Predict: The Next Generation Console Tech
- Thread starter Acert93
- Start date
- Status
I think next gen consoles will use a single chip design like Larrabee or a next gen Cell fr cost purposes. PC's will probably still be using the CPU-GPU model at the high end which will offer more raw power but will cost a lot nore.
The big difference between future consoles and last gen ones will be backward compatibility. Because they are multi-core, they will be upgraded in a way that allows full backward compatibility without any redundant components.
PS4 - Cell with 2xPPE, 32xSPE with larger local store, 1024MB RAM, 512MB VRAM, GPU with 3x current performance.
XBOX360NG - 6 core Xenon with more cache, 1024MB RAM, 25MB EDRAM
Input and peripherals will be another big difference. 3D stereoscopic goggles with built in LCD display for each eye and 3D orientation sensing; video cams for user input - eg. eye-cam pupil tracking to aim and to get emotion feedback; voice input; brainwave bio feedback feedback for user emotional input; more sophisticated positional sensors - for example separate glove hand controllers with positional sensors for each finger and hand and positional sensors for multiple body points eg. wrists, elbows, sholders, waist, knees and ankles, allowing full body movement for control; separate rumble feedback units which can be strapped to feet, wrists etc, and multiple options for input units in games, for example games can be played with different controllers, different number of rumble units, position sensors etc. All of these will be wireless, so you don't have to sit in front of the console.
I can also see current consoles being cost, size and low power engineered to turn current models portable or embedded devices that come as portable or embedded in media/entertainment or tablet PC devices, while maintaining backword compatibility, or at least straightforward porting of games from current gen.
PS4 - Cell with 2xPPE, 32xSPE with larger local store, 1024MB RAM, 512MB VRAM, GPU with 3x current performance.
XBOX360NG - 6 core Xenon with more cache, 1024MB RAM, 25MB EDRAM
Input and peripherals will be another big difference. 3D stereoscopic goggles with built in LCD display for each eye and 3D orientation sensing; video cams for user input - eg. eye-cam pupil tracking to aim and to get emotion feedback; voice input; brainwave bio feedback feedback for user emotional input; more sophisticated positional sensors - for example separate glove hand controllers with positional sensors for each finger and hand and positional sensors for multiple body points eg. wrists, elbows, sholders, waist, knees and ankles, allowing full body movement for control; separate rumble feedback units which can be strapped to feet, wrists etc, and multiple options for input units in games, for example games can be played with different controllers, different number of rumble units, position sensors etc. All of these will be wireless, so you don't have to sit in front of the console.
I can also see current consoles being cost, size and low power engineered to turn current models portable or embedded devices that come as portable or embedded in media/entertainment or tablet PC devices, while maintaining backword compatibility, or at least straightforward porting of games from current gen.
Dominik D
Regular
Perhaps optical drives are just not an option anymore? You can buy HDD for as low as $30 and it has more space and better speed than DVD. Back to SNES carts?
Iomega REV disk is $70 for 70GB. Obviously read speeds for those are better than for optical. Will we gat tapes and go back to C64?But seriously: we need price drop on flash media or we'll go back to magnetic storage. Consoles need flash storage for data streaming and memory paging and we need more data delivered by the game. What I guess we'll see (and I for one would be glad to see this change happening) is game prices going up with the amount of content delivered multiplied too. That is a must if we change media we deliver games on. Instead of 140 folks working on 5h long FPS with cool multi, we would have 400 developers on the team working on a customizable game with 20h of gameplay and multi like no other. All delivered on magnetic or flash media. Do we really have to buy 3 WWII games for $60 each? How is it better than one for $200?
I think what we'll see from MS is GPU embracing some of the stuff CPUs do. Virtual memory (memory manager for GPU) is something Carmack's latest engine would benefit from.
I disagree with that. It'll be more in MS's favour to use the 'x86' multicore architectures ofthe time giving developers another easy ride in PC friendly development. I think Xenon was a stop-gap to offer affordable multicore when there were no other viable solutions and never a long-term architecture like Cell. By the next console, Intel and AMD will both have architectures to choose from. AMD will probably do a sweet bundle deal of CPU+GPU, and at the moment that's my guess for the next XB's innards.
Because of of those games offers improvements over the original only possible through an evolution of the design process. Either that $200 game would have to be launched at the beginning of the cycle and so be a 'first-gen title' that's not amazingly fantastical++, or it released later when the developers have had more experience with the hardware. Oh. Which they won't have had. So basically, you'd be constricting games to low efficiency.
Also not everyone wants all three WWII shooters. They'd rather spend $60 n a WWII shooter, $60 on a racer, and $60 on Madden. If they can only have one of those games at $200, no matter how good it is, that'll not offer the diversity of gaming they want.
There have been a few consoles with very expensive 'super experience' games, and they all tanked. People aren't going to spend $200 per title on gaming. Anyone pursuing that model won't last long!
I doubt it very much, for a number of reasons.
Firstly MS got badly burned on the 1st XBox, because of that they want to be able to take the chip to any Fab and get it for a different price. Neither Intel or AMD would allow that, although with AMD going "fab light" they may allow it in a few years.
Secondly MS will want a high end part but at a decidedly low end price, Intel or AMD may be able to supply the part they want but I doubt MS will want to be trying to sell a $1200 console. Xenon is really geared for narrow vector processing, when Xenon shipped there was nothing on the market which could match it, they can only just do it now and even that's questionable as Xenon has optimisations that traditional processors don't.
GPUs have higher peak figures but they're much wider vectors and are much less flexible, that's changing but there are still fundamental differences between them which may not change.
I certainly think Intel and AMD will make a pitch for the contract but I strongly doubt they'll offer the type of thing IBM will be able to do. In fact MS may end up going for an even more custom part and end up with something more like Cell.
One thing I'm fairly certain is that the almost linear extrapolation most people here assume for all components won't hold for the next generation. For example, I find the guesstimates about 4-8 GB ridiculous.
Don't forget the goal of the IHV is not to create the most powerful architecture they can, but the most ballanced one, where every component is just good enough not to hinder the rest of the system.
I think there are two or three fixed numbers for next-gen, and the desire for a ballanced architecture can probably derive the others from them. One is the resolution, 1080p; the second is the memory bandwidth bus - whatever is practical to run over a 128-bit interface in 2009-2010; maybe we can take the XDR of the PS3 and multiply it by 1.2-1.5. The third, uncertain, variable is the media size - or rather, whether there is a physical disc anymore. If it is, it'll be Blu-ray-sized, if not necessarily Blu-ray (Nintendo have proven time and again the non-standard media works, so MS and Nintendo can pick what remains of HD-DVD or even that third contender which arrived recently, which is even more low-tech and DVD-like than HD-DVD, while still providing decent sizes) - on the order of 30-50 GB. The other possibility is that DVD-sized games live on in the form of downloadables.
I think a Cell successor with 4 "normal" cores and 16-24 SPUs can serve as a good evolution path for both Xbox360 and PS3. Add to that something like Xenos, with no more than twice the number of execution units, just enough EDRAM to go by (including tiling for 1080p 4xAA), 1 or 2 GB of relatively off-the-shelf RAM, 16 GB of flash for the cheaper SKUs and a HDD for the more expensive ones.
Don't forget the goal of the IHV is not to create the most powerful architecture they can, but the most ballanced one, where every component is just good enough not to hinder the rest of the system.
I think there are two or three fixed numbers for next-gen, and the desire for a ballanced architecture can probably derive the others from them. One is the resolution, 1080p; the second is the memory bandwidth bus - whatever is practical to run over a 128-bit interface in 2009-2010; maybe we can take the XDR of the PS3 and multiply it by 1.2-1.5. The third, uncertain, variable is the media size - or rather, whether there is a physical disc anymore. If it is, it'll be Blu-ray-sized, if not necessarily Blu-ray (Nintendo have proven time and again the non-standard media works, so MS and Nintendo can pick what remains of HD-DVD or even that third contender which arrived recently, which is even more low-tech and DVD-like than HD-DVD, while still providing decent sizes) - on the order of 30-50 GB. The other possibility is that DVD-sized games live on in the form of downloadables.
I think a Cell successor with 4 "normal" cores and 16-24 SPUs can serve as a good evolution path for both Xbox360 and PS3. Add to that something like Xenos, with no more than twice the number of execution units, just enough EDRAM to go by (including tiling for 1080p 4xAA), 1 or 2 GB of relatively off-the-shelf RAM, 16 GB of flash for the cheaper SKUs and a HDD for the more expensive ones.
One vote for common sense.
Microsoft should continue their PPC road. Strong CPU powered by a very strong GPU except this time they should remove the obvious flaws from the get go.
I really hope Sony is gonna stay on the Cell road and get some payback for the hard times they have now with the Cell adaption. Regarding GPU, it´s easy to say whatever Nvidia has on the tables.. so i´ll do that, just something that is planned a bit better
Microsoft should continue their PPC road. Strong CPU powered by a very strong GPU except this time they should remove the obvious flaws from the get go.
I really hope Sony is gonna stay on the Cell road and get some payback for the hard times they have now with the Cell adaption. Regarding GPU, it´s easy to say whatever Nvidia has on the tables.. so i´ll do that, just something that is planned a bit better
Would it be? With multiplatform development progressing even today and to the end of this generation of consoles, I wonder if they shouldn't just stay with Xenon/Waternoose as the basis for the next gen. It's in the same vein as Sony keeping the Cell architecture.
There are few developers out there that aren't working on the Xbox 360 and many will be working on it for the next few years (the exceptions being Sony-owned etc
). The big guns on PC are already developing multi-platform middle-ware for Microsoft's console. By 2011-12, there will be big-name engines from id software, Capcom, and Epic etc already optimized for a base architecture. I don't think going back to x86 will be as big of a win by then, especially given the proliferation of Unreal Engine 3. I know it's a win when considering new developers, but... if everyone else already has an existing optimized console engine by that time, is it that much better to go back to the PC model 
As nice as it would be for x86 to return to consoles, I have to keep in mind that MS does own the IP designs in the 360. It would make sense to develop that further. Factoring in the royalties and monies being paid out, I would think MS would rather keep going with what they already own.
Have AMD ever been given a real shot at the consoles? IIRC they didn't have the production capacity to be viable (in MS's eyes) for XB. The only reason MS got burned with XB was their stupid contracts that didn't set a pathway for cost reductions at their end. I think from AMD's POV, they'd love to land XB720. It'd make their systems, CPU+GPU, the core of a closed-architecture box meaning lots of optimization, and as such would have some feed-back into their PC space, meaning software improvements developed on XB3000 having a beneficial effect on PC position. And though Xenon will benefit from engines developed for it over this gen, a single x86 system would allow MS to have one developer system rather than need to develop and maintain two - XNA for x86 and XNA for console. The new PC software architectures created to run on the new x86 PCs will be directly applicable to XB5000GTX saving cost to developers of improving existing engines, assuming existing engines aren't
very scalable. And I think that's the case. With 3 cores, it's more straightforward to assign tasks per core and not worry about low-level task management like Cell. Thus where a later Cell engine may just share jobs around available SPEs, scaling up to a 32 SPE system, a Xenon engine would likely only use 3 cores of a 12 core Xenon system (XB360 devs correct me if I'm wrong here and you're virtualizing a lot more on Xenon).
The only negative side to going to an x86 derivative is the cost-reduction, part management thing, which I'm sure some sensible lawyers and eager IHVs could work out to everyone's benefit. The whole concept of the XBox was a box to run DirectX, and taking that forwards, the most sensible box is one using the same hardware as the most widespread DirectX platform, the PC.
very scalable. And I think that's the case. With 3 cores, it's more straightforward to assign tasks per core and not worry about low-level task management like Cell. Thus where a later Cell engine may just share jobs around available SPEs, scaling up to a 32 SPE system, a Xenon engine would likely only use 3 cores of a 12 core Xenon system (XB360 devs correct me if I'm wrong here and you're virtualizing a lot more on Xenon).
The only negative side to going to an x86 derivative is the cost-reduction, part management thing, which I'm sure some sensible lawyers and eager IHVs could work out to everyone's benefit. The whole concept of the XBox was a box to run DirectX, and taking that forwards, the most sensible box is one using the same hardware as the most widespread DirectX platform, the PC.
True. In fact they originally referred to it as the "DirectXBox" which is obviously where the name came from. Xbox was supposed to be the perfect embodiment of a DX8 gaming platform.
Interestingly though Xbox 360 does seem to have moved away from a strict adherance to a particular DX version (to its benefit).
I expect the next xbox will be roughly in line with DX11.
Thanks Alstrong!
If Ms stay with IBM with the cpu, I think they will pursache something more advance than xenon with more core.
At the beginning of the cell project, it seems that IBM wanted to put a lot ppc core all together, while Toshiba wanted a spu only design (which they're currently working on, spurs engine if my memoy doesn't fail again...).
But what would have been the cell if done only by IBM, a xenon? I don't think so (nobody in fact) may be IBM was considering something close to what the larrabee will be.
So if Ms doesn't manage to convince Intel it will be interesting to see what IBm or AMD have to offer.
It also clear that MS will have an tigher silicon budget with their next hardware than they had with the 360. I think especially to cpu and gpu.
A nice to keep show a nice bump in performance should be to grant the hardware with a lot of RAM say 4GB, and huge bandwidth.
GDDR5 would provide +50GB/s on a 128bits bus but Edram will still be needed as a they will have to feed a huge cpu and keep some bandwith for texturing and others operations.
I hope they will aim for 720P so ~30MB which would allow for 4xAA without tiling and so let a lot of liberty to the devs to choose theirs rendering technics.
A top of the class scaler could help there!
Way faster communication between cpu and gpu.
Overall Ms will need clever design to ensure an pretty noticable performance bump.
I don't think that the gpu will include way more transistors than today top of the line gpu, but will be better features wise and more flexible.
If Ms stay with IBM with the cpu, I think they will pursache something more advance than xenon with more core.
At the beginning of the cell project, it seems that IBM wanted to put a lot ppc core all together, while Toshiba wanted a spu only design (which they're currently working on, spurs engine if my memoy doesn't fail again...).
But what would have been the cell if done only by IBM, a xenon? I don't think so (nobody in fact) may be IBM was considering something close to what the larrabee will be.
So if Ms doesn't manage to convince Intel it will be interesting to see what IBm or AMD have to offer.
It also clear that MS will have an tigher silicon budget with their next hardware than they had with the 360. I think especially to cpu and gpu.
A nice to keep show a nice bump in performance should be to grant the hardware with a lot of RAM say 4GB, and huge bandwidth.
GDDR5 would provide +50GB/s on a 128bits bus but Edram will still be needed as a they will have to feed a huge cpu and keep some bandwith for texturing and others operations.
I hope they will aim for 720P so ~30MB which would allow for 4xAA without tiling and so let a lot of liberty to the devs to choose theirs rendering technics.
A top of the class scaler could help there!
Way faster communication between cpu and gpu.
Overall Ms will need clever design to ensure an pretty noticable performance bump.
I don't think that the gpu will include way more transistors than today top of the line gpu, but will be better features wise and more flexible.
Last edited by a moderator:
Dominik D
Regular
I don't agree with this. First of all although working on 3 games in parallel is different than working on one game three times as big as one today (additional level of management) at the same time you have some benefits. Having one technology base utilized by more people is a huge plus: more people, more bugs fixed. Instead of developing 3 engines, you develop one. Middleware market is growing for this exact reason.
There's no reason why huge game cannot be completed on the same schedule a small one would as long as team is proportionally larger.
This is true and that's why low budget games won't die. That's why we have an increased activity of Indie developers. I was referring strictly to the AAA titles. More on that below.
Some specific examples?
This is something I'm not sure about. It's valid for US market, even more valid for Japanese market, but not valid for others. Broadband connection assumption doesn't work for most emerging markets and currently markets with fast connection are pretty much saturated with consoles. In order to grow and achieve 100mln+ consoles sold you need low price tag and a huge variety of cheap titles available. About 75% of PS2s were sold after the price dropped to $200. You need "epic", expensive titles for hardcore gamers and people who love certain type of games and a huge amount of inexpensive ones for the majority of "late" customers. In order to satisfy the first group I believe the way they get developed should change (described previously). The second group mostly cares about volume and less about quality (gameplay experience matter though).
Sorry for dblpost - can't edit own messages.
Great name, seriosly. Xbox 3K became my bet for the next Xbox.
Not entirely true. .NET Framework on the console is something MS would bring to 360 sooner or later anyway. It makes perfect sense: MS invested a lot of resources in .NET and C# despite the huge problems with Windows 200x .NET idea.
Maintaining XNA on both platforms isn't really that much of a problem because a lot of the differences are "hidden" below the CLR. Well, not a problem for XNA, problem for app developers - you get a lot of performance problems using XNA on 360.There's yet another problem with x86: it's a general purpose CPU. In order to make it good for the console AMD (or whoever) would have to strip all the unnecessary extensions. CPU should do the state machine type of logic and number-crunching should be entirely pushed to the GPU. So I believe that with x86 one would go back to early x86 CPUs and strip all the SIMD extensions trying to improve out of order execution at the same time. Multiple simple cores with deep pipe and branch prediction, large cache, separating code and data, CPU even closer to GPU - that's something I would bet on. I don't think anyone will go as multicore as Sony did with PS3. 4 cores seems to be a safe bet.
At the same time I believe consoles would benefit from some of the virtualization technologies of modern x86 CPUs. Back-compat would be a little bit easier (and safer).
You assume that 5 years down the line PC will still be a large gaming platform. I don't think that's our future.
Last edited by a moderator:
I've just read a review of the Phenom the die size is 285MM² @65nm
intel in the time needs 2 143mm² cores @ 65nm but there's more cache and logic in the same area than what AMD offer.
I don't think AMD will be able to offer enought ressources and will have good enough process to provide something valuable for the next Xbox.
In the 360 @90nm
xenon ~170mm²
xenos ~180mm²
edran ~70mm²
= 420 mm²
The next xbox will be launched mostly @45nm (may be the cpu could be @32nm).
But Ms (and others actors) will have less legs for cost reduction due to skrinking.
The combined size of all chip will be tinier than what the 360 offered.
I would bet slightly over 300mm² for the whole package.
intel in the time needs 2 143mm² cores @ 65nm but there's more cache and logic in the same area than what AMD offer.
I don't think AMD will be able to offer enought ressources and will have good enough process to provide something valuable for the next Xbox.
In the 360 @90nm
xenon ~170mm²
xenos ~180mm²
edran ~70mm²
= 420 mm²
The next xbox will be launched mostly @45nm (may be the cpu could be @32nm).
But Ms (and others actors) will have less legs for cost reduction due to skrinking.
The combined size of all chip will be tinier than what the 360 offered.
I would bet slightly over 300mm² for the whole package.
- Status
Similar threads
- Locked
