D
Deleted member 11852
Guest
How did you come to this conclusion? How do you know Sony haven't been working on this for a year or more? How do you know this wasn't the plan right from the start?
PS4 Pro Speculation (PS4K NEO Kaio-Ken-Kutaragi-Kaz Neo-san)
- Thread starter mpg1
- Start date
-
- Tags
- sony sony roach motel
- Status
It's not a conclusion. It's speculation, like everything else being put forward in this thread. Hence the use of the phrasing, "This seems..." as opposed to "This has...".
You can already build a PC with multiple GPUs and CPUs. I'm guessing it won't be extremely hard to implement that with a node reduction and would make it extremely easier to optimize for Ps4/whatevertheycallthenewps4. They could also do some cool stuff with VR, like rendering each eye with each APU separately 
McHuj mentioned that it was brought up in the NeoGAF thread by an "insider" in this post.
No reason to consider the rumor as especially valid beyond the ways that it "fits" the market scenario that I mentioned above. I think it would be a very interesting choice.
D
Deleted member 86764
Guest
Unless I'm being stupid, don't the PC dual GPU solutions replicate their memory across each GPU. So the effective 8GB is still only able to use the 4GB per card and the data is duplicate per card?
Surely it's a waste to have 16GB on PS4, when you can only use 8GB.
Sony would have to have a clever solution to fix the problem, right? With double bandwidth too.
Surely it's a waste to have 16GB on PS4, when you can only use 8GB.
Sony would have to have a clever solution to fix the problem, right? With double bandwidth too.
With low level apis you can use that 4+4 as 8. I think Dx12 supports this as well but you need devs and GPU drivers to support it. You can even use AMD + Nvidia cards together.
Edit: here
This is precisely the kind of thing that the new APIs will fix, with the caveat that it creates a lot more work for the developers. In short, the reason that multi-GPU has worked this way in the past is so that, to the developer, the multiple GPUs look like one "super GPU". The burden then falls on the GPU developer's driver team to make this work on the physical hardware. With DX12 and Vulkan this "implicit" support for multiple GPU operation will be supplemented with "explicit" support meaning that the developer can choose to be aware of the presence of multiple GPUs and specifically target work to them. With the explicit model, limitations like the duplicated memory go away.
Edit: Or you can look at the pretty pictures above.
New or old architecture, it doesn't matter. If it's a new chip, meaning anything is different, it will take a similar amount of time to design the chip. This is the advantage of synthesizeable designs.
Tkumpathenurple
Veteran
The dual APU solution got my electronics-fantasising nerves tingling.
Would it be conceivable of Sony to make the PS4K essentially 2 PS4's? Two users can log in and each play a different PS4 game simultaneously or, if a game supports PS4K enhancements, utilise the full system for a single user.
If I remember correctly, one of the standards in HDMI 2 is the ability to feed 2 sources to a 3DTV, and for each user to receive a full image. I know that's been implemented on TV's in the past, but I should imagine that's from separate cables. If a single console with a single cable can achieve full-console local multiplayer - split screen or full screen, depending on the user's display - I think Sony would have quite a device to tout.
It would also offer incentive to developers/publishers to ensure that the standard PS4 version of a game must run well, and allay a concern held by some. Not that it's a concern I share, but I still think it would send quite the message.
Would it be conceivable of Sony to make the PS4K essentially 2 PS4's? Two users can log in and each play a different PS4 game simultaneously or, if a game supports PS4K enhancements, utilise the full system for a single user.
If I remember correctly, one of the standards in HDMI 2 is the ability to feed 2 sources to a 3DTV, and for each user to receive a full image. I know that's been implemented on TV's in the past, but I should imagine that's from separate cables. If a single console with a single cable can achieve full-console local multiplayer - split screen or full screen, depending on the user's display - I think Sony would have quite a device to tout.
It would also offer incentive to developers/publishers to ensure that the standard PS4 version of a game must run well, and allay a concern held by some. Not that it's a concern I share, but I still think it would send quite the message.
I think the probabilities of the PS4K (or any console) having two APUs, two GPUs or two of anything are lower than CERN creating a black hole at the LHC that would devour the whole solar system.
EDIT: In my defence, I wrote this while I was watching The big bang theory. I'm not usually that geek-apocalyptic.
EDIT: In my defence, I wrote this while I was watching The big bang theory. I'm not usually that geek-apocalyptic.
Last edited:
It depends on the cost, if the die shrink enables them to use two chips with the same architecture as the Ps4 at the same TDP then i think it's very possible that it may happen. First, you don't have to redesign a new chip which saves R&D time and money, second you can have two similar SKUs on the market, sell the "single" Ps4 for 250$ and sell the "double" Ps4 for 400$. Scaling software will be much easier as well assuming that two apus won't create too many h/w conflicts within the system. Third, first, second and third parties already know the architecture, should be much easier to transition to the new system and make engine optimizations.
D
Deleted member 86764
Guest
You wouldn't be able to play two different games at the same time because the two chips would share the same memory.
Having a separate view on a 3D TV is possible on any 3D and split screen game. I remember testing it out for myself on an Xbox 360. It's easier with a passive 3D TV coz it's a simple modification to the glasses; i.e. putting two left lenses in one and two right lenses in the other.
Having a separate view on a 3D TV is possible on any 3D and split screen game. I remember testing it out for myself on an Xbox 360. It's easier with a passive 3D TV coz it's a simple modification to the glasses; i.e. putting two left lenses in one and two right lenses in the other.
Whatever "PS4k" turns out to be, it won't have two APUs. The CPUs in these chips don't have multi-socket support, and having to double up memory chips on top of the APU (as well as power delivery and whatever other bits and pieces) will make the motherboard large and filled with expensive components.
You'd reasonably need to double up storage as well btw, as you'd have to load two instances of a game, and doing it from one drive would slow everything down hugely.
IE... Nope!
colinisation
Newcomer
For what it's worth I saw this discussion about 2 APU's last night and was here to post pretty much what Grall has said. Also how would the 2 CPU's maintain coherency - I assume you would not disable the CPU bank on the second APU.
We have just moved away from having separate memory pools to go back to that would be a step backward IMO.
On to the PS4k itself - I still cannot clearly see what is driving the need for Sony to release a mid gen update - Sony have the more powerful console and the marketshare lead things are going as swimmingly as can be expected.
A new console for PSVR and 4k - really, Sony have already said PS4 can do VR and have announced a product already and 4k support does not need a whole new console. But there is no smoke without fire and given that I have been wrong about this thing at every turn till now. I would imagine PS4k would shape up as follows
An updated APU at 20nm (think 14nm is too risky/expensive)
An updated GPU generation say Polaris - this should maintain backwards compatibility
More shaders available in the premium version - 32 Compute Units @ 1Ghz
Defective APU's sold as bog standard PS4 slim with only 18 enabled CU's
An updated CPU - probably just Puma+
I remember in the X360 / PS3 gen there some talk about die size, bus width and pad limitations - being that if you have a wide bus you could not shrink die size below a certain limit as you could not find the space to route the various connections. Could something like that play a role here - ie you cant shrink orbis below a certain size so you might as well use the area for something?
One thing I would like to see is a whole separate SoC for running the OS - makes no sense to spend all this money on high speed memory then cut out 25% of it to use as OS / background app memory IMO.
We have just moved away from having separate memory pools to go back to that would be a step backward IMO.
On to the PS4k itself - I still cannot clearly see what is driving the need for Sony to release a mid gen update - Sony have the more powerful console and the marketshare lead things are going as swimmingly as can be expected.
A new console for PSVR and 4k - really, Sony have already said PS4 can do VR and have announced a product already and 4k support does not need a whole new console. But there is no smoke without fire and given that I have been wrong about this thing at every turn till now. I would imagine PS4k would shape up as follows
An updated APU at 20nm (think 14nm is too risky/expensive)
An updated GPU generation say Polaris - this should maintain backwards compatibility
More shaders available in the premium version - 32 Compute Units @ 1Ghz
Defective APU's sold as bog standard PS4 slim with only 18 enabled CU's
An updated CPU - probably just Puma+
I remember in the X360 / PS3 gen there some talk about die size, bus width and pad limitations - being that if you have a wide bus you could not shrink die size below a certain limit as you could not find the space to route the various connections. Could something like that play a role here - ie you cant shrink orbis below a certain size so you might as well use the area for something?
One thing I would like to see is a whole separate SoC for running the OS - makes no sense to spend all this money on high speed memory then cut out 25% of it to use as OS / background app memory IMO.
14/16nm is 20nm; it had a new marketing name slapped on it, shrunk only just barely, and got finfet ("3D") transistors added on top. It wouldn't surprise me if most or all of the 20nm manufacturing facilities have been, or will be converted to 14/16nm as soon as manufacturing contracts run out.
AMD have been doing this exact thing for around 15 years with HyperTransport and implemented CPU to GPU coherency in their APU line. Not a deal-breaker there.
It wouldn't be seperate memory pools.
Whatever "PS4k" turns out to be, it won't have two APUs. The CPUs in these chips don't have multi-socket support, and having to double up memory chips on top of the APU (as well as power delivery and whatever other bits and pieces) will make the motherboard large and filled with expensive components.
You'd reasonably need to double up storage as well btw, as you'd have to load two instances of a game, and doing it from one drive would slow everything down hugely.
IE... Nope!
Add a bus to the memory controller to let it talk to the memory controller in the other APU. Again, not a deal-breaker. I agree that the added motherboard complexity is an issue, though.
Prophecy2k
Veteran
Wouldn't two APUs sharing a single memory pool on a single bus end up bandwidth starved when they are both trying to access data from main memory at the same time?
On PS4 contention between the GPU and CPU already causes BW for each to drop disproportionately. I can't think how that will work for two entire APUs.
On PS4 contention between the GPU and CPU already causes BW for each to drop disproportionately. I can't think how that will work for two entire APUs.
Two APUs will not happen, that will just add unnecessary complexity and increase of cost.
BTW I wonder if there is a chance of this happening:
Sony and AMD design a new beefier APU that can perfectly handle native PS4 games, and then put that APU in two consoles - PS4 Slim and PS4K. Users who purchase cheaper PS4 Slim get "permanent low performance" mode so they can only experience PS4 content in a ordinary way, but with a PSN Store options to unlock higher clocks/inactive CUs and transform their console into PS4K. Sony could justify the use of big APU for both consoles by eliminating the need to manage two chip production lines.
This is a unique chance for Sony to release console that can almost truly "download more ram"!
BTW I wonder if there is a chance of this happening:
Sony and AMD design a new beefier APU that can perfectly handle native PS4 games, and then put that APU in two consoles - PS4 Slim and PS4K. Users who purchase cheaper PS4 Slim get "permanent low performance" mode so they can only experience PS4 content in a ordinary way, but with a PSN Store options to unlock higher clocks/inactive CUs and transform their console into PS4K. Sony could justify the use of big APU for both consoles by eliminating the need to manage two chip production lines.
This is a unique chance for Sony to release console that can almost truly "download more ram"!
- Status