Sun, allies broaden open-source chip push

(continued from previous page)

"We believe in the strength that comes from the use of the GPL license," said Fabrizio Fazzino, one of Simply RISC's two main designers. "There are many success stories in the world of software, and we believe that it is time for something similar to happen in the hardware world."

"There are many success stories in the world of software, and we believe that it is time for something similar to happen in the hardware world."

--Fabrizio Fazzino, Simply RISC designer

The choice of the GPL has interesting implications. The license permits anyone to change designs--but if an organization distributes products based on the modified design, it must publish those changes.

The GPL serves as a convenient check on rivals who might want to profit from Sun's multicore investments by wrapping their own interface on a Sun Niagara core, Yen said.

"With the GPL, if our direct competition--IBM, Intel and AMD--wants to copy what we have done in high thread count per core, they will have to put (their changes) back. That will potentially be a barrier to just copying what we've got," Yen said. Those rivals might be leery of revealing not just their own designs, but also might be forbidden from revealing third-party intellectual property, he said.

Cores and threads
Sun's Niagara chip is the industry's most aggressive example to date of general-purpose multicore chips--those with multiple processing engines on each slice of silicon, as opposed to the comparatively old days when each chip had only a single processor engine. Niagara has eight cores, and each core can process four instruction sequences called threads.

Niagara 2 still has eight cores, but each can handle eight threads, and the chip has better number-crunching abilities as well as built-in encryption, input-output control and 10-gigabit-per-second networking. Niagara 2 servers are due to ship in the third quarter of 2007, Sun Chief Executive Jonathan Schwartz said in April.

Sun recognizes that this multicore, multithreaded approach isn't easy for some necessary partners in the software industry to digest. Trying to encourage their support is another reason Sun opted for the OpenSparc project.

"We strongly believe this multicore, multithreaded direction is the way to go, both for efficiency in computing and efficiency in power consumption," Yen said. "But this cannot be done alone by hardware processor vendors. It requires the software community, all the way from system software to application software, to adapt and participate," he said. "By open-sourcing OpenSparc S1, it's also sending strong signals urging the IT community to go in that direction."

But it's notable that the two companies so far working on open-source derivatives aren't as bold. Both Polaris Micro and Simply RISC have designs with a single four-thread core.

There are complications to making Niagara 2 an open-source project. One of them is export control, because the U.S. government imposes restrictions on encryption technology.

"Suppose today I want to publish an implementation of elliptic curve cryptography algorithm. I'm not sure the government will allow us to do that," Yen said. "There are more things we have to work out, clarify, maybe get certain permissions."

Programmable chips
Making a chip design an open-source project is a very different beast compared with open-source software projects.

It's easy to get a computer, download some software source code, and start programming away. But when it comes to hardware design, fewer people have expertise, and a chip foundry isn't likely to be interested in allocating fab capacity to build a handful of some amateur's experimental processors.

But for would-be open-source chip designers, there's another way: field-programmable gate arrays, or FPGAs. These programmable chips are blank slates from companies such as Xilinx onto which designers can load whatever hardware logic is desired.

To that end, Sun has released a version of OpenSparc that runs on an FPGA, said Shrenik Mehta, Sun's senior director of the OpenSparc program. The design has just a single core with a single thread so students can figure out ways to beef it up, he said.

"It works like a building block, so students can experiment in the lab or course work to do different designs with one core and two threads, two cores, etc.," Mehta said.

Indeed, that's just what the University of California at Santa Cruz is doing in one of its courses. Sun established a center of excellence at the university to try to foster collaboration, and a U.C. Santa Cruz assistant professor, Jose Renau, is a member of the OpenSparc community governance board.

Though Sun is pleased with the open-source chip effort so far, including 4,700 downloads of the design, the company recognizes that progress so far is very early.

"Though the publicity is getting broadened," Yen said, "most people are still digesting and trying to understand what we are offering."

[Blito]

I still think Open Spec type Hardware is safer and more competative

I really think it's more efficient and safe to be able to view the hardware and have it fixed by outside companies. Maybe you wouldn't fix the actual chip BUT it can be replaced more effectively by modification and also other chips built around that chip by alternative companies instead of just relying on the "Papa" company to just replace your stuff with a warrenty. No seious innovation and especially no customization. It's much more efficient to build around chips then to constantly reinvent the wheel with just a few companies.
Also, safer, meaning I know what I am getting and can protect myself from hackers etc. allot better. It's bad enough that specs aren't released to the public for microchips (and atomic sized nano-chips are here) but AMD's plans to start DRMing would be disastrous.

I might be to extreme in some areas of my opinion but this is where I stand now. I just think the move from tubes and wires to microchips hasn't realy been that enlightening.

I still think the constant short term (2 years) computer upgrading was and is a scam.

[bob donut]

The people at Sun don't get it

I know some people who've worked at Sun. one thing I've found is that they get so into the free-love, free software thing they forget that business is basd on competition.

Let's recap: a business is an organized group of people (usually not an individual) designed to provide goods and services in exchange for something else, usually money, AND does so in a "market", in which there are others attempting to do the same.

It's amazing that Sun hasn't died yet, but culturally, they aren't the type to seek advantage in the market. Eventually, they'll just end up being a service provider like Red Hat. Their cash cows, their software and hardware, are almost dead.

[American 1st]

Really?

Bob's comment misses the fact that Sun is doing more for the industry than any of their competitor's combined. It is easy to underestimate Sun if you don't have any vision or insight into where they're going; and, frankly, that's Sun's fault. They're "wasting" money on innovation rather than advertising. Sun powers more of the world's computing than most other vendors. The open source model provides many thousands of "developers" that produce code constantly which is reviewed for possible inclusion in Solaris, for example. There is a new release every month that has potential to become part of the OS. Many folks don't "get" Sun including many of their employees. But, to count them either "down" or "out" is to misunderstand everything about them. They are a for-profit company and the efficacy of their vision will be clear more and more as time goes on. Bob's company must not have to compete with Sun because he would be virtually "bleeding" from the battle. Bob needs to get out more to see what reality looks like. Look for more: lower cost Java-based cell phones; a more user-friendly Solaris; higher performance, lower power consumption, denser cpu chips; and many other things from SunLabs. Sun not only "gets it" but they will count on folks like Bob to win.

[bradford653]

Hardware not dead

I was at a major data center today that hosts some big name
companies... I can assure you that Sun's hardware is not dead, and
in fact makes up a substantial percentage of the thousands of
boxes that I saw among the racks. Just because you wouldn't buy
one for your desktop computer doesn't mean that it's worthless.