[oscarMaxwell]

Hi,

Maxwell is not 32 bit single precision float for rendering.

Best regards.

[bpositive]

Oh this is a interesting topic.

I currently have a p4 3 ghz and im saving money for a dual core dual processor opteron sytem, which is the fastest at the moment. But i have also read about the cell processor and i sometimes think that it would rock if you could render on such a machine... Is there any chance that maxwell in the future could be used on a cell equipped machine??

I know i ask the same qustion as Killerx, but im not sure what your answer means Oscar?? Probably a big NO

[Maximus3D]

LoL, you're asking if Maxwell will work on future Cell based computers. The processor is just out fresh of the factories now and PS3 is the only piece of hardware so far to use it for anything practical. Predicting that far into the future at this moment in time is just not possible. No one knows if there will even be Cell based homecomputers, all we can do is hope there will be

/ Max

[bpositive]

To Maximus.

I actually read that IBM was going to use the cell in some of there pc's, so i think the cell will be available in other than consoles, and why shouldn it.

And of course its impossible to predict if maxwell can run on one of those, but my thought was if it was possible at all, since there might be a conclusive difference in the calculation method that would make it impossible.

[Maximus3D]

Yes that's ofcourse IBM's own machines, but i was refeering to other manufacturers which i forgot to specify (sorry). Wonder tho how well those Cell based machines will run Windows probably not, or only via emulation.. which could be one way to do it. Probably a modified/adapted version of Linux will run on it. I guess we just have to wait and see, and hope for the best!

/ Max

[daros]

I think it will be very difficult to write code for that cpu.
Some one know if they are high level languages for that cpu?
If not it will be terrible i think. This is only a intuition and probably i'm worng.

[Mihai]

I posted this earlier about the Cell, it's opinions expressed by people on cgtalk who I got the impression they know what they are talking about:

"but ray tracers are very branch-heavy, and the Cell has a long pipeline, and no out-of-order execution, so it's not well suited to running branchy code. The SIMD engines, being SIMD engines, aren't well suited to running raytracing code, either. It's going to require new algorithms and techniques for the Cell to be good for 3D rendering as we know it now"

"Also, the coprocessors used by Cell have only static branch prediction. "

So as the Cell is now, it is kind of seriously crippled for running any sort of raytracing code.

I have recently read though that IBM demoed a Blade server prototype running with two Cell processors, so they are obviously trying to use it not just for the PS3 and HD TV's. But I think the Cells in that server were basically the same design as those going into the PS3, so we'll have little use for it.

IBM though is trying hard to gather support from software developers and it will be more difficult to program for it because so much of the control mechanisms which were previously handled by the processor hardware will now have to be handled in code, by the programmer. So they have more responsibility in writing efficient code to take advantage of all those cores. But we have to remember multi-core is the future, whether it's the Cell or AMD's quadcore's so programmers will be forced to adapt.

I really do hope IBM makes new Cell based chips geared more towards the scientific community.

[Maximus3D]

You're probably correct about that Daros, coding on the Cell is probably a Hell (wheeee! it rhymes)
But i'm sure Sony developed a strong solid SDK for the PS3 and by the looks of it from the demos i've seen i'd say they done a good job on that part so they must have found a way to rather easily code the Cell.
And it'd surely would be interesting to take a closer look at that SDK to see exactly what can be done with it..

Also i cannot imagine they would make it any more difficult to code for than any other Intel or Risc cpu or else you would see gamedevelopers rip their hair out in anger. More likely is they made it easier
And my guess is that since the Cell cpu gives the coders so much more freedom and that it's much more "open" to develope. Less restrictions which opens up the paths for cleaner faster and more optimized simpler coding (i know it's a contradiction but why not..)

/ Max

[DELETED]

DELETED

[Mihai]

There's already a company who made a custom Cell design for just that. I forgot it's name...

Also i cannot imagine they would make it any more difficult to code for than any other Intel or Risc cpu or else you would see gamedevelopers rip their hair out in anger. More likely is they made it easier

Maximus, that's exactly what PS2 programmers were doing, in fact the poor bastards had to spend all their hard earned cash on hair transplants.

Cell will not be easy, for two main reasons: new ground to explore and multithreading. Combine those two and you have maybe 100 people in the world capable of understanding how to get started with it writing good efficient code. That's why IBM has let out so much information about it, hoping it will be the community that drives tool development, because they need a lot of help to make it accessible to programmers.

[bpositive]

Buffos you are right. we waste to much time to speculate.... We just have to wait and see..

There is probably no reason to wait for some new technology, because at that time another new invention is waiting in the shadows.

So i just wait for my dual core dual processor opteron and then ill be a happy user of maxwell... oh yeah