PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)
- Thread starter Love_In_Rio
- Start date
- Status
Huh? i'm not sure where I supposedly said that (maybe I meant the PS4 has 40% more compute now after the XBO clock bump? Only thing related to ~40% I can think of...)
PS4 and XBO have the same CPU...
Silent_Buddha
Legend
Pretty sure Silent Buddha is a dev
Oh dearie me, no. I've never said that and would hope that I've never implied it either. So, let me put a stop to that before any rumors about that start up. We've got enough rumors floating around as is.
Regards,
SB
I could see this gen's cycle be slightly shorter actually because the hardware is relatively easier to get into and max out, and cost reduction may go faster this time too. Which is to see it could go back to the more normal 5 year cycle. However, Sony tends to envision a total life-cycle of about 10 years for each of their generations regardless of when the next one starts, so he may just as well be referring to that.
pMax
Regular
My description was vague, for someone that was not familiar with CPU/GPU intricacies... but you raised the interesting question of "what makes a core?"
Probably, what makes me not considering them a core is:
...and the fact that you cannot index them directly - that is, like a SPU (an embedded RISC processor would still be a full processor, even if you cut the outside access, but it is in principle there, at least).
An embedded RISC processor with no access to the outside, would still support interrupt and all the techs that are part of a processor. CUs do not (at least, yet).
My lowest-level definition is that "core" is a physically distinct path that an instruction stream's associated signals traverse between stages that reside in a generic Von Neumann or related architecture's memory.
This does still require at least one ALU, and at least one port to memory, but exclusivity for either is not absolutely necessary--although it gets iffy once this gets reduced to near-absurdity.
At a bare minimum, there would need to be a physically independent pipeline control unit, a mechanism for generating the program counter, and some physically independent context storage necessary for the the prior two elements to work through an instruction stream.
So, it's a physically separate pipeline control, and at least enough storage for a program counter.
The gray area, since I didn't require ALU exclusivity as an absolute, is if the necessary ALU for generating the PC were somehow shared between two cores. However, that would require an arrangement where an ALU is subject to two schedulers or pipeline control units that are still somehow physically distinct, and I haven't thought up an arrangement where that's workable.
Everything beyond that is increasingly negotiable in terms of sharing or necessity.
Decode? Could be shared, and for some defunct VLIW CPUs, not necessary.
FP? Could be shared, and CPUs predate that functionality.
Cache? Same.
A Von Neumann architecture, in its simplest theoretical implementation, makes no provision for interrupts. The preponderance of all cores in the world do service interrupts, but at least the possibility of a simple system that utilizes no interrupt signals is allowed for, as odd as that might be.
The choice to hide direct indexing of the hardware is more of an API or platform issue. At least from the POV of the GPU's internal firmware, it can make the distinction.
The PS4 in particular may be allowing lower-level access for compute, although what that entails is not disclosed.
In the other direction, the increasing virtualization of execution resources or hardware-level sleight of hand makes it increasingly less likely that the cores software exist or are in use are what it actually gets.
Not yet, but the architecture has been drilling down to very few features out of a large list of attributes ascribed to cores.
I think it's close enough to consider them for the most part primitive cores, with some haziness when it comes the scheduling hardware that kicks in at the boundaries of kernel execution--which would include scheduling new contexts or someday potentially switching them.
Apparently you'll be managing your own apps?
I don't think they are shipping the dev kit UI to retail. BTW what exactly does managing your own apps even mean?
Specifically closing apps instead of the system doing so for you may imply some apps can run in the background (or the user is doing memory management on their own). Or maybe it's just a devkit thing.
It's the same on iOS, Android and Vita, fyi, the option to actively close an app. Except I think on Vita you can only explicitly close Apps, with a maximum of 6 open at once. And I have heard Sony imply that Vita's OS may be the foundation for the PS3's OS as well.
Don't forget the Ask and Yahoo toolbars.
http://www.examiner.com/article/geo...ch-ram-you-can-actually-touch?cid=db_articles
Makes sense as discussed here before.
I don't particularly mind if the PS4 only gets 5~6 GB of ram to the game.
176GB/30= 5.87 GB theoretical max usable BW per frame, and we know it's hard to even come close to this number.
Bandwidth will still be the real limiter.
Makes sense as discussed here before.
I don't particularly mind if the PS4 only gets 5~6 GB of ram to the game.
176GB/30= 5.87 GB theoretical max usable BW per frame, and we know it's hard to even come close to this number.
Bandwidth will still be the real limiter.
The way I understand it, yes you do need to re-access pretty much everything every frame, even if the data is carried over from the previous frame. Some stuff you won't have to re-write, but if it's needed to render a frame, you will have to read it at some point.
Of course stuff like textures that don't appear in this particular frame won't be accessed.
There's a lot more in memory than just graphics, and even not every texture in memory would be displayed all the time, every frame.
djskribbles
Legend
- Status
