Slashdot Mirror
China Develops Their Own CPU: The "Dragon Chip"
vaxzilla writes "China's People's Daily
Online is reporting in
this article that the Computer Institution of the Chinese Academy of
Science have developed a new CPU, which they're calling the Dragon Chip.
The report isn't clear on the technical details of the chip, though it
does state, somewhat confusingly, that it, `is based on the RISC
structure, a totally another standard. Therefore, it will not fall into
the intellectual property right trap.' They're running Linux on the chip
and have built a server around it, Soaring Dragon. It looks like China is
starting to tell both Microsoft and Intel to take a hike. Interesting
times are ahead."
This being not a for-profit fly-by-night sweatshop, but a research institute, rumour has it that they cloned Alpha.
I hope they did, because there is no microprocessor architecture that holds more promise then the Alpha, and it is a shame on the US supposedly pro-competitive, efficient culture that it has been cancelled due to Digital being inefficient in marketing it and then Intel not wanting the competition.
Leandro Guimarães Faria Corcete DUTRA
DA, DBA, SysAdmin, Data Modeller
GNU Project, Debian GNU/Lin
I read an interview with one of the Dragon Chip
project leader (Dr Hu) a few months ago in a magazine. It gives a lot more details if I can
still recall correctly.
The reporter interviewed him after their team booted into Linux successfully with their prototype chip (or I should say FPGA implementation). Follow the common practice, they have written a C simulator for the chip, followed by hardware logic verification with FPGAs. I think the latest news is refering to
the completion of the initial silicon design.
The team focuses on the hardware design. The proposed chip is compatible with the MIPS instruction, IIRC. For the floating point
arithmatic, it follows the IEEE 754 standard. That's why they can boot to Linux to verify their
design quite early on without too much tweaking.
The targeted performance is close to PII. Not too bad for an embedded microprocessor at this moment... But, maybe a bit old when they commerically release it. But, as long as they can find applications into consumer electronics, the chip may get a good life like our good old Z80, HC11... Nevertheless, it is a good achievement consider the fact that the bulk of the team has no previous MCU design experience.
China hasn't been communist for quite a long time; there's plenty of private industry (both local and foreign), and significant disparity in wealth between the rich and the poor. It's essentially a capitalist one-party state. It's still socialist in some ways, though mostly unofficially (a lot of the large private companies are indirectly controlled by people in high places in the government).
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
The Chinese RMB, on the other hand, is worth a lot less. It's worth 1/8 of a dollar, and average people earn only about 1,000 RMB a month, if they even have a job. A halfway decent, probably barely usable computer costs well over 8,000 RMB, making it out of reach for most workers because they spend most of that money on food and housing anyway.
One reason for the high prices is because of the fact that much of the parts are imported, and only assembled in China under the brands Legend, iBuddie, etc... If this archetecture of chip gets popular in China, more of it will be produced within the nation, making it less expensive, then soon after will come cheaper motherboards, the cases are already made in China anyway... This would mean lower prices, making personal computers within the reach of a lot more Chinese. So, this chip, I say, is a Good Thing(TM), and a step in the right direction.
The distinction is fading quite a bit. Modern x86 chips have RISC cores, but have additional hardware outside the core to translate the CISC instruction set to the core RISC instruction set. On a true RISC chip, the translation from higher-level constructs to lower-level opcodes happens in software at the compilation stage. The functional and performance difference between the two approaches isn't really that huge anymore, since this CISC->RISC translation doesn't slow things down a whole lot.
Now what does slow things down is the hardware having to deal with parallelizing code in the pipeline and avoiding all the variou ssorts of problems that can cause. Both RISC and CISC chips generally do this in hardware. The Itanium is the first to abandon that approach, and say "it's up to the compiler to make sure stuff doesn't mess up when we pipeline." Speeds things up a lot, but makes writing compilers damn near impossible, and writing hand-coded assembler completely impossible.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
The Naval Academy is a real university, and it's better than most.
Jimmy Carter was trained as an engineer probably moreso and better than the average Slashdot reader who self-identifies as "engineer".
Sheesh. "I hoped this has helped a little." Yeah, right.
You're correct only insofar as it's true that the American public doesn't think much of anyone that smacks of intellectualism and rarely do contemporary candidates emphasize their academic credentials. Carter's status as a real engineer, in fact, worked against him as it was used to validate the view that he was a hopelessly naive scientist/engineer type out of his depth in big-time politics. And, honestly, there was probably truth to that at the time.