Russia's new home-grown CPUs

[Chuck(G)]

Interesting, yet--

They're not exactly wonderful

 

[commodorejohn]

Tragically, they no longer have DEC to swipe from.

That said, I'm all for less homogeneity in computers, even if it is only happening thanks to Putin's paranoid isolationist policies.

 

[eeguru]

Why can't they just buy Allwinners from China like the rest of the world when it needs something cheap and bad?

 

[krebizfan]

It would have made more sense if the manufacturing plant was in Russia. Overpriced and slow but at least the entire production chain would be kept in country. The references I see say the actual production is done by TSMC.

 

[k2x4b524[]

Wouldn't it be weird if they produced cpus, in-house, that smeared the floor with intel and amd?

 

[krebizfan]

Wouldn't it be weird if they produced cpus, in-house, that smeared the floor with intel and amd?

In one sense, they have. Elbrus is using a 90 nano-meter process so the die size must be enormous and the power consumption outrageous. It has the same transistor count as a quad core Phenom II built on 45 nm. Not having those two process shrinks would suggest quadrupling of both die size and power consumption. Working at all at 800 MHz is impressive.

 

[Stone]

And the prices are so reasonable.

 

[Chuck(G)]

And the prices are so reasonable.

Well, they'll be "gotta-haves" for the chip collectors.

 

[Stone]

Well, they'll be "gotta-haves" for the chip collectors.

Who collects these newly manufactured, half-baked, under powered garbage Russian chips?

Do they collect bridges , e.g., Golden Gate and Brooklyn, as well? If so, ...

 

[cthulhu]

I wonder how many of the x86 extensions it supports. MMX and SSE are pretty much essential. SSE2 would be very useful.

 

[KC9UDX]

Tragically, they no longer have DEC to swipe from.

That said, I'm all for less homogeneity in computers, even if it is only happening thanks to Putin's paranoid isolationist policies.

Hear, hear! This is good for everyone.

So the processor runs x86 under emulation: is it SPARC?

 

[krebizfan]

Not SPARC but Elbrus 2000 which is a VLIW processor closer to Itanium in concept.

 

[Chuck(G)]

Yup, 256x84-bit register file VLIW architecture.

Interesting CPU. Apparently, the firmware for the x86 translation is kept on a CF card:

 

[cthulhu]

Yup, 256x84-bit register file VLIW architecture.

Interesting CPU. Apparently, the firmware for the x86 translation is kept on a CF card

Strange. I wonder what the performance hit is. It was horrific for Itanium.

 

[ClassicHasClass]

It may not be as bad as you think: see, for example, Transmeta Crusoe and Efficeon. Clock for clock, they were slower, but they also used substantially less power at the time.

For another example of VLIW that didn't suck, see Apollo PRISM, which HP cancelled after purchasing Apollo but used some aspects of in PA-RISC.

 

[Scali]

Strange. I wonder what the performance hit is. It was horrific for Itanium.

Itanium's hardware x86 was very bad.
The hybrid software JIT that Microsoft put in Windows was a lot better. Eg, an 1.5 GHz Itanium would perform about as well as a 1.5 GHz Xeon.
Which is actually better than Crusoe I suppose.

I guess a lot of it would depend on just how good the x86 emulation software is.

 

[mark.ablov]

Yesterday appeared another one - Baikal T1, 2 x MIPS Warrior, 28nm.