Considering AMD gets better performance per Watt and seems to offer better licensing deals, I can't see why Sony would pick nVidia over AMD given a choice.
Granted, they will want to have a different deal with Nvidia to control costs, but backwards compatibility in this digital age is VERY important. All the content that has already been purchased by ps3 owners would be wasted if the architecture is too different and incompatible. Granted, Sony could go the route of MSFT and try to emulate the GPU, but I don't think the AMD advantages are enough to make up this cost.
Yes, they would technically be better off going with AMD, but there are business reasons to stick with Nvidia that outweigh the technical advantage of AMD designs.
For all parties involved, the bottom line will be the driving force of next gen consoles.
As long as Nvidia can get them close to AMD performance (within 15% or so) then they will be the choice to carry the existing ps3 user-base.
I believe a G71 without any exotic high bandwidth link such as EDRAM can easily be emulated by a 5770, which will surely be eclipsed by the next console generation.
On the other hand, the EDRAM in the current 360 will make EDRAM in the next xbox mandatory if they want to keep BC.
I believe a G71 without any exotic high bandwidth link such as EDRAM can easily be emulated by a 5770, which will surely be eclipsed by the next console generation.
On the other hand, the EDRAM in the current 360 will make EDRAM in the next xbox mandatory if they want to keep BC.
Technically it shouldn't be an issue, but Sony would have to pay for Nvidia to open their architecture for AMD engineers to emulate it as MSFT had to for their xbox1 emulation.
I think we get the point here, cause we could believe BC will be an important factor for next generation with many developers have spent a lot of money and probably will not want to invest in the same level of the order of 10 times more (ps2/xbox average -> US$ 2/3 million to US$15/30million in ps360) like transition occurred in the past generation(today consoles->ps360) for next gen. In fact next generation consoles first games should be (as always..) something like Version 1.5 which and we have to wait at least 18/24 months(almost 2 years) to see engines "pure next gen".
Perhaps today's games are now being created or even thought to be "post compatible" (or something like that), it reminds me a stage 1998/1999 where games like Dragon Quest VII and others for psone received some effects that were better handled in "smooth mode"(mip mapping) in ps2.
About gpu "PS4"... sony if I remember correctly they have closed a partnership with the creators of the PowerVR?
Technically it shouldn't be an issue, but Sony would have to pay for Nvidia to open their architecture for AMD engineers to emulate it as MSFT had to for their xbox1 emulation.
Just like DX11 graphics cards are also DX9 compatible, all ATI/Sony engineers need to do is to make sure the new GPU implements all the libgcm/psgl functions at least as fast as the G71, which shouldn't be too hard. I don't think the whole chip needs to be emulated. They could even replace the QAA calls with 4x/8x MSAA calls. Changing rendering resolution for PS3 games or increasing framerate would probably not be possible due to Cell's involvement in the rendering pipeline, but AA and filtering should be easy improvements only running on the GPU.
Sorry all this off topic, but this interesting article(Hiroshige Goto Article about x360 cpu/gpu integration)* deserves to be mentioned and we can go further in imagining the challenges sony and microsoft will have to eventually integrate its GPU and eDRAMs and that will impact if they adopt for next generation.
Even with the eDram as a separate chip on the GPU package it's still got to work out cheaper than a much wider bus and additional memory chips though, surely ...?
And i think most would consider Sony's lack of BC a huge mistake given their dominance the generation prior to ps3.
#3 Illustrates exactly why Sony (and MSFT) need to implement BC in their next gen consoles. Not merely a "it works for some titles" but a full transfer of purchased digital content to the new machine. Breaking this compatibility breaks any ties of momentum either company built over the past 5 years and discourages ANY future digital purchases.
The other reason for BC is reduced development costs and reduced time to market. Sony or MS can bring up-scaled versions of their consoles to market in a fraction of the time it would take to develop a bottom up brand new console.
The benefits are also there for content creation. Getting games up and running quicker and hitting the shelves for purchase to extend profitability. Building on the investments which were already in place for both consoles will pay off for both companies and are a lot easier for investors to swallow than starting all over from scratch.
Your post may have just been limited to the GPU issue, but it stretches further than that. The only reason Sony will abandon Nvidia IMO is if Nvidia kills the deal themselves. They are technically in the ballpark with AMD and any deficiency there can easily be made up with Cell. All the while, they have full BC with every PS3 game and digital content generated for PS3.
Sony may pressure nvidia for better pricing by shopping AMD, but its truly in both Sony's and Nvidia's interests to work out a deal.
And i think most would consider Sony's lack of BC a huge mistake given their dominance the generation prior to ps3.
#3 Illustrates exactly why Sony (and MSFT) need to implement BC in their next gen consoles. Not merely a "it works for some titles" but a full transfer of purchased digital content to the new machine. Breaking this compatibility breaks any ties of momentum either company built over the past 5 years and discourages ANY future digital purchases.
I really don't think that matters. I don't play any of the Spectrum games I bought, nor the Amiga games, nor the Sega Master System games, nor my PS1 games, nor PS2, nor my PC games of yore, nor any of my PS3 games that I've finished. I don't care if my current content doesn't work on PS4 because I'll be playing even better, new content!
I'm sure BC has been considered at length on this board before. Indeed, here's a poll of the time that showed only 1/9th would not have bought a current-gen console if it lacked BC, and of 104 voters, 2 said they wouldn't buy a console without BC. A search throws up lots of other threads about BC, so I don't think this one needs be derailed, although if the points need be thrashed again, there's nothing stopping a new thread being made!
While this may be true, my point was in response to assen's WRT older content brushed up and sold as new. I seriously doubt people will be "happy" to re-up and again purchase the same old content for ps4/xb720 which they already purchased on ps3/xb360.
While some may be happy to buy a download only game only to play it once and forget it, I think most would like the ability to use that content as they wish after having paid for it. Otherwise, a rental model is more fitting for that one and done purpose.
In contrast to your behavior, most PS3 owners I know used their launch ps3's as a wireless ps2 player and bluray player for the majority of the first couple years.
BTW - your poll showed 65% usage of BC on the small sampling of B3D users. With emphasis this gen put further in online interaction giving software longer legs than usual, I'd expect this percentage to be higher next gen as more and more users take their experience online with multiplayer.
In addition, I'm not sure if you read my theory on next gen content, but I think if the software is written right, we may see forward compatible games similar to what we see in the PC space, but potentially more improvement from the baseline to the top end.
Just my opinion on where I see things from a business perspective - Money/Economy is leading technology direction for the next gen (note lack of Kutaragi, success of Wii, losses by MSFT and Sony, soft World Economy).
Performance will be compromised to protect existing revenue streams and investments (BC, building on existing architecture) and limit liability in uncertain economic markets (content creation, die-size, heat, failure).
If times were different, we may see more aggressive designs. The extension of time this gen is going some ways to make up the difference, but on the other end, the limited ability to die shrink going forward is also putting a cap on projected savings through die shrinks.
I really don't think that matters. I don't play any of the Spectrum games I bought, nor the Amiga games, nor the Sega Master System games, nor my PS1 games, nor PS2, nor my PC games of yore, nor any of my PS3 games that I've finished. I don't care if my current content doesn't work on PS4 because I'll be playing even better, new content!
While I expect this is true for the most part for most people, there are exceptions, such as the way the Halo 2 was the major draw for Xbox Live on the 360 for quite a while. In this case, not only did MS continue selling Halo 2 itself, but it helped move subscribers over to the new system with an upgraded Live experience where they could be milked even harder.
Once you move away from considering games as a physical thing you stick into a physical console, and look at it as content you've paid for that is linked in some way to a service that you pay for and rely on, then BC becomes a slightly different beast.
The transition between generations is where battles and billions are won and lost, and being able to take hugely popular games with ongoing relevance (thanks to online components) across the generational gap may prove to have value next generation as it did this.
I wouldn't rule out MS (or even Sony) keeping an eye on the possibility of BC while they're designing their next gen systems.
While I expect this is true for the most part for most people, there are exceptions, such as the way the Halo 2 was the major draw for Xbox Live on the 360 for quite a while. In this case, not only did MS continue selling Halo 2 itself, but it helped move subscribers over to the new system with an upgraded Live experience where they could be milked even harder.
Once you move away from considering games as a physical thing you stick into a physical console, and look at it as content you've paid for that is linked in some way to a service that you pay for and rely on, then BC becomes a slightly different beast.
The transition between generations is where battles and billions are won and lost, and being able to take hugely popular games with ongoing relevance (thanks to online components) across the generational gap may prove to have value next generation as it did this.
I wouldn't rule out MS (or even Sony) keeping an eye on the possibility of BC while they're designing their next gen systems.
However, I think as time goes on, BC becomes even more and more relevant. In much the same way app stores allow users to transfer apps to new phones, consoles need to adopt the same concept. It's an expected feature that most every other medium recognizes. BR is backwards compatible with DVD, Windows keeps their users on this model, Ipod songs etc.
Content that is transferable to at least the same brand branch is expected by consumers. Nintendo gets this and has had a firm grip on the handheld space, generation after generation by following it.
Of course there are always limitations to this. VHS tapes don't work in DVD players, but whenever there is a logical and consumer friendly way to allow BC, it should be done. (see CD's to iPod)
Now all of these examples go out the window if there is a compelling reason to abandon BC such as a vastly superior but incompatible architecture which explodes onto the scene.
Baring this though, extending architecture makes more business sense than scraping it for the sake of having ~10% more performance.
If Sony goes with Nvidia again I'm predicting it will be a much more custom product this time that has evolved from some serious co-research and production. There was no time for that last time, but that's what I would be betting on this time. Nvidia also has a strong drive of 3D technology in common with Sony ...
I personally think that the importance of backwards compatibility is waning, and we'll see more optimised remakes instead. We'll see ... In a few years.
If Sony goes with Nvidia again I'm predicting it will be a much more custom product this time that has evolved from some serious co-research and production. There was no time for that last time, but that's what I would be betting on this time. Nvidia also has a strong drive of 3D technology in common with Sony ...
It would make sense still I feel 3D is so over rated it's not even funny. One have to wear glasses and it's not OK by it-self. Might be ok for a special / impressive / all out movies but that's it.
And how about the others people in the room? I remember playing in the living room when I lived at my parents place, I can even think of their face if on otp of having to suffer watching me playing if the screen had been on top of it a "mess"... Really 3D is a geek/high tech fan thing.
I personally think that the importance of backwards compatibility is waning, and we'll see more optimised remakes instead. We'll see ... In a few years.
I agree on top of it nowadays Sony and MS have informations about the games solo or MP that still get usage long after their release. They could launch with software emulation solution ready for a handful of important games.
An extension of this thought of what I expected in next gen entertainment:
In looking at the strides that were made from last gen to this gen, Art asset creation is IMO at near sufficient levels currently. Sure there are improvements that could be made, but for the most part when artists are modeling multimillion poly models and crunching them down with normal maps I think this level of detail is sufficient for next gen.
Indeed assets are already there. Adding tessellation and displacement mapping on top of normal map could a long way in improving in-game models
Asset creation in the environment is another issue, but detail is again very good. Improvements could/should mostly come from procedural generation for smaller details.
This puts a cap on content creation costs which as we know, the industry is extremely concerned (see EA losses etc).
With that in mind, how would the next gen be differentiated over current gen?
Image quality of these existing assets.
Lighting, shadow, resolution, AA, and procedural generation detail.
None of these are asset creation additions so the cost is fixed squarely in the corner of hardware.
Another potential boon for this model, games could be made essentially the same for next gen as current gen. In fact, it may be possible to ship the same exact game on current gen as next gen and have the next gen version look markedly better than the current gen just through improvements in how it handles lighting, shadow, resolution, and AA.
Keys to making it happen:
Procedural generation detail- cpu burden of this technique for multiple detail requirements will be higher. How much will depend on how far they go with detail obviously, but the more processing available, the better. Uses are not only textures, but geometry, and animation as well.
The other image quality improvements are obviously gpu limited and in many ways, 3d rendering engine related.
Quiet some devs seem to want easy, low latency communication between the CPU and the GPU.
By 192 stream processor you means 12 64wide SIMD array, right? Something in between the HD5770 and HD5830 (minus based on up to date technology).
Some raw numbers:
Valhalla: 168mm² Combined CGPU + 45mm² ED-Ram = 213mm² total
Vs
Jasper: 135mm² CPU + 121mm² GPU (256mm^2) + 64mm² ED-Ram = 320mm² total
So I grossly assume a 50% split in CPU and GPU.
Grossly the thing would be using @45nm process:
168mm² for CPU (a x2) + <200mm² for the GPU (gross estimate based on hd5770 die size,remove RBE, add SIMD array, improvements, and a less dense process)
=> < 368mm² => @32nm x0.7 => < 260mm²
So it may be possible to fit everything on a single die. Actually depending on power and clock consideration they could have less SIMD array running a bit faster. It sounds reasonable.
Till I think 4 cores would be enough as long as they are "better" with wider SIMD units (say 8 wide matching X86 counter parts).
They should also consider make the Edram chip more versatile (ie Shaders could read and write it, by passing or not the RBE).
Overall I feel like Ms should go with whatever AMD offers them within their transistor budget. In the same time I would like to see them listen to Epic again and say to hell to any fixed function hardware no matter how fewer pixels they push in the end.
yes I'm re-reading "End of the GPU roadmap" from Sweeney
Sony:
cell2 - 1 ppe, 16spe 512kb cache/local store each
2gb ram
Nvidia off shelf dx11 300mm²
Honestly depending on how PC like the GPU's become I could see them preprogramming upgrades. A resolution bump some better AA/AF maybe so better shaders and alpha could all be put in a PS3/360 game with a flag to enable them for PS4/Xwhatever generation.
I wasn't sure if ask here or on the edram actual utilization thread
An ignorant question from an ignorant person:
Considering the gpu evolution, and the shift to shader driven AA and such, would edram still make sense in a future console?
And with this i suppose that neither bc will be a problem because ati would be able to emulate this in some way, maybe not producing exact images, but changing the aa method with one that has little or no cost on the next gpu
