IBM to offer Open Source for some Cell software tools

[ShootMyMonkey]

Wow, so they are shooting for temperature at which chip can actually operate for an extended amount of time without a refrigerator attached to it? Those people at Sony sure set ambitious goals for themselves.

At 3.2 GHz, it shouldn't be a problem --

I'm assuming worst-case scenario that these are truncated figures (meaning 3 W could actually be as high as 3.99 W)

But, assuming that this means you can run at 1.0 Vcc at 3 GHz (these figures are for a single SPE)... I'd assume this means somewhere around 35-40 W at 3.2 GHz for CELL, although I'm really just guessing on the PPE consumption there.

Reminds me of the engineering in joke about the specific heat rejection of CPU cores approaching that of nuclear reactors.

I believe the data for P4 suggests that Intel crossed the halfway point to that with the very first Prescotts.

It seems like it wasn't that long ago that we were looking at the first heatpipe CPU coolers, and just thought it was absurdly superfluous and just plain overkill -- toys for those who would want to overclock to hell and back and play with Peltiers. Now we don't even blink at the sight of a stock cooler with 6 or 7 pipes and some 500 fins.

 

[Carl B]

I don't get it - wasn't this thing supposed to happen in Barcelona today? I would imagine then that we'd already be seeing some of the architecture details all over the Internet.

And to re-address some of the points I made previously on the first page of this thread, I wonder what Sony sees the risks as of possibly competing against a Cell derived chip next gen. Also, since IBM seems quite willing to sell Cell chips to whoever wants them, I have to imagine that Sony and Toshiba must also have plans in place to monetize on this should a robust software library indeed create itself around Cell - assuming of course that they had to give their 'ok' on this open-sourcing plan. Which is my assumption.

EDIT: Well, the actual conference is tomorrow - THAT's when we get the goods. Speaking of which, I notice Toshiba to be a noticeable abstainer from the 'Power.org' movement. After reading about the intitial strife between the three companies when determining the base architecture to build Cell around, I wonder what the future holds for the partnership.

 

[one]

Speaking of which, I notice Toshiba to be a noticeable abstainer from the 'Power.org' movement.

Toshiba got a PowerPC license back in 1994 so I don't think they are particularly out of Power. Also Toshiba is selling IBM's Linux POWER cluster servers. Toshiba is a big company and using all architectures including ARM, MIPS, and Intel, so it may be a reason not to join Power.org though. But it may change in future with Cell.

http://www.ee.surrey.ac.uk/Contrib/Edupage/1994/10/02-10-1994.html

POWER PCs BY TOSHIBA

IBM has granted Toshiba Corp. a license for IBM's PowerPC microprocessors and AIX operating software. The agreement also gives Toshiba the right to develop its own version of the PowerPC. Toshiba plans to use the chip in its mid-range computer line. (Investor's Business Daily 9/30/94 A7)

http://www.techworld.com/opsys/news/index.cfm?NewsID=2732

Some prominent Power licensees like Toshiba and Apple are not on the list however, but IBM said it expects to announce new members in the weeks ahead.

Apparently http://www.power.org/news/events/barcelona/broadcast/
is going to have an event broadcast from June 9th, including

17:40 - 18:20
Unleashing the power: A programming example of large FFTs on Cell
Alex Chow, IBM

 

[Carl B]

Thanks for the knowledge on Toshiba One - didn't know that.

Still, just like it mentioned, Apple is also a big licensee, and we all know what's going down there...

Interestingly it seems both Apple and Toshiba signed on for Power the same year.

I saw that presentation on the schedule on the power.org website as well; should be interesting. It may be the moment at which they 'open-source' the Cell as well, being that it'll be tomorrow at some point or another.

 

[Panajev2001a]

and since Sony paid the majority of R&D costs, my logic thus follows...

Too bad that the core which was used to deliver both the PPE's and the Xbox 360 cores is IBM's property and IBM did the majority of the work on it. The Xbox 360's CPU is also quite customized .

 

[nAo]

What kind of documentation about CELL will be released today?
Where? When?
I can't find anything on power.org

 

[Titanio]

What kind of documentation about CELL will be released today?
Where? When?
I can't find anything on power.org

Theres a presentation at 5.40 local time in Barcelona. I've tuned into the webcast, but evidently they're on a break at the moment. They say presentations will start again at 4.00pm CET (i.e. in a few minutes).

We should be all watching, as we may be coding on Cell sooner than we think! (if Kutaragi isn't blowing smoke about Linux on PS3 HDD at any rate).

 

[Vennt]

What kind of documentation about CELL will be released today?
Where? When?
I can't find anything on power.org

Theres a presentation at 5.40 local time in Barcelona. I've tuned into the webcast, but evidently they're on a break at the moment. They say presentations will start again at 4.00pm CET (i.e. in a few minutes).

We should be all watching, as we may be coding on Cell sooner than we think! (if Kutaragi isn't blowing smoke about Linux on PS3 HDD at any rate).

Do you have a link for the webcast?

Thanks,

V.

 

[Titanio]

What kind of documentation about CELL will be released today?
Where? When?
I can't find anything on power.org

Theres a presentation at 5.40 local time in Barcelona. I've tuned into the webcast, but evidently they're on a break at the moment. They say presentations will start again at 4.00pm CET (i.e. in a few minutes).

We should be all watching, as we may be coding on Cell sooner than we think! (if Kutaragi isn't blowing smoke about Linux on PS3 HDD at any rate).

Do you have a link for the webcast?

Thanks,

V.

Try here:

http://www.gginternet1.co.uk/powerorginbarcelona/

There's a box to submit questions too..so the more learned amonst you may have an opportunity to ask something pertinent if they're taking Qs during the Cell presentation.

 

[Carl B]

Too bad that the core which was used to deliver both the PPE's and the Xbox 360 cores is IBM's property and IBM did the majority of the work on it. The Xbox 360's CPU is also quite customized .

Yeah after reading some of the interviews/stories dealing with the creators I realize just how much that PPE core has to do with IBM now. So my comment on Sony funding Microsoft's R&D was a little out of line.

Still, I imagine Sony and Toshiba have full rights to all the IP in Cell (inclusive of the PE core) and I wonder what sort of contingencies have been put in place to make sure that the Cell architecture doesn't boomerang around and get 'used against them' so to speak with this open-sourcing deal, because on it's face it really seems to benefit IBM most of all, even though all three parties would have had to sign off on it.

 

[Titanio]

Alex Chow is on now. Talking about programming models on Cell, and an example of it (FFTs).

 

[Carl B]

Ack, I'm getting no audio whatsoever. :?

 

[Titanio]

I'm getting the audio, but I'm sure someone else would better understand everything. I'm capturing the slides as I get them.