No doubt that it can handle Cell. Programmers can do just about anything if given the incentive. The question is will Cell function well as a general purpose CPU? That's for the brainiacs to decide.
No doubt that it can handle Cell. Programmers can do just about anything if given the incentive. The question is will Cell function well as a general purpose CPU? That's for the brainiacs to decide.
A SMT PPC primary core with AltiVec, running between 4.0 and 4.6 GHz, should handle Tiger quite fine. Whether Tiger could use any of the extra SPU's to get real Cell bang for the buck is the open question.
Let's assume Tiger can handle the Cell in some way. What form would the performance boost come in? Video? Rendering? Gaming? Other desktop apps (browsers, etc)? Everything?
Cell is a departure from traditional computer CPUs and software would have to be rewritten completely to make use of Cell parallel processing efficiently.
Cell is a departure from traditional computer CPUs and software would have to be rewritten completely to make use of Cell parallel processing efficiently.
Good point, well made - don't think of it as OSX working well on a cell architecture, thing about how cell-tuned GCC can become
Cell is a departure from traditional computer CPUs and software would have to be rewritten completely to make use of Cell parallel processing efficiently.
But the PSU is essentially a G5, and the APUs are at least "Altivec-like" vector processors (the opinions I've seen are that it's not quite Altivec, but converting between them shouldn't be difficult). So how hard could it be to tweak Altivec calls to use the APU's? If Apple has plans to use Cell, I'd imagine they'd make sure current software could run on the PSU without modification - but with no benefit. And that Altivec-enabled software could use all the PSUs, with a concomitant huge performance boost. Maybe that would require a recompile, but maybe not. Even if so, Apple itself could make sure CoreAudio and CoreVideo were ready-to-go when Cell is released, so any Cocoa apps that leverage them would get an instant payoff.
The main cell processor includes VMX (Altivec), but performance is probably only on the level on early G4 chips. i.e. weak. Too weak for Apple who has a heavy investment in Altivec.
The main cell processor includes VMX (Altivec), but performance is probably only on the level on early G4 chips. i.e. weak. Too weak for Apple who has a heavy investment in Altivec.
Wouldn't 4 GHz of early-G4 equivalence blow a 2 GHz G5 out of the water? G5's don't have much (if any) clock-to-clock benefit over G4's, right?
Comments
with enough work I would imagine that any OS could run on any processor.
WinNT used to work on PPC/68k
There is a new version of the NT kernel that apparently works on PPC as well since MS XBOX engineers are using PMG5s for dev work...
anyway, AFAIK, FreeBSD is/was extremely agnostic in regards to hardware. The prob is getting all the other stuff (QT, etc etc) collaborating as well.
In other words, I have no idea
Apple has known about CELL for many many years and will have done what is necessary in that time to adapt. At least we so hope.
Originally posted by hmurchison
No doubt that it can handle Cell. Programmers can do just about anything if given the incentive. The question is will Cell function well as a general purpose CPU? That's for the brainiacs to decide.
A SMT PPC primary core with AltiVec, running between 4.0 and 4.6 GHz, should handle Tiger quite fine. Whether Tiger could use any of the extra SPU's to get real Cell bang for the buck is the open question.
Originally posted by kim kap sol
Cell is a departure from traditional computer CPUs and software would have to be rewritten completely to make use of Cell parallel processing efficiently.
Good point, well made - don't think of it as OSX working well on a cell architecture, thing about how cell-tuned GCC can become
Originally posted by kim kap sol
Cell is a departure from traditional computer CPUs and software would have to be rewritten completely to make use of Cell parallel processing efficiently.
But the PSU is essentially a G5, and the APUs are at least "Altivec-like" vector processors (the opinions I've seen are that it's not quite Altivec, but converting between them shouldn't be difficult). So how hard could it be to tweak Altivec calls to use the APU's? If Apple has plans to use Cell, I'd imagine they'd make sure current software could run on the PSU without modification - but with no benefit. And that Altivec-enabled software could use all the PSUs, with a concomitant huge performance boost. Maybe that would require a recompile, but maybe not. Even if so, Apple itself could make sure CoreAudio and CoreVideo were ready-to-go when Cell is released, so any Cocoa apps that leverage them would get an instant payoff.
Can they eventually use Cell?
Sure.
Will they?
Probably.
Soon?
Nope. Think years.
Originally posted by Electric Monk
The main cell processor includes VMX (Altivec), but performance is probably only on the level on early G4 chips. i.e. weak. Too weak for Apple who has a heavy investment in Altivec.
Wouldn't 4 GHz of early-G4 equivalence blow a 2 GHz G5 out of the water? G5's don't have much (if any) clock-to-clock benefit over G4's, right?
We have to let this whole CELL thing settle a few days/weeks before making any kind of deductions.