-
Hey Guest. Check out your NeoGAF Wrapped 2025 results here!
RUMOR: Sony chooses Imagination Technologies for the PS4
- Thread starter Lagspike_exe
- Start date
Gorgon
Member
Pimpbaa said:
I wont either. But there's people who will not esitate to buy the last iPhone model for that price and complain that a PS3 for 499 USD is too expensive. I know one :lol
However the sweet spot is no doubt at 299 USD at launch and I think this is what we are going to see next gen, with a possible second SKU at 399 at most.
Gorgon said:
You cant delete them. But leave them empty and say ou accidentally multiposted. They will be deleted by a mod if he/she strolls through.
I agree in principle. But having skus with various packins like the PS3s is not a big deal to me. Small or big harddrive, its all the same. Game included or not, its all the same.
H_Prestige
Banned
thuway said:
Multiple models isn't a big deal, especially when all that differs is HDD space. Sony puts out the premium models because they have greater margins on them. That 250gb model probably breaks even for them and there's actually a decent number of people buying it.
fortified_concept
Banned
camineet said:
Great bump to a joke thread citing a ridiculous site.
A PC Watch Impress article on new rumours and speculation about Sony's choices on the CPU side.
http://pc.watch.impress.co.jp/docs/column/kaigai/20091224_339258.html
http://kotaku.com/5435384/report-sony-working-on-multi+core-design-for-ps4
Apparently Sony's now decided against Cell, a new design based on modified SPUs and Larrabee. Apparently they want 'a bit more horsepower' than these solutions offer, and are working on a new multi-core design with ease of development in mind.
It seems all a bit outlandish to me, but it's just a musings/rumour article so...take with a grain of salt...
edit - reading the original...or the google translate of the original...and not Kotaku's take, the basic thrust is that CPU plans for PS4 are in a state of flux. That Sony seems to have been turned off Larrabee because of graphics performance concerns, they think it might be a solution for the longer term but not necessarily PS4. He goes on to say there are two other plans on the table - 1) a SPU based design with a modified memory architecture vs Cell (to overcome difficulties devs have had with grasping Cell's memory architecture - the LS vs cache and so on) and 2) a more 'regular' PC-like multicore design which would obviously be easier still for many developers. In the latter case he suggests that the GPU could play host to a lot of general purpose processing that a more highly parallel CPU (like a Cell based design) might otherwise have taken care of.
He says this insight comes from developer interviews and feedback Sony's been conducting on next-gen plans from Summer 2008 onward. In short, the whole picture on their CPU plans seem quite 'shaky' and uncertain right now.
At least that's the gist of what I got from the article..
http://pc.watch.impress.co.jp/docs/column/kaigai/20091224_339258.html
http://kotaku.com/5435384/report-sony-working-on-multi+core-design-for-ps4
Apparently Sony's now decided against Cell, a new design based on modified SPUs and Larrabee. Apparently they want 'a bit more horsepower' than these solutions offer, and are working on a new multi-core design with ease of development in mind.
It seems all a bit outlandish to me, but it's just a musings/rumour article so...take with a grain of salt...
edit - reading the original...or the google translate of the original...and not Kotaku's take, the basic thrust is that CPU plans for PS4 are in a state of flux. That Sony seems to have been turned off Larrabee because of graphics performance concerns, they think it might be a solution for the longer term but not necessarily PS4. He goes on to say there are two other plans on the table - 1) a SPU based design with a modified memory architecture vs Cell (to overcome difficulties devs have had with grasping Cell's memory architecture - the LS vs cache and so on) and 2) a more 'regular' PC-like multicore design which would obviously be easier still for many developers. In the latter case he suggests that the GPU could play host to a lot of general purpose processing that a more highly parallel CPU (like a Cell based design) might otherwise have taken care of.
He says this insight comes from developer interviews and feedback Sony's been conducting on next-gen plans from Summer 2008 onward. In short, the whole picture on their CPU plans seem quite 'shaky' and uncertain right now.
At least that's the gist of what I got from the article..
This article does a fairly good job translating the details as well actually: http://www.andriasang.com/e/blog/2009/12/25/ps4_cell/gofreak said:
You should probably make this a new thread though. This is actually the second or third unrelated Sony hardware rumor thread this has popped up in, and it probably will keep coming up in more until it gets its own thread. :lol
Edit:
Joystick's take on the article's translate also got a quote mentioning that a Larrabee design is still being considered for the PS5 and PS6, but due to the timeframe they want to release the PS4 in, newer versions of Larrabee won't be ready: http://www.joystiq.com/2009/12/28/ps4-larrabee-sony-will-reconsider-for-ps5/
This meshes well with the speculated 2012 release date in the Andriasang translation.
Panajev2001a
GAF's Pleasant Genius
gofreak said:
The updated CELLv2 design + regular PC GPU (ATI or IMG, I doubt they will work with nVIDIA for PS4) could be interesting... but maybe a bit of a big R&D project for Sony to commission (CELLv1 was co-developed by 3 large corporations, this CPU would really be PS4 only...)... as it would have to consider plans for the LS that can both offer ease of development as well as not make BC with PS3 software too hard.
If per title emulation is acceptable and they can get a good PS3 emulator up and running (PS3 titles would still look good on PS4 while many PS2 titles do not look incredibly well running on a PS3) I still see the multi-core CPU + flexible GPU as the better approach...
hence...
Modern x86 quad-core CPU + LRB based GPU (the fact that LRB1 for PC has been canned as a discreet GPU does not mean console plans have been canned too...). If Intel can keep working well on LRB's roadmap... the fact that some HW issues + SW stack issues hurt its launch as PC GPU... well they won't matter that much.
HW wise: PS4 is some years away and SCE could ask Intel to push a bit harder on the manufacturing process side (instead of being one manufacturing process behind the Desktop CPU side)... plus there could be other improvements to LRB's HW architecture (maybe full Pentium ISA compatibility could be altered and a faster yet leaner x86 core could be used instead).
SW wise: PS4 does not need a fully working up-to-date DirectX and OpenGL stack... early PS4 software could use Intel's developed Software Renderer and a PS4 optimized OpenGL stack... PS3 compatibility would only need a modified DX9 stack targeted to RSX's quirks.
Generally the other reason I don't advocate posting new articles in threads like these is it just gets people talking about the OP again, as opposed to the new article, which is why we have seen this article pop up in a few other Sony hardware threads as well, since most people miss them after the bump happens.
NemesisPrime
Banned
Goldrusher said:
Hmmm that made a lot more sense than I thought. You might have a point there.
D
Deleted member 30609
Unconfirmed Member
that's all well and good, but 2012 is too early. I'm happy with my PS3 for a fair while yet.
D
Deleted member 30609
Unconfirmed Member
oh yeah, 12 is conceivable, I just don't want it to be so. =p
Early release (well, early in the sense that this generation doesn't seem to have moved toward its potential as quickly as others did) is a really tough call this time around though. I'd say there's a real chance of the general public not picking up on a new system if it's out too soon - it really feels like most people out there are only just starting to get rid of their PS2s and jumping on board current gen systems.
I have no basis for that statement, just FYI. Just anecdotal blather.
I have no basis for that statement, just FYI. Just anecdotal blather.
Panajev2001a
GAF's Pleasant Genius
avaya said:
PSP2 will be SGX based though...
.Software support in a closed box is not that problematic, they are not trying to provide a WDDM or X.org driver, but something co-developed alongside the entire PS4 software stack and that some developers will access at a low level. Still, maybe their solution scales better on a portable system rather than a big beast like PS4 (the MIMD approach they take with SGX might be a bit excessive when dealing with a very large number of cores/execution units needed in a high-performance part). So, we agree on that kind of
.Again... fast task oriented x86 CPU with OOOe cores + throughput/data oriented LRB based GPU hub + single and fast shared memory pool == PS4... IMHO.
Chuck Norris
Banned
Sure, if it's an incremental upgrade (just graphics) that costs too much like PS3 was. But surely besides graphics they will be going for something innovative, whether it's controls or online. Graphics alone won't cut it anymore.alistairw said:
I don't see any of the Big 3 starting at much more than $299 (if at all) next gen.
Glad to hear Sony is looking at ease of development as a focus next gen. It bit them hard this gen.
EDIT:
And I feel like you BUT I would've said I'm not ready at the end of 2003 for next gen. By 2005 though I was more than ready. Who says you won't be in 2011 or 2012?
Lagspike_exe
Member
Chuck Norris said:
Not really. The only major problems were back in 2005. and 2006. and those were only with multiplatform games. From 2007. multiplatform was ok, and 1st party was pupming out some seriously impressive games - graphics wise.
Panajev2001a said:PSP2 will be SGX based though...
Software support in a closed box is not that problematic, they are not trying to provide a WDDM or X.org driver, but something co-developed alongside the entire PS4 software stack and that some developers will access at a low level. Still, maybe their solution scales better on a portable system rather than a big beast like PS4 (the MIMD approach they take with SGX might be a bit excessive when dealing with a very large number of cores/execution units needed in a high-performance part). So, we agree on that kind of
Again... fast task oriented x86 CPU with OOOe cores + throughput/data oriented LRB based GPU hub + single and fast shared memory pool == PS4... IMHO.
Can't say I disagree with that but I stil don't like the idea of legacy x86 wasting silicon. A Nehalem core has 10-15% of the space dedicated to it. How much of that would they be able to stip out without making it too alien?
KAL2006 said:
Exactly i want a handheld with some stuff like an Iphone. Ya know how easy to get apps in iphone' install them and sycn with your computer. They should release something like an Iphone with better gpu cpu and more control options than a touch screen. (PSP GO sounds like that but still not very accessible as Iphone) DO it like APPLE!
Chuck Norris
Banned
It's improved, but multiplats still show inferiorities because PS3's main capabilties were focused on things that are rarely utilised outside of exclusives, like Cell for graphics processing.Lagspike_exe said:
At worst we still get ports like Bayonetta
Chuck Norris
Banned
I'm hoping Sony can find an interesting way to mix a multi-touch interface with a 3D capable screen.KAL2006 said:
Panajev2001a
GAF's Pleasant Genius
avaya said:
I do not think they would take much of that out, but they certainly could if they decided to design an x86-64 (x64) only CPU removing the 32 bit support completely... They could remove the x87 unit and the support for it completely and simplify the decoders.
Such a CPU could still run Windows x64 probably (minus any application that has legacy x86 code).
It would be an interesting exercise to have a lean and mean x64 only CPU, but would the engineering effort be worth the R&D costs? Wouldn't it be better to waste a bit of die space on transistors that would be idle anyways (they would not consume power), but to save custom R&D work on the custom bus and UMA solution as well as the LRB based GPU?
My idea is this:
PS4:
-CPU: Quad-core Sandy-bridge CPU with AVX extensions (out in the PC space in 2010, shrunk and power optimized on the "Tick" phase of the design cycle Intel uses... "Tick" being "same" micro-architecture on newer manufacturing process while "Tock" is new micro-architecture on the same manufacturing process as the one used for the "Tick" phase).
--the goal of the CPU is a fast, low latency optimized processor good to manage a few number of independent tasks really quickly (OS, Sound, Networking, JIT compilation, etc...).
-GPU: next-generation LRB based architecture (graphics, physics, A.I., rendering, video decoding, etc...).
--the goal of the GPU would be a data/high throughput oriented/heavily multi-threaded processor flexible enough to be adapted to a high variety of data parallel tasks.
--the GPU would also be the memory hub for the console.
-CPU-to-GPU link: very high-speed two-way I/O link. PS3's suffers IMHO from having an unbalanced design in this regard (the CELL CPU can read/write to the GPU's memory pool at a very slow speed).
-RAM: the challenge would be to have a very fast CPU-to-GPU link while avoiding GPU-to-RAM bus saturation with CPU related traffic if the CPU is feeding from main RAM. Keeping as much work done in LRB's cores L2 memory as possible (without hitting main RAM) would be a big optimization goal for most game developers as a 512 bit wide super-fast memory interface could be a bit costly for a game console which will likely need at least 2 GB of RAM.
--considering an UMA design 2 GB would be the minimum target (only a 4x increase over what the current generation of consoles uses). It is true that the target resolution for PS4 will likely be 1080p so we won't have the big 640x480 -> 1920x1080 resolution jump some PS3 game developers have been facing compared to their previous generation's efforts. 4 GB would be the recommended target IMHO (an 8x increase in RAM capacity does not sound implausible... PS2 == about 40 MB of RAM vs 512 MB on PS3, so more than 12x the RAM in PS3 compared to PS2).
How complex would PS3 titles emulation be? IMHO they will go for per-title optimized emulation, with constant work on the PS3 emulator and emulator packs for titles. Titles available on PSN would be the first targets for the PS3 emulator and the compatibility packs, the disc based releases would follow based on sales IMHO. Hopefully the CPU+GPU combo would be up-to-the-task
.Only thing I would want over that would be Haswell instead of Sandy Bridge. Might mean a 2013 release but it would Intel's leanest and meanest. If they go full Intel then I don't think Chipzilla would baulk at the idea of a quad-core Haswell (Haswell is 8 by default on 22nm). Haswell will have brand new power saving features built in which would be the biggest attraction in addition to new cache design, don't know if Haswell's FMA on the CPU would be that beneficial in your design though.
EDIT: I think any R&D work that reduces chip size is never going to be a waste, since it will payoff immediately, especially with process innovation entering a possible stall post 22nm.
Also 4GB XDR-II?
EDIT: I think any R&D work that reduces chip size is never going to be a waste, since it will payoff immediately, especially with process innovation entering a possible stall post 22nm.
Also 4GB XDR-II?
Jason's Ultimatum
Member
I think what I want most as far as graphics goes is realistic lighting, like some awesome CG lighting we currently see in games. That would make me happy.
Lagspike_exe
Member
Jason's Ultimatum said:
Well, with things we're seeing right know on this hardware, if either Sony or MS goes high/mid end in 2012. you may just get stuff pretty close to what you want. If that happens, hardware performance would stop being the biggest problem and dev costs could be a much more serious one.
So I guess this means PS3 emulation isn't happening?
I've heard everything from PowerVR to Larrabee, and more of an emphasis on attracting programmers who are more used to x86 architecture.
I've spent a lot of money on PS3 games, and wouldn't appreciate it if I am denied BC once again. It's already a damn shame that Sony couldn't figure out a way to keep a ~$30 EE-GS chip on the PS3 for everyone.
But anyway, a decision to use PowerVR would be... strange.
I've heard everything from PowerVR to Larrabee, and more of an emphasis on attracting programmers who are more used to x86 architecture.
I've spent a lot of money on PS3 games, and wouldn't appreciate it if I am denied BC once again. It's already a damn shame that Sony couldn't figure out a way to keep a ~$30 EE-GS chip on the PS3 for everyone.
But anyway, a decision to use PowerVR would be... strange.
Shao Kahn Brewing a Stew
Banned
Just release something that:
1. is easy to develop for
2. has backward compatibility
3. is as awesomely designed and plug-n-play as PS3
4. is on par with PC gaming (atleast till 2012)
5. partners up with Valve and integrate Steam as their default online infrastructure*
and I'll be happy!
*
1. is easy to develop for
2. has backward compatibility
3. is as awesomely designed and plug-n-play as PS3
4. is on par with PC gaming (atleast till 2012)
5. partners up with Valve and integrate Steam as their default online infrastructure*
and I'll be happy!
*
which will NEVER EVER EVER happen 
proposition
Banned
If Sony learns from their mistake of shooting for higher performance at the expense of ease-of-use (...for the second generation in a row) then that can only be a good thing for consumers.
shagg_187 said:Just release something that:
1. is easy to develop for
2. has backward compatibility
3. is as awesomely designed and plug-n-play as PS3
4. is on par with PC gaming (atleast till 2012)
5. partners up with Valve and integrate Steam as their default online infrastructure*
and I'll be happy!
*which will NEVER EVER EVER happen
Why would anyone want Steam on a console? Oh and all consoles in history are plug and play AFAIK.