China chip an Intel rival?
China's Godson-3 chip is ambitious if anything. It proposes to be everything a world-class processor should be--and then some.
Developed by the Chinese Academy of Sciences, it also has a larger goal: microprocessor independence for China. "Their motivation is pretty clear. They don't want to be totally dependent on the outside world for something as important as microprocessors," said Tom Halfhill, an analyst at In-Stat.
The current Godson chip: the Loongson-2E High-Performance General-Purpose CPU
(Credit: The Institute of Computing Technology of the Chinese Academy of Sciences )But its singular head-turning feature is the proposed Intel "x86" compatibility mode.
"The most interesting part of the chip is that they're adding about 200 new instructions to assist with x86 compatibility," Halfhill said.
At its core, it is a MIPS RISC processor--but one that proposes to run Intel-compatible software efficiently enough that most Chinese may not notice the difference. "It won't be an x86 processor. But the 200 instructions will optimize the (Intel) performance," Halfhill said.
Halfhill says it's not an outright Intel competitor. "I don't think they're doing this to compete with (Intel) x86 per se (but) If somebody has to run some software that's only available to x86 you can do it," Halfhill said.
The upshot is that it will compete on some level. "Every processor that China sells is one less processor that somebody else sells," he said.
Oh, and that sticky licensing question. "They don't have an Intel license. It remains to be seen if these 200 instructions they're adding violates any Intel patents or not. They have not disclosed those new instructions," according to Halfhill. (Intel would not comment on the processor.)
But the support from the Chinese government may help it avoid the fate of other x86-compatible processors that had similar aspirations, such as the Cyrix and Transmeta Crusoe processors. To be sure, most Intel-compatible processors have died a slow, painful death.
If the processor succeeds, it will be compelling because of the overall low cost of the platform. The Chinese government can put together a hardware-software package "that's a lot less expensive than if they went with buying chips from Intel and licensing from Microsoft," Halfhill said.
"They want to get a lot of computers out into their schools and companies and they want to make them in China," he said. "With a MIPS-compatible processor, they can stick Linux on it, they can stick Open Office on it, and adapt open-source Web browsers."
The silicon is no slouch. The Godson-3, due in 2009, will have four cores, and an eight-core version is also planned. "The four-core chip is the most sophisticated chip we've seen come out of China," according to Halfhill.
The later eight-core version is a "heterogeneous" processor, he said. That is, different types of processors will be put on one piece of silicon. Typically, in a quad-core Intel processor, for example, all four cores are identical.
This would put the eight-core version in the same elite camp as IBM's cell processor or Intel's future Havendale and Auburndale processors. Those Intel chips due next year put a graphics core and a Nehalem processor on the same piece of silicon--a first for Intel.
And Godson may go far beyond desktops. By 2010, China plans to build a petaflop high-performance computer based on the Godson-3. (A petaflop translates to a quadrillion floating-point operations per second.) IBM currently offers one of the fastest supercomputers in the world that achieved 1 petaflop in June.
The original Godson-1 was launched in 2002. In 2005, the Godson-2 debuted, with new versions appearing successively with increased performance.
Godson processors, also known as Loongson, are manufactured by STMicroelectronics.
Brooke Crothers has served as an editor at large at CNET News, an editor at Dow Jones' Asian Wall Street Journal Weekly, and a senior editor at InfoWorld. His CNET blog covers chip technology and computer systems, and how they define the computing experience. He also contributes to The New York Times' Bits and Technology sections. He is a member of the CNET Blog Network and is not an employee of CNET. Disclosure. Follow Brooke on Twitter @mbrookec. 
Good luck China, your social problems are just about to come....
By the by, China did build a Linux distro years ago (RedFlag - google it), based off of RedHat. Everyone using Linux in China pretty much uses something else.
Unless there is true innovation on behalf of the Chinese, the will only be producing good copies of yesterday's technology.
"gen" doesn't just mean counting cores - it means much more than that.
the fact that they are planning to compete with ibm in making the worlds fastest computer is not reverse engineering, no where in the article they say that they are reverse engineering. intel cannot even dream about competing with ibm for real pro grade mission critical computing.. intel should be honored that godson is making a processor that is compatible with their years behind technology.
commenting about politics and world domination isn't the way to go in a tech blog, its discourse, out of topic... well maybe world domination is not because it sounds cool but he is right about elections. you don't elect a president by listening them since they won't do anything they talk about anyways. you elect a president for the companies and people who back them up. there is a reason why MS moved all their r&d to india.
On the other hand, mandating a single solution and approach from on high is also doomed to failure. The trick for a country like China is to come somewhere in between - to avoid the waste inherent in the western system without the mismanagement of the government directed system. It will be interesting to see how it works out.
FWIW, the MIPS RISC designs turned out to go against the need for low instruction stream bandwidth and the high work/cycle needed for power-efficient operation. Grafting 200 new instructions onto a RISC design is bass-ackwards === RISC means REDUCED instruction set. computer. So it spoils the point. Yes, clock rate is high, but unless you're looking for a clothes-iron it's not too useful. Certainly no good in a laptop. Hence Intel's switch in processor architecture from P4 back to lower-clock rate Core designs with more work/cycle. So even if the Chinese engineers succeed in producing a working design, and there's no reason to suppose they won't eventually, a compromise in power consumption or performance is pretty much a given.
It's one thing for a politician to say "we're going to do this" and (in China) back it up with large resources. But that doesn't make the learning curve or laws of physics go away magically.
On the other hand, the Godson project is a national project by the Chinese government, if necessary the Chinese government can just mobilize all the Chinese scientists within China for this project.
See the drift ??
Intel will carry on as usual. So will AMD. Microsoft OTOH would end up losing huge in China (and large parts of Asia with it) if this thing ever gets any altitude.
and for it to survive, it has to sell cheaper than Intel.
my guess is that the new chip might appear in appliances and mobile devices.
nonetheless, some of the money will go to Intel under patent law.
and i think Intel got enough patents in the chip business that even if you sneeze you have to pay them.
What makes you think China needs to adhere to ANY western patent ?
Who's going to enforce it ?
I personally think it's a good thing that some of these patents get smashed..
I'll leave it up to the madding crowd to explain as to why.
if China wants to sell the equipment to the U.S., they have to comply with U.S. patent law.
it's just simple as that.
and China economy largely depends on western imports.
heck, ask any rich Chinese today.
they wouldn't buy any Chinese-made product themselves!!! :-)
About 10years(or older) ago, I red an article saying that a 4times faster processor was developed by a South Korean company for similar motivations described in above article. I thought WOW this is great! 4x faster processor? But that's it. No company's buying that in terms of incompatibility and lack of applications. Judging from this old article, several attempts to develop faster chip was successful in the past but giant Intel is yet here. Firefox is better than IE but minor in market share. Think about it.
by the way, I disagree with the firefox-vs-ie example. Firefox has majority share in some regions of the world. In those regions, the uptake has been amazingly fast and part of that is due to the moderate size of the population and educated population. The U.S market is extremely large so it'll take a little longer for firefox to claim king. Having said that, firefox already broke some 15-20 percent share. That's the biggest of any browser. So give it some more time and it'll grow even more.
Intel and AMD are private corporations, they are big but to certain extent they have limited number of R&D staff and workers working for them.
On the other hand, the Godson project is a national project by the Chinese government, if necessary the Chinese government can just mobilize all the Chinese scientists within China for this project.
See the drift ??
On the other hand, the Godson project is a national project by the Chinese government, if necessary the Chinese government can just mobilize all the Chinese scientists within China for this project.
See the drift ??
to certain aspects, both intel and amd are private corporations, but they do receive our gov on their R&D. further they use or apply tons of new findings from academic field to their current applications. and those researches are funded by our gov. china does not have the system similar to us, so they would do things differently. further collectively gathering scientists on one or two projects in US is not unheard of either: atom bomb and apollo projects.
Go ahead. Put all chinese scientists on this one project. That's like 10 million other more useful things they could have been doing. Yet they won't be able to change the Laws of Physics, or be able to catch up with Intel and AMD incredibly much faster by themselves. It's the same reason multi-threaded programming is hard, and government committees rarely accomplish much. Here too, the learning curve must be borne by individuals. And you can't run 10 successive generations of chip design through fab in any less time no matter how many people you have.
So having a lot of people work on something sounds good, and can produce results in the long term, but in the short term you wind up with simplistic least-common-denominator approaches, such as grafting a big band-aid onto the side of an someone else's antiquated design.
- by anakin2006 September 16, 2008 10:30 AM PDT
- contencreator-2008
- Like this Reply to this comment
-
(30 Comments)from zero to build up, even an aged design is better than nothing. the important thing is the first step. chinese achievement might not be matched in commercial field unless they laid the egg and relayed the result to private sector for them to copy and hatch more variants. then i would worry that intel/amd will get hammered commercially.
because of complicity of OS compatibility and user acceptance, chinese might not start it in desktop product directly. instead, they could build embedded and bigger server using linux. this will be the big threat to intel/amd.