Not-exactly-new player in the x86 market

DrMrLordX

Lifer
Apr 27, 2000
22,705
12,658
136
It looks like the race towards domestic technology independence continues in Russia:

http://russia-insider.com/en/business/russias-mcst-unveils-homegrown-pc-microprocessor-chips/ri6603

If anyone was wondering who wants to throw billions of dollars into a hole to bring an x86 processor to market, now you know!

IN SOVIET RUSSIA, billions of dollars waste YOU!

Here's another fun bit: the 65nm Elbrus-4C is said to be about as fast as an Intel processor from about 5 years ago. Ha! ha ha! Queue up the AMD jokes gentlemen! I expect nothing less.
 

bononos

Diamond Member
Aug 21, 2011
3,928
186
106
Its not a waste, its necessary for Russia to build up its homegrown semicon industry to maintain from being overreliant on the US at the expense of its own economic and military security. Japs and Koreans do it, China is also following suit by putting more money into hw/sw development.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
When it comes to national security, you can no more blame the NSA for doing what it does than you can fault Russia for doing what it aims to do.

History shows no counter-example, if you neglect national security then it is simply a matter of time before you have no nation left to worry about.

Were the shoe on the other foot, Russia the base for the most common high performance CPU on the planet, the USA would find itself with no other acceptable option but to attempt to develop a home-grown solution of some form or another.
 

wolf_squad

Junior Member
Dec 2, 2014
22
0
0
five years behind is an overly optimistic estimate. from the few performance tests they've shown, these things are more like 10 years behind.

I also don't think this will impact the consumer market in any way.
 

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
Here's another fun bit: the 65nm Elbrus-4C is said to be about as fast as an Intel processor from about 5 years ago. Ha! ha ha! Queue up the AMD jokes gentlemen! I expect nothing less.

At least the Russian government isn't saying it will ever make money with it...
 

crashtech

Lifer
Jan 4, 2013
10,682
2,280
146
It's too bad they are so far behind. It'll be interesting to see what they do with their tech over time.
 

MrTeal

Diamond Member
Dec 7, 2003
3,916
2,700
136
five years behind is an overly optimistic estimate. from the few performance tests they've shown, these things are more like 10 years behind.

I also don't think this will impact the consumer market in any way.

Got a link to the performance tests? The i3 moniker wasn't even introduced until Clarkdale in 2010 so the timeline of i5/i3 level of performance being 5 years behind seems right, but I have a hard time believing they're comparable to a 32nm Intel chip on the 65nm node in anything but well threaded code where four real cores beat the two in CD.
 

Fox5

Diamond Member
Jan 31, 2005
5,957
7
81
It looks like the race towards domestic technology independence continues in Russia:

http://russia-insider.com/en/business/russias-mcst-unveils-homegrown-pc-microprocessor-chips/ri6603

If anyone was wondering who wants to throw billions of dollars into a hole to bring an x86 processor to market, now you know!

IN SOVIET RUSSIA, billions of dollars waste YOU!

Here's another fun bit: the 65nm Elbrus-4C is said to be about as fast as an Intel processor from about 5 years ago. Ha! ha ha! Queue up the AMD jokes gentlemen! I expect nothing less.

Probably only as fast as an Intel processor from 5 years ago under the most favorable conditions. At 800mhz, even as a quad core, you're looking at performance more on par with 10-15 years ago.
 

DrMrLordX

Lifer
Apr 27, 2000
22,705
12,658
136
At least the Russian government isn't saying it will ever make money with it...

Because they don't have to say it, they aren't saying it. It's priced so high that the "I don't want to buy stuff backdoored by the NSA" crowd will provide MCST with a bunch of customers right off the bat. All it had to do was be "good enough" for that crowd, which is a feat that took them awhile.

For a second, I thought Cyrix got reincarnated :p

In Russia, Cyrix would become Cyrillix. Da, tovarich.

Probably only as fast as an Intel processor from 5 years ago under the most favorable conditions. At 800mhz, even as a quad core, you're looking at performance more on par with 10-15 years ago.

The uarch is so different from what we use in desktop PCs that speed estimates are sketchy at best. It's a VLIW architecture, no? It's performance will depend heavily on compiler/software support. Running custom-tailored benchmark software, it is said to pull off something like 50 GFlops, or so I have heard, but what you're really looking at is a delayed chip that was once said to be a competitor to the Itanium. This chip is not really something you'd want in a desktop. It can do Transmeta-style instruction translation to handle x86, and it's probably very slow while doing it.

They also have the elrubus-8s, an octacore fabbed at 28nm

Really? Is that in production?
 
Last edited:

imaheadcase

Diamond Member
May 9, 2005
3,850
7
76
I guess you guys missed the part that these chips don't have to rival current CPUs for the uses they are intended for.
 

cytg111

Lifer
Mar 17, 2008
25,665
15,166
136
It's a VLIW architecture, no? It's performance will depend heavily on compiler/software support. Running custom-tailored benchmark software, it is said to pull off something like 50 GFlops, or so I have heard, but what you're really looking at is a delayed chip that was once said to be a competitor to the Itanium.

- Big things have small beginnings. Heres to hoping! (and hoping it wont happen during putin and a new cold-war).
 

DrMrLordX

Lifer
Apr 27, 2000
22,705
12,658
136
- Big things have small beginnings. Heres to hoping! (and hoping it wont happen during putin and a new cold-war).

Perhaps MCST can do something interesting with VLIW eventually. It did not work out very well for Intel. I would not mind having one of those octo-core chips in a 1P configuration as a curiosity.

Its not a waste, its necessary for Russia to build up its homegrown semicon industry to maintain from being overreliant on the US at the expense of its own economic and military security. Japs and Koreans do it, China is also following suit by putting more money into hw/sw development.

Sorry to have taken so long to address what you said here. Bear in mind that the Russians gain two advantages from using domestically-produced hardware: not only are they better-protected from NSA backdoors and similar hardware-level hyjinx, but they also have the ability to incorporate their own backdoors for use by Russian intelligence agencies. The downside for the Russians is that they are currently reliant on TSMC for production, which greatly increases the chance that foreign intelligence agencies might meddle in production somehow.

Framing the expenditures involved in producing such hardware as a "waste" was more-or-less a parody of posts in other threads on this forum where others have claimed that nobody in their right mind would ever want to pour money into x86 research/production.
 

ClockHound

Golden Member
Nov 27, 2007
1,111
219
106
History shows no counter-example, if you neglect national security then it is simply a matter of time before you have no nation left to worry about.

Were the shoe on the other foot, Russia the base for the most common high performance CPU on the planet, the USA would find itself with no other acceptable option but to attempt to develop a home-grown solution of some form or another.

So the Vodka & Caviar cold war is already lost?
 

TheELF

Diamond Member
Dec 22, 2012
4,027
753
126
Got a link to the performance tests? The i3 moniker wasn't even introduced until Clarkdale in 2010 so the timeline of i5/i3 level of performance being 5 years behind seems right, but I have a hard time believing they're comparable to a 32nm Intel chip on the 65nm node in anything but well threaded code where four real cores beat the two in CD.
Youtube has some clips,one is of the Elbrus-4C running Doom 3 BFG
https://www.youtube.com/watch?v=PUNJ_tkq2hk

( ... at ~20FPS)

But hey come on, it's a great effort on their part and as was said before these cpus don't have to reach modern speeds they just have to be able to do the job.
 

Leyawiin

Diamond Member
Nov 11, 2008
3,204
52
91
Funniest part of the article are the comments from the Russophiles at the end of the article.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,809
1,289
136
This is still going on. Especially, with Rosatom getting a 100% stake of MCST in 2023.

2021+ generation
Elbrus-2C3 = 16nm FinFET, 2-core
<-- same generation of dynamic binary translation x86-64 cores -->
Elbrus-16C/12C = 16nm FinFET, 16-core/12-core
Elbrusv6 "additions":
elbrusv6.png

The next parts will be from 2021 document for the 2025+ generation.
Elbrus-32C (2x32-core)
Elbrus-7L (16-core)
Elbrus-7M (4-core)
Elbrusv7 "additions" not fully out: AI/Crypto extensions? Plus whatever this is:
elbrusfuture.png

Rather than going from 16nm FinFET TSMC to 6nm FinFET TSMC as planned. Russian academics are recommending going from 16nm FinFET TSMC to 12nm FDSOI GlobalFoundries for the 2025+ generation.

If the geopolitical event continues on: Domestic 300mm and sub-28nm by Mikron RU is expected to be ready by 2028.
[December 10th, 2012 -> Mikron will develop 65-45-32nm processes at the Molecular Electronics Research Institute (MERI).]
It is <<Molecular Electronics Research Institute>> that is promoting the usage of FDSOI for MCST and Mikron's STMicroelectronic RISC-V MCU/MPU clones. Where they have cloned 28FDS already at Academy. Which is probably a leftover from...
"MERI carries out joint projects with global science centers of microelectronics:
CEA-Leti (Grenoble, France)
STMicroelectronics (Geneva, Switzerland)"
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,809
1,289
136
The issue is that won't be possible unless they uses SMIC for it.
SMSC (Semiconductor Manufacturing South China Corp) is the leading edge component of SMIC. The one doing FDSOI in China with NSIG (Okmetic, ZingSemi, Simgui) is Huali (HLMC).

Where Phase1 (22nm FDSOI) has been delayed to 2H26.
Phase2 (12nm FDSOI) would be two-three years after that 2028/2029.
Phase3 (10nm and 7nm FDSOI) will be two years after phase2 being 2030/2031.
Wait time between GF with lapsing sanctions and Huali with sanctions is about the same.

SMSC N+3 (6nm) is already out.

However, the cores/products are absolutely massive on FinFETs for reliability reasons:
2C3 = 222 mm2 (Dual-core being larger than the PowerVR GX6650 iGPU (192 ALU))
16S = 618 mm2 (16-core with reduced amount of cache than planned)

2C3... look at those cores... yeesh!
elbrus2c3.png
 

DZero

Golden Member
Jun 20, 2024
1,289
467
96
SMSC (Semiconductor Manufacturing South China Corp) is the leading edge component of SMIC. The one doing FDSOI in China with NSIG (Okmetic, ZingSemi, Simgui) is Huali (HLMC).

Where Phase1 (22nm FDSOI) has been delayed to 2H26.
Phase2 (12nm FDSOI) would be two-three years after that 2028/2029.
Phase3 (10nm and 7nm FDSOI) will be two years after phase2 being 2030/2031.
Wait time between GF with lapsing sanctions and Huali with sanctions is about the same.

SMSC N+3 (6nm) is already out.

However, the cores/products are absolutely massive on FinFETs for reliability reasons:
2C3 = 222 mm2 (Dual-core being larger than the PowerVR GX6650 iGPU (192 ALU))
16S = 618 mm2 (16-core with reduced amount of cache than planned)

2C3... look at those cores... yeesh!
View attachment 111868
damn... is so big like SD 810!