Predict: The Next Generation Console Tech

[archangelmorph]

The ps3 will most likely use a new cell chip. Perhaps a 4x32 if its avaible at a decent price. If not a 2x16. Mabye a 3x24 ? It will most likely have a large mount of ram also. It will use a bluray drive , I'm expecting 12x. It will also use a new nvidia gpu. Most likely not nvidia's newest card or anything custom , just a modified desktop card from a few months before the ps4s release. I expect this in 2012.

Highly unlikely considering nVidia began working on the next GPU part for the next gen playstation a while ago. There's little reason I can see for Sony to go with a modified desktop chip a la RSX as they aren't constrained by time this time around. Not to mention the fact that alot of the system imbalance of the PS3 comes from the lack of a more custom RSX design which would provide greater scope for load balancing & much tighter Cell communication..

I'm pretty sure the next solution (GSRSX maybe?) will either be an entirely custom chip or a lighter custom hardware co-processor (a la GS) with a very-high bandwidth interface integrated into the Cell..

 

[eastmen]

Highly unlikely considering nVidia began working on the next GPU part for the next gen playstation a while ago. There's little reason I can see for Sony to go with a modified desktop chip a la RSX as they aren't constrained by time this time around. Not to mention the fact that alot of the system imbalance of the PS3 comes from the lack of a more custom RSX design which would provide greater scope for load balancing & much tighter Cell communication..

I'm pretty sure the next solution (GSRSX maybe?) will either be an entirely custom chip or a lighter custom hardware co-processor (a la GS) with a very-high bandwidth interface integrated into the Cell..

We've had two nvidia chips in two consoles. Both were modified desktop parts . The first one was more modified and it almost cost nvidia their pc busniess.

I don't see nvidia diverting that many resources again for a full custom chip.

 

[PS3MAN]

As I know, PS4 project is in the air now. Cell BE and RSX on PS3 only showing up 20-30% performance on its exclusive way. However multiplatform games are lower than 10% as
we see. If developers can unlock to 60-80% of PS3 performance now. You'll see more clear information of PS4 architect as soon.

Since PS and PS2 ages. I think SCEI may announce some kind of PS4's shadow around 2010
or 2011.

For me. I think my PS4 must be some kind of Cell BE 2 architect with wide dynamic OoOE (8-way issue) 512-bit PPE Core @ 6.4 GHz (Base on Power6 or beyond) with 64 x OoOE SPE (8-way issue) 512-bit register wide/entry/SPE @ 6.4 GHz on-die Rambus Terabyte/sec memory
controller with 8 channal 64 Terabyte Rambus (TB/sec Model)

FlexIO Advance from Rambus to kick 500TB/sec bandwidth overall the system from lower and
high bandwith component.

RAID-0 4x Solid State HDD 500 Terabyte(each) with 64 Terabyte buffer(each)
64x Next generation Blu-ray Drive Technology with 64 Terabyte buffer

GPU nVidia GPU with 64-way SLI and 512 Terabyte Rambas (TB/s Model)
output 11520x4860@120FPS

Oh god! that you all need to pick more than 3000 USD. for PS4 console

 

[archangelmorph]

We've had two nvidia chips in two consoles. Both were modified desktop parts . The first one was more modified and it almost cost nvidia their pc busniess.

I don't see nvidia diverting that many resources again for a full custom chip.

Almost cost nvidia what? where did you get that idea from?

Nvidia have the resources to spare & if Sony have contracted them for a fully custom chip then that's what they will deliver..

Not to mention the fact that Sony will be looking to own the IP rights of the chip & thus it would be in nvidias best interests to design something as specialised as possible so as to not infringe on their own desktop IPs..

 

[archangelmorph]

As I know, PS4 project is in the air now. Cell BE and RSX on PS3 only showing up 20-30% performance on its exclusive way. However multiplatform games are lower than 10% as
we see. If developers can unlock to 60-80% of PS3 performance now. You'll see more clear information of PS4 architect as soon.

Since PS and PS2 ages. I think SCEI may announce some kind of PS4's shadow around 2010
or 2011.

For me. I think my PS4 must be some kind of Cell BE 2 architect with wide dynamic OoOE (8-way issue) 512-bit PPE Core @ 6.4 GHz (Base on Power6 or beyond) with 64 x OoOE SPE (8-way issue) 512-bit register wide/entry/SPE @ 6.4 GHz on-die Rambus Terabyte/sec memory
controller with 8 channal 64 Terabyte Rambus (TB/sec Model)

FlexIO Advance from Rambus to kick 500TB/sec bandwidth overall the system from lower and
high bandwith component.

RAID-0 4x Solid State HDD 500 Terabyte(each) with 64 Terabyte buffer(each)
64x Next generation Blu-ray Drive Technology with 64 Terabyte buffer

GPU nVidia GPU with 64-way SLI and 512 Terabyte Rambas (TB/s Model)
output 11520x4860@120FPS

Oh god! that you all need to pick more than 3000 USD. for PS4 console

It's a good thing Sony won't be building that PS4 for you then huh?

 

[PS3MAN]

Hee hee may be its support build to order like some luxurios car camp.

 

[archangelmorph]

Hee hee may be its support build to order like some luxurios car camp.

Which goes against the entire ethos of what a console is supposed to be..

As good as it may sound to you it's probably not going to happen & if you're really looking for hardware like that i your projected timeframe then you're looking in the wrong places (maybe PC gaming would fit your bill a bit better..)

 

[Mobius1aic]

Well the PS4 I'm sure will have a build upon the Cell BE with Power 6 archy in it, however I can see it possibly having 2 P6 cores, not one that way it's a bit more suited to running "normal" programs made for Power architecture as well as for controlling many more SPEs. On the SPE side, I have a feeling they'll revise the design some, keep it's specialized nature, but make it more user friendly with normal code, while increasing the number of SPEs on the Cell BE 2 die, something like 48 of them. So Power6 core x 2, SPE x 48. I don't see any reason to increase the clock speed, as energy efficiency gets the worst loss in that area of chip design I think.

 

[PS3MAN]

Increase clock speed may produce higher bandwidth.

 

[pascal]

That all sounds about right to me. I think since the tech industry has started to show signs of slowing, and games industry showing us they are better off not launching with very high prices and big consoles (see Japan)

We either have to have 40% less performance in the same time frame. or wait 40% longer for an equivalant performance jump. (ok you dont have to, but this gen has shown us its wise to)

Either way you end up with a console that can be close to 60% of the size of consoles this gen. Which actually feeds quite well in to why I think Sony will extend the PS3 life cycle:

a) they cant really afford to lose another billion so soon for a new console release.

b) Graphics are a commodity not a feature or differentiator. Therefore you will not get the return on investment for making an uber high tech high graphics machine. I still dont think the PS4 will be a slouch though. And actually with some clever GPU design they could do a much better job with the real estate than the RSX (near) off the shelf component ever did.

EDIT:

c) They do have a high def player. MS dont. For me that is a big big big plus.

If I were Sony I would design the PS4 for a cheap, cool and fast type of PS3. I would use the same system and development software with improvments and make certain full backward compatbility.
The video target is still the same 1080p signal then maybe something like:


- Cell 6.4 GHz, 45nm (small, fast and cool)
- nvidia 9xxx compatble
- 1 GB ram
- BD/DVD player
- slim form factor
- 100 watts
- $250

 

[Mefisutoferesu]

What if sony were to scale back the number of cores planned for the PS4 CELL revision and instead increased the size of the LS? For instance, if CELL PS4 was going to be 32 SPEs with 512KB, sony could instead scale it down to, say, 24 SPEs with 1MB LS each. It seems like it'd be a good idea, since it'd - I imagine - cut down on DMAing stuff which many devs seem pretty unhappy with and it would allow the SPEs to have more room to do more work, which I imagine would speed things up, since the SPEs are freaky fast at calculating things WHEN their in the LS, but things seem relatively normal outside that. Yeah, it's kinda counter intuitive to the original design, but console developers don't seem to thrilled the with original design... and granted, it's more a factor of having the LS sized to what's considered reasonable code/data set size... I remember in the presentations way way back Sony talked about the research they did in coming up with 256... I think they said something like 512 seemed to be the sweet spot, but only 256 was reasonable to get the size/performance/price ratio they were after...

 

[Shifty Geezer]

The structure of the Cell memory model means keeping the SPU's supplied with data isn't generally a problem for the tasks that use their processing throughput. To increase LS would mean a significant reduction in available cores and peak processing. It would ease memory access in some cases, probably making some tasks a bit easier to code, but to be honest a lot of work requires big datasets that'll exceed umpteen megs of LS or cache. Buffered data streaming is going to be inevitable. The only real difference in Cell is developers having to manage it themselves rather than letting the cache handle it automatically. Okay some devs don't like that, but on the flip side forcing devs to think about it is good practice for all architectures. As long as you have enough room for code and data enough to keep the core busy while another batch of data is fetched, that's as much LS as you need, and indeed the optimum for performance per mm^2. PS2 managed with tiny little stores! So we may go up to 512 KB LS, if code is too constrained, or if RAM bandwidth can't keep up with the increase in cores. I can't imagine more than that being a win.

 

[PS3MAN]

My PS4 Spec is better choice for STi.

 

[pjbliverpool]

My PS4 Spec is better choice for STi.

Just for my own peace of mind... you were being sarcastic with that spec weren't you?

 

[kyetech]

1 GB ram

That would be pitiful for 2011/12/13, I think the quantity of ram and its bandwidth is one of the easiest ways to improve fidelity and performance. As a result, spending your budget on more ram is a better long term investment than trying to cut corners with it.

Therefore I think we will likely see at least 2 GB ram and pssibly up to 4 GB or somewhere in between , while other aspects of the system will not see the usual performance/spec increases. I still think the PS4 will be a good performer though.

And lets not forget by the time we have the new consoles BR drives will be cheap as chips, so just that in itself will be a huge cost saving for the PS4.

 

[PS3MAN]

Just for my own peace of mind... you were being sarcastic with that spec weren't you?

No. I don't mean that. I think we need CGI like in realtime games. I think PS3 power isn't
enough for. To match CGI like in realtime it may 100-1000 X PS3 Power. May be my PS4 spec can handle it.

 

[StefanS]

Dear Beyonders!

There're some basic standards for posting in these parts of the Console Forum. For those who don't like long reads I'll sum them up in 3 words:

A) Rationality
B) The aim to add signal to the discussion (hopefully even on topic [that earns you extra brownie points ])
C) The ability to listen

If you're not contributing anything meaningful (that also includes responses to some rather "exotic" postings), don't post. It just adds more noise. So please save us the trouble of cleaning up and spanking some ass.

 

[kyetech]

Nice one StefanS.

It was making this thread stupid

 

[Heinrich4]

Whatheaver it will be next generation, I desire frequently see radiosity illumination(not only in demos,tech demos etc) in game at least something like this :

http://www.youtube.com/watch?v=9rZJrdxDkZE&feature=related

Video from this guys: http://www.geomerics.com

http://www.geomerics.com/enlighten-sdk.htm

Very interesting:

" Performance
Enlighten is available on three platforms.

Sony Playstation 3

* Enlighten is currently optimised so that the core algorithm can run sufficiently on 1 SPU, with minimal GPU usage. On a game running at 60fps, Enlighten will consume 1/3rd of the resource of 1 SPU, leaving the remaining SPU resource for the rest of the game. This may of course be increased to add detail to the radiosity solution, giving an exciting amount of scope for creativity.

Microsoft Xbox 360

* For the 360 Enlighten can be configured to balance work load across either the GPU or on one of the main threads. How you choose to distribute the work can then reflect the balance of resource demand in your system.

PC

* The PC version of Enlighten can also be configured to balance work load between the GPU and the CPU. This allows Enlighten to target the most suitable resource across a wide range PCs with graphics cards of differing capacity and those PCs with multi-core CPUs."

 

[Nesh]

Are you saying that this demo is running on the PS3 by just using 1/3 of a SPU and no GPU?

That would be impressive

edit: btw isnt what is shown here what Splinter Cell's developers claimed was not doable on the PS3 and excused their decision for going exclusive on the 360 with Convinction?

If yes then their claim is pretty much debunked.