日電腦打敗「天河一號」 日志

lo_pak

【明報專訊】「TOP500」全球超級電腦排行榜昨日公布，日本富士通（Fujitsu）製造的超級電腦「K」擊敗中國「天河一號」，成為目前全球最快的超級電腦，其運算能力比「天河一號」快足足3倍，達每秒8000萬億次，相當於100萬部普通個人桌面電腦。

「K」部署在神戶的RIKEN高級研究所，主要進行科學運算。

MoiRhapsody

You are using every chance To spin any piece of news into toes-licking CCP.

peter236

Nono , he was doing what you use to do.
MoiRhapsody 發表於 2011-6-24 17:32

nah, Mr Lik was using every chance to spin any piece of news into anti-Chinese material.

MoiRhapsody

Nono , he was doing what you use to do.

peter236

You start the bashing first, you moron
MoiRhapsody 發表於 2011-6-24 17:27

nah, it was Mr Lik who started the anti-Chinese bashing in post #2 of this thread in a sarcastic way.
We have already seen a lot of these self-hating posts here.

MoiRhapsody

You start the bashing first, you moron

peter236

I wouldnt say Loongson is not 'chineses' but in the other hands, bashing fujitsu SPARC fx by the ISA ...
MoiRhapsody 發表於 2011-6-24 16:12

But BiscottiGelato has already explained the differences.
Why are you so stubborn? For the purpose of basing the Chinese?

MoiRhapsody

I wouldnt say Loongson is not 'chineses' but in the other hands, bashing fujitsu SPARC fx by the ISA it is using is totally invalid.

rockypath

To completely develop something in-house just to claim that it's a all China/Japan/etc design makes no business sense what-so-ever.
BiscottiGelato 發表於 2011-6-24 14:51

I didn't read this thread and I don't know what had been arguing about in here.

However, I agree with what you said about the "in-house" developing.

I would say this would be the least and last option any company or country would take. They will "copy" or "improve" on a design or product whenever it is possible. Why do we need to "reinvent the wheel" when the wheels are already available to be taken.


China has been trying to develop a lot of technologies on its own because the western countries has been embargoed exporting technologies to China until NOW. I would deliver my respects to the Chinese scientists for their achievements.

BiscottiGelato

Moi, no need to shout all caps...

I see where peter is going.
Fujitsu bought HAL, including designs personel and expertise for a SPARC compatible core.
Longsoon designed their own MIPS compatible core.
That's the difference.

MIPS32/64/x86/SPARC, etc is just an architecture standard. It's just a spec on register space and instruction set. When we say we are designing, we are talking about RTLs, VHDL/Verilog codes. In this case, Fujitsu bought out HAL which had design/personel/expertise, and thus is far from starting from scratch. Longsoon started from scratch (or we'll presume unless proven otherwise).

If Longsoon is paying and buying a MIPS core (which MIPS also makes aside from just doing license for just the architecture), then it'd not be developing the core themselves in this case.

That said, even if a core is purchased, you still need to design a bunch of stuff around the core to call it a microprocessor. Memory bridges, various memory controllers, bus controllers, interrupt controllers, memory spaces, interblock communication channels, etc etc... Fujitsu probably got some of these off HAL also.

This doesn't mean that Fujitsu might not have their own ASIC design capability however. Just happened that their SPARC expertise is acquired instead of in-house developed. Which at times, acquisition is a much smarter move, both economically and technically than in-house design from scratch. To completely develop something in-house just to claim that it's a all China/Japan/etc design makes no business sense what-so-ever.

I work in the ASIC industry. Not one of the sharpest pencil in the box, but I think I can explain more if there's still confusion over design vs architecture.