A month after releasing the architecture specifications and hypervisor APIs for its UltraSPARC T1 processor to the open-source community, Sun is putting out more information on the chip. Sun on March 21 will publish the hardware design for the T1 – formerly codenamed Niagara – and the Solaris operating system simulation specifications for the chip. The move is the latest in the company’s OpenSPARC project, designed to enable hardware and software developers to build atop the new chip’s architecture. The goal is to build an ecosystem around the processor that will expand the reach of the SPARC platform. In addition, Sun’s grid will finally launch this week.
Why? I understand sharing the details about the Sparc core for the Niagara. What I don’t understand is, why give away the piece of tech that is so unique to Sun.
Because they know that nobody can go an make a cheaper version of the Sparc CPU and undercut them (unlike software). You’d have to have a fab and an assembly plant to put things to gether – it’s not easy like software.
This simply buys some good brownie points from RMS and to rub it in IBM’s face.
Because they know that nobody can go an make a cheaper version of the Sparc CPU and undercut them (unlike software). You’d have to have a fab and an assembly plant to put things to gether – it’s not easy like software.
This simply buys some good brownie points from RMS and to rub it in IBM’s face.
Sun itself is fabless, and the fab they are renting is TI, which is definitely not the cheapest. Companies that do have fabs themselves, can produce these ICs for less than Sun does, and have spent US$0 developing them.
Asfor “brownie points from RMS” and “rub it in IBM’s face”, even IF that was an achieved result, what’s theusefulness of it?
It’s simple really. Sun gives all these companies ability to use there state of the art chip, and make there own custom versions of it with special modules in it say 3 cores and one module to do dsp or whatever, and what are they going to develop and test all there code on? Sun Workstations (running the same architecture) and they might as well store all there data on Sun servers as well. What OS are they going to put on there custom chips? Sun’s hope’s it will be Solaris, since its already a stable and robust OS. Its much easier to debug only one variable at a time, in other words why debug the OS and the chip at the same time, go with the ready made solution for one and you suddently have a lot less work. Of course running Solaris on your embeded solution gives you the power of Dtrace to debug your problems.
See the lock in? Once your company relys on Sun hardware for a part it soon will rely on Sun’s hardware for all. Sun isn’t in the custom chip business so releasing plans for there chip really costs them nothing its all possitive unless someone can make a better OS for the chip than Sun.
Have a problem with your design chip? Call up Sun I’m sure they will provide you with one on one time with the guys who developed it and/or the OS if you pay the price or have secured service contracts on your Sun gear. Has any company scheduled one on one time with Linus Torvalds? And he may be great with Linux you still have to find the guys who created your chip parts and how well will they work together if you did manage to get them at the same table? The Sun/Solaris guys have a working relationship that is years in the making.
Which chip are you going to want on your next multithreaded embeded project?
Which chip are you going to want on your next multithreaded embeded project?
The one that provides the required performance yet still does not require much power. Also, being easy to program for is important. But the architecture you develop on does not have to be the same as what you develop for, especially when you start modifying the design of your chip.
Sun may score some nice contracts from this, but I don’t see how the OpenSPARC initiative introduces lock in. Sure, consumers would be more friendly towards Sun, but no lock in.
SPARC has always been an open ISA. In the 90’s several companies competed to supply CPUs to Sun’s own workstations. Sun slipped behind in openness with the UltraSPARC III, aparently (recalling OpenBSD complaints about documentation), but Niagara is a big step back out into the open.
Considering that it is brand new and non-x86, the fact that Linux already boots on Niagara is quite amazing, IMO. This alone is more valuable to Sun than worrying about IBM re-fitting a fab to make SPARCs.
Also, a keyword is “patents”. I’m sure the IP in Niagara is locked up pretty tight.
Now that the dreadful and expensive x86 ISA is conquering all markets, IBM and SUN rush to open their platforms, trying to emulate the success of ARM owners.
Good luck, but I fear it’s too little too late…
Edited 2006-03-21 23:10
x86 ISA owns the home desktop…. a market of diminishing importance. More and more desktop duties are being offloaded to other devices and these offer oppurtunities for other architectures. For example: while ARM on the deskto has failed ARM is the most common ISA on PDAs AFAIK. Also, HPC is not bound to Windows and thus an arena for different ISAs. As consoles pick up more power and the ability to boot Linux (and other *nix I’m sure) different ISAs get a chance to shine. I think the PS3 will raise the interest in PPC Linux overall. I’m sure I’m not the only one getting a hard-on thinking about OpenSolaris/Cell.
Hello,
don’t worry, Sun knows what they are doing. Since no backend data are provided, it still takes a lot of effort to produce such a large and complex design. I think when first chip from competition will be on market, Sun will already have a successor on start. The whole thing about open sourcing the Verilog data is to put these data in simulator and simulate the behavior of the hardware with the software. Another point is the education, so future hardware-developer will be much better educated in design of SPARC processor than in any other.
Regards,
Anton