Russian Electronics: Semiconductor and Processors

Honestly, there's little "real-world difference" between the two architectures; product design is largely driven by other factors. However, I'd say that availability and existing code base are often the deciding factors in choosing one architecture over another; it can be challenging to buy a sub-10k quantity of high-performance MIPS SoCs, whereas you can easily get an i.MX/OMAP/Sitara/Tegra2/etc. solution; if the necessary OS/software/drivers are only available for one architecture, that will be your architecture.

Architecture-wise, ARM has fewer registers but more addressing modes. MIPS has more registers, but fewer addressing modes. --> I suppose that MIPS would have a slight performance advantage when the memory bus is slow and you're in a position to take advantage of having more registers.

Can you identify what is the single biggest constraint anyone has when working on smartphones or tablets? Give yourself a minute to think before reading further..Assuming you got it right...its battery aka power...Since vendors are making the phone "smart" by adding many functions to it, consumers also want to do lot more , but still they don't want to drain the battery out...So any architecture should be power efficient ..As it turns out, ARM processors has the best power efficient processors or in another words they have advantage of in power performance ratio...Now add another dimension to this. You need an OS and android primarily is based on RISC based architectures and so OS portability is important..And next is marketing..ARM did an amazing job in marketing and selling its solutions..It targeted the right segment...MIPS used to have larger share in Settop box markets etc...

But now MIPS and Intel with ATOM ver2 processor are closing the gap with ARM ...so its matter of time when there will not be much technical advantage going ARM processors...There are also other business factors like pricing, availability and application support which also largely influence a vendors decision to go with a specific architecture...

sepheronx wrote: But in the end, a MIPS processor can easily do what arm one does, and vice versa with little differences.

Morpheus Eberhardt wrote:http://www.rostec.ru/news/4516582

sepheronx wrote:I dont see the point really. If they produce mips now, they can accomplish what arm does with mips. Would also save them money.

kvs wrote:

sepheronx wrote:I dont see the point really. If they produce mips now, they can accomplish what arm does with mips. Would also save them money.

It may be a code base issue.   There is a lot of software written for ARM.   Even though MIPS dates back to the 1990s, it never had the sort of
large scale adoption.   If all one cares about is compiling programs from source code then any decent CPU is enough.   But as I am sure you know,
that is not how this industry works.

mutantsushi wrote:Well if you want to make CPUs to go in commercial products like smartphones that are supposed to use
generally available Android apps from Google Play App Store, yes it's important.
Sure, China also has MIPS projects, but also pursues commercial ARM projects.
(Android in fact HAS a MIPS branch maintained by Google, and it may become commercially relevant.  Not yet though)
There is really no difference in software jobs in general, since the issue is recompilation,
a "hassle factor" to promote local ISA infrastructure is just not going to accomplish much.

Remember, America's "Silicion Valley" is all about commercial synergies after all.
Ultimately, swapping MIPS execution units for ARM designs does not impede Russia's larger Semi design/fab projects,
and if an ARM licence produces profitable commerical product, that helps things in the larger picture.

mutantsushi wrote:I'm saying it exists but you can't trivially download majority of Android apps, thus not commercially relevant at present moment.
(for consumer-focused software platform markets, as opposed to niches which don't involve consumer software market)
I actually think MIPS has interesting potential (aside from current niche), considering Android potential as well as other OS,
but that is obviously not the same as short term commercial viability...
I would say comparison to multiple competing Intel proprietary ISAs is not relevant, because this is not about pushing new ISA but about
cashing in on existing ARM market with instantly viable commercial product, which MIPS apparently is not at current moment (outside existing niche).
The fact they are choosing to go ahead with it while already having MIPS projects (and knowing all costs involved) suggests there is a good reason to do so.

The new non-Neumann ISA that basically obsoletes OOO Branch Predictors seems really interesting and potential long-term...

According to the website of the Russian developer of processors and microelectronics MCST, submitted in the spring of 2015 first Soviet mass-production rack server based on a 4-core Elbrus-4C is an integral part of the new computing building Supercomputers with performance of 13.8 teraflops! New Russian computer system is a rack-mounted cluster based on domestic servers, in which a total combined 64 module 4 processors each. All 256 processors 4-core Elbrus-4C (800 MHz per core) on domestic motherboards, where the slots of RAM installed dies a total volume of 6 TB, and the total amount of disk space of the entire complex is 32 TB. The performance of such computing systems Supercomputers equal to 13.8 teraflops. At the end of 2015 is expected to create the next generation of Russian computing systems based on 8-core processor Elbrus-8C (1.3 GHz per core). The developers also acknowledge the ongoing processor Elbrus-16C and a series of decisions based on it.