://spaceplace.nasa.gov/blue-sky/en/

Aerosols and water vapour give a white-gray tint. When we had the large forest fires here in Kanada in the last few years the sky
was tinted white in spite of no clouds and a high pressure system. Aerosol particles and water vapour droplets are subject to
Mie scattering whereas the sky colour is due to Rayleigh scattering.

Anyway, it is good to see some information on the EUV/X-ray lithography activity in Russia. But the equipment posted by SmoothieX12
has been around for a while. We are lacking information about the efforts to industrialize this lab equipment.

Russia could have done this decades ago.. They would have a 14nm by now. Why does Russia wait to get sanctioned before initiating these projects? This project If I remember right started in 2014 and it just paused until what? 2021?