(Rumour) XB2 CPU @ 65nm
- Thread starter Dave Baumann
- Start date
- Status
RaolinDarksbane
Newcomer
You mean MS and Nintendo, since Sony has their own fab(I think....).
RaolinDarksbane
Newcomer
Aren't OS Kernel and API two really seperate things? They aren't completely dependant on each other.
...
Yes, you can have DX10 subsystem run on CE or Linux kernel with proper development, but it will take far more time and delay the Xbox Next launch. As you may recall, it is always the software delay that pushes back a console launch, never the hardware delay. MS has no reason to not include NT Kernel with Xbox Next, since they own it and can include it at no cost.
The whole point of my arguement was that MS needs to run NT kernel on Xbox Next and this excludes the possibility of MS licensing some kind of CELL device, since large NT kernel and CELL APUs are mutually incompatible.
Not necessarily. The point I am trying to make is that MS needs the latest DX subsystem to drive XGPU2 for Xbox Next and NT kernel has the most upto date DX subsystem development.
Yes, you can have DX10 subsystem run on CE or Linux kernel with proper development, but it will take far more time and delay the Xbox Next launch. As you may recall, it is always the software delay that pushes back a console launch, never the hardware delay. MS has no reason to not include NT Kernel with Xbox Next, since they own it and can include it at no cost.
The whole point of my arguement was that MS needs to run NT kernel on Xbox Next and this excludes the possibility of MS licensing some kind of CELL device, since large NT kernel and CELL APUs are mutually incompatible.
That's a total non-sequitur. DX interacts with the kernel purely through the driver interface, which won't even be there in an embedded console OS that targets a single hardware platform. Not to mention, the version of DX that targets the hardware that the Xbox Next will have probably hasn't even been written yet, for any OS (well, ok, it's probably being worked on, assuming the capabilities of the target hardware are reasonably stable).
Re: ...
No DLL support in XBox has got nothing to do with one application running.
At some point the latest DirectX will be handed to console team and they will reengineer alot of it. Thats what they've done with DX XBox and likely to happen on the next OS. DX Xbox is almost completely rewritten from the early version of DX 8 is split from.
Deadmeat said:
No DLL support in XBox has got nothing to do with one application running.
At some point the latest DirectX will be handed to console team and they will reengineer alot of it. Thats what they've done with DX XBox and likely to happen on the next OS. DX Xbox is almost completely rewritten from the early version of DX 8 is split from.
More than likely.Wunderchu said:
IBM is not producing the versions of CELL Sony will be using in the PS3. That will come out of Toshiba and Sony fabs. (You can search around for the specifics--it's certainly been mentioned enough in other threads.)
Re: ...
IIRC the APUs are running their own show, the PowerPC is “onlyâ€￾ there to do housekeeping and distribution of tasks and such. So it doesn’t really matter what kind of OS it's running.
Seeing as the APUs really are independent processors, I am a bit puzzled as to why the engineers didn’t give a specialized APU like processor those duties? That would maybe have been more elegant, than the PowerPC, with all its legacy technology. But the architecture is very flexible, so maybe that's one for future iterations of Cell?
Re NT kernel: I don't think it is really a true microcore in any way. It probably started development as one, but soon, for some reason, it mutated into something else.
I would have thought a true micro kernel like QnX Neutrino or Mach, would have been more suitable for PS3, than a cut down proprietary Linux.
It would have been able to reside in the eDRAM buffer of the PowerPC, and be real stabile and efficient.
Gubbi said:
IIRC the APUs are running their own show, the PowerPC is “onlyâ€￾ there to do housekeeping and distribution of tasks and such. So it doesn’t really matter what kind of OS it's running.
Seeing as the APUs really are independent processors, I am a bit puzzled as to why the engineers didn’t give a specialized APU like processor those duties? That would maybe have been more elegant, than the PowerPC, with all its legacy technology. But the architecture is very flexible, so maybe that's one for future iterations of Cell?
Re NT kernel: I don't think it is really a true microcore in any way. It probably started development as one, but soon, for some reason, it mutated into something else.
I would have thought a true micro kernel like QnX Neutrino or Mach, would have been more suitable for PS3, than a cut down proprietary Linux.
It would have been able to reside in the eDRAM buffer of the PowerPC, and be real stabile and efficient.
Panajev2001a
Veteran
nobie said:
APUs have their own Program Counter Register and they have their Stack and their Heap ( in the 128 KB of Local Storage they have ).
APUs can DMA to and from DRAM by talking to the DMAC and they can also do some form of I/O to and from I/O devices.
That's why I wrote, "specialized APU like processor".
It would have special extensions to make it better suited for the job, but at the core, would still remain the basic APU.
Of course, I'm not qualified to say if it would indeed be a better solution, but nevertheless, to me, it would seem somehow "neater".
It would have special extensions to make it better suited for the job, but at the core, would still remain the basic APU.
Of course, I'm not qualified to say if it would indeed be a better solution, but nevertheless, to me, it would seem somehow "neater".
Re: ...
If CELL is an option, it would boil down to performance compared to a traditional CPU like PowerPC. I doubt Operating System issues will play a major factor in the decision process if the performance gap between PowerPC vs CELL is large. If your opinion is that Microsoft won't go with CELL, even if it is vastly more powerfull, because it would require additional work on the Xbox 2 OS, I don't agree with that.
Deadmeat said:
If CELL is an option, it would boil down to performance compared to a traditional CPU like PowerPC. I doubt Operating System issues will play a major factor in the decision process if the performance gap between PowerPC vs CELL is large. If your opinion is that Microsoft won't go with CELL, even if it is vastly more powerfull, because it would require additional work on the Xbox 2 OS, I don't agree with that.
Panajev2001a
Veteran
nobie said:
You could in theory take an APU, the way they designed it, and use it as a CPU in another system with few modifications here and there.
APUs are independent processors which can be commanded by this super-CPU which is the PU in the case of the PE.
I understand your point, but still what you said about APUs being like "a bit specialized ALUs" is not correct.
...
But the effect of DirectX is that Xbox ended up running on a stripped down NT kernel, which is about to happen again on XB Next. And this NT kernel requirement makes the "CELL" an impossibility. XBox Next is definately going for a Power5-style SMT.
DLL exists to share library code between multiple running applications. Since Xbox Next will run only one application(game) at a time, there is no need for library sharing, so DLL can be omitted without a problem.
But the effect of DirectX is that Xbox ended up running on a stripped down NT kernel, which is about to happen again on XB Next. And this NT kernel requirement makes the "CELL" an impossibility. XBox Next is definately going for a Power5-style SMT.
Yes it is. MS has lots of code to port over, and it won't run on something like CELL.
The rule of semiconductor biz is that no chip fabricated on identical process around a similar timing blows the other away by more than a factor of three. CELL won't be "vastly more powerful" than Power5 lite that MS is putting inside Xbox Next.
- Status
