Slashdot Mirror
Sun Moves Into Commodity Silicon
Samrobb writes "According to Sun Microsystems CEO Jonathan Schwartz, Sun has decided to release its UltraSPARC T2 processor under the GPL. Schwartz writes, 'We're announcing the fastest microprocessor we've ever shipped this week — delivering 89.6 Ghz of parallel computing power on a single chip — running standard Java applications and open source OS's. Simultaneously, we've said we're entering the commodity marketplace, and opening the chip up to our competition... To add fuel to the fire, the blueprints for our UltraSPARC T2... the core design files and test suites, will be available to the open source community, via its most popular license: the GPL.'" Sun is still working on getting these released; early materials are up on OpenSPARC.net.
Because MAJC, picoJava, aJile, and Jazelle don't count, right?
Javascript + Nintendo DSi = DSiCade
How many cores does this chip have?
8 cores, 8 threads each.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
The original T1 design was released under GPLv2 a while back, and the new T2 design will be released under GPLv2 for now as well. Using GPL v2 creates an implicit patent license so while using GPLv3 would make the situation cleared GPLv2 is probably sufficient.
And yes if you look at the map on the opensparc.net page (when they get some quota back after being slashdotted) you'll see they are getting a vast amount of interest from China, where I gather a company is already producing an OpenSPARC T1-derived chip for embedded use.
Well, apart from Simply RISC, who used the design to build a single-core chip (S1) for embedded applications.
And Polaris Micro in China, who are doing the same.
And David Miller & friends, who made Linux run on it.
And Canonical who support Ubuntu running on it.
And the other Linux distros picking it up.
And... Oh, sorry, you were just trolling, right?
Indeed, someone just did:
More details on Simply RISC's web site.
The T1 is NOT abandonware, in fact I would say it is one of Sun's greatest strengths. We are doing a design for a JD Edwards data warehouse and while our JDE system is on Oracle on Windows we are looking at Unix platforms as strong choices for the data warehouse. Thanks to only needing 6 total boxes for the middleware layers for 4 different environments vs 16 Windows boxes Sun is 10% cheaper and 10% lower in 3 year operating costs despite having power sucking, expensive DB servers.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
It is GPLv2:
http://opensparc-t2.sunsource.net/
That comparison is between a 1.0 GHz UltraSPARC T1 - a three-year-old chip. There are 1.4 GHz chips currently available. How many months has the Intel Q6600 been on the market?
How well will that Intel architecture scale to over 4 CPUs, anyway. At least AMD can do that.
1.4 GHz * 8 cores * 8 threads = 89.6 fake GHz.
I wonder how many BogoMIPS that is equivalent to.
Just to quell the concerns of "abandonware" and cries of "performance benchmarks"
:-)
Linky on numbers
Summary:
* This puppy comes ahead of Power5 and top-dog (till now) Power6
* Highest single CPU integer and floating point performance
Oh, and it has 2 10G network interfaces on chip... and EIGHT crypto cores to keep them running full throttle too. All this with 8 core each with its own floating point unit and 8 threads.
Oh and BTW, Ubuntu guys just booted their distro on this puppy
So yeah, it runs Linux (too)!
- mritunjai
Engineering is the art of compromise.
what's the hardware meaning of a thread?
As another poster has pointed out: You build a core with multiple copies of the register set and replicate (or take turns on) the associated instruction-dispatching logic.
But these multiple CPUs share a common set of arithmetic/logic execution units, along with arbitration logic. Different threads will be doing different things at any given instant, and thus using different sections of the ALU.
The arbitration logic decides which thread gets which hunk of ALU at any given moment. And some threads will be stalled waiting for data and won't need any ALU function at all. Of course when more threads want a particular kind of execution unit than are available, one or more of them must stall. But by having the right number of copies of the commonly-used types of execution units you can keep a number of threads running at or near full speed most of the time and the ALU components mostly busy, with much less silicon than if each thread had a full-blown CPU and most of the ALU logic was idle at any given moment. With less silicon logic you can put things closer together and speed it up still more.
This approach has been around since Cray was at Control Data.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
If I remember it right they max out at 72 watt for 8 cores running 8(?) threads each.