Predict: The Next Generation Console Tech

[Heinrich4]

Maybe this new Fox Engine going to next generation console like capcon Framework MT coming from ps2.


http://www.next-gen.biz/news/kojima-unveils-multiplatform-fox-engine

video:

http://www.youtube.com/watch?v=HjFWLEBFlOc&feature=player_embedded


Off topic here...Sorry all

I see video and have HP Z800 servers...chipset..intell 5520CH* with Geforce 8400GS** same 8500GT specs.(i see x360 controllers but hardware maybe could be Z800 for showing all platforms)

*
http://techreport.com/articles.x/16656/4

**
http://techreport.com/discussions.x/12699

 

[rpg.314]

GPU is only suitable for large scale computation tasks. You need large scale calculations that can be scaled to hundreds of GPU threads to get peak performance out of the system. You cannot for example instruct the GPU to multiply one matrix with other, and get the result back in less than 100 CPU cycles, like you would with CPU vector instructions. GPUs are deeply pipelined, and run asynchronously. Synchronous CPU-GPU calculation would cause the GPU to run idle most of the time.

CPUs do a lot of vector math in games. You need to do stuff like determine the geometry that is visible in the viewport (viewport culling), setup the object matrices for rendering (camera transform), apply animations and bone transformations for skinning, determine collisions between objects, simulate/apply physics to game objects, do ray casting (for various reasons). Many AI algorithms also need lots of vector processing (AI is interacting in a 3d world after all). Some of these tasks can be offloaded to GPU, but many tasks remain that either require too much logic (too much branching and data structure traversal), require immediate results (GPU calculation has basically one frame latency) or cannot be grouped to large tasks that run hundreds of identical calculations at the same time (GPU calculation model).

Surely, if one can make these jobs fly on SPUs, one can make them work on a GPU, no?

Note that the cpu<->gpu link will be just as low latency on next gen console as the ppu<->spu link on cell today.

 

[3dilettante]

3dilettante, what are the inherent advantages that make ARM more suitable than POWER in devices ?

I'm not familiar enough the embedded market to say anything definitive.
From a very high-level view, I do not think there is any real advantage.

ISA is a second order consideration for performance.
The TDP range for consoles is comfortably high enough that any power advantage or disadvantage is going to be lost in the noise.

ARM has a number of disadvantages. While it has Neon in the ISA, it is only recently really being treated as standard. IBM's experience with FP and SIMD extensions is much longer and has a history of good results compared to the patchwork results for ARM.

Both ISAs are not new, and each has cruft. ARM is able to shift with each instruction (or many variants do, I'm not sure how the new lines behave), and can use predication on all its instructions.
IBM has a more complex memory translation method, and POWER does do cracking of its more complex instructions.

ARM is not known for high-performance system interfaces (limited bus and memory bandwidth are chronic bottlenecks for even the strongest ARM), while IBM is one of the best.
IBM Semi has some of the best engineering available for this task, and IBM has had a business designing custom designs for several console generations.
ARM does do some design work for its IP, but I haven't seen enough to say if it would do what IBM does. I have not seen anything that would say ARM can match the engineering effort IBM Semi is willing to bring to the table. In addition, IBM can fab chips as necessary, while ARM does not.
IBM has an additional incentive in that it has this large amount of semiconductor capability it has trouble utilizing, so is willing to sell its services. ARM is plenty busy as it is.

 

[Rangers]

If Apple were to enter the console space in the next couple of years what kind of hardware do you think they'd include in the box?

And before someone mentions it, no the Bandai Pippin wasn't a very serious effort.

I've wondered about a Samsung console. One trend that seems to be consistent over the last decade is the rise of the Korean companies and the fall of the Japanese ones, see Samsung V Sony or Hyundai V Toyota. I'm a big phone aficionado and Samsung has done extremely well in that business. As in pretty much any other business Samsung is in.

I dont see an Apple console, Xbox and PlayStation seem like everything Apple hates, aka good products, OK I kid, male oriented, high powered, hot, low margin, devices with lots of flaws lets say. If there was an Apple console strategy I would guess it would be basically, allowing the Apple TV to play crappy iPhone games on your TV, or something organic like that.

 

[Gubbi]

While it has Neon in the ISA, it is only recently really being treated as standard.

And not even that well. The Cortex A9 ROB doesn't handle data dependencies between NEON instructions, which means it stalls on RAW hazards. This is probably the reason Nvidia killed off NEON support when the Tegra 2 die-diet set in, the performance just wasn't there.

Cortex A15 can handle the data dependencies and issue NEON ops out of order.

Cheers

 

[rpg.314]

I'm not familiar enough the embedded market to say anything definitive.
From a very high-level view, I do not think there is any real advantage.

ISA is a second order consideration for performance.
The TDP range for consoles is comfortably high enough that any power advantage or disadvantage is going to be lost in the noise.

ARM has a number of disadvantages. While it has Neon in the ISA, it is only recently really being treated as standard. IBM's experience with FP and SIMD extensions is much longer and has a history of good results compared to the patchwork results for ARM.

Both ISAs are not new, and each has cruft. ARM is able to shift with each instruction (or many variants do, I'm not sure how the new lines behave), and can use predication on all its instructions.
IBM has a more complex memory translation method, and POWER does do cracking of its more complex instructions.

ARMv8 could change the status quo wrt cruft.

ARM, would likely go in, if it does at all, at the hands of an architectural licensee.

 

[3dilettante]

That means Microsoft may have the option to do so.
There could be some monetary advantages to having the design in-house, but the ISA isn't the hard part.
A screwup would be more costly, and the RRoD was rumored to be a case where Microsoft tried to go it alone and wound up getting outside help in the end.

 

[rpg.314]

That means Microsoft may have the option to do so.
There could be some monetary advantages to having the design in-house, but the ISA isn't the hard part.
A screwup would be more costly, and the RRoD was rumored to be a case where Microsoft tried to go it alone and wound up getting outside help in the end.

MS's architectural license is the key puzzle here. They took the license in 2010. They had no expertise for it before hand and it will take a huge team recruitment or acquisition to get it started. We haven't seen signs of either one.

 

[liolio]

Just an iffy affirmation whatever the system BOM from all the news we got Ms is the only manufacturer that may invest a lot in R&D, right?
Sony've just been clear about it. It doesn't seem N went crazy either with the Wii2, so that lets MS if they think it's worse it.

 

[patsu]

I'm not familiar enough the embedded market to say anything definitive.
From a very high-level view, I do not think there is any real advantage.

ISA is a second order consideration for performance.
The TDP range for consoles is comfortably high enough that any power advantage or disadvantage is going to be lost in the noise.

ARM has a number of disadvantages. While it has Neon in the ISA, it is only recently really being treated as standard. IBM's experience with FP and SIMD extensions is much longer and has a history of good results compared to the patchwork results for ARM.

Both ISAs are not new, and each has cruft. ARM is able to shift with each instruction (or many variants do, I'm not sure how the new lines behave), and can use predication on all its instructions.
IBM has a more complex memory translation method, and POWER does do cracking of its more complex instructions.

ARM is not known for high-performance system interfaces (limited bus and memory bandwidth are chronic bottlenecks for even the strongest ARM), while IBM is one of the best.
IBM Semi has some of the best engineering available for this task, and IBM has had a business designing custom designs for several console generations.
ARM does do some design work for its IP, but I haven't seen enough to say if it would do what IBM does. I have not seen anything that would say ARM can match the engineering effort IBM Semi is willing to bring to the table. In addition, IBM can fab chips as necessary, while ARM does not.
IBM has an additional incentive in that it has this large amount of semiconductor capability it has trouble utilizing, so is willing to sell its services. ARM is plenty busy as it is.

Sounds like they should stick with IBM, and follow up with their R&D. Let's see if they can bring Watson into PS3/4.
I think system architecture and software library/maturity may be more important than CPU/GPU choice so far.

Binary compatibility with ARM will be broken/incomplete anyway if Sony continues to use SPU.

 

[liolio]

There are pretty interesting discussion on the matter in the realtechworld forum for those interested
For binary compatibility, well if you don't use SPUs there is no problem SPUs have their own ISA by the way I'm not sure I get your post properly by the way.
Honestly for low power but hight perfs I would trust IBM (as INtel is not in the party it seems) more than ARM, I'm sure their POWER A2 will deliver outstanding perfs with low power characteristic (according to the perfs they reach) at the job they designed for as SPUs were.

 

[patsu]

I meant they will have to at least change and recompile the code anyway if they want to use the SPUs.

 

[Heinrich4]

I dont know if latest ARM have even 50%* (counting with "tweaks"...) power of Cell processor,but we have some people talking very well about port from ps3 to NGP.

http://www.computerandvideogames.com/288500/news/sony-ngp-porting-ps3-games-simple-and-quick/


3 hours for ps3 and 20 minutes doing same thing in NGP(now PSVita..i know... less resolution etc but they claimed great results...):

http://www.vg247.com/2011/02/04/new...source-calls-the-handheld-a-developers-dream/

With some SPUs as co-processor/stream processor helping...maybe ARM(and GPU Tile-based deferred render on die...) could handle very well backwards compatibility for ps4.

*
(Developers really needs a "very powerfull cpu" doing everything and low range GPU or extremelly eficiently and "reasonable-powerfull-GPU-for-five-years-cycle" + cpu "medium range like/enough power" extremely user-friendly?... Yeah.. we know the answer... )

 

[msia2k75]

A GTX 560 could be possibly in a console at 28nm?

 

[AlNom]

Anything is possible unless you define a set of restrictions.

 

[Brad Grenz]

I dont know if latest ARM have even 50%* (counting with "tweaks"...) power of Cell processor,but we have some people talking very well about port from ps3 to NGP.

http://www.computerandvideogames.com/288500/news/sony-ngp-porting-ps3-games-simple-and-quick/


3 hours for ps3 and 20 minutes doing same thing in NGP(now PSVita..i know... less resolution etc but they claimed great results...):

http://www.vg247.com/2011/02/04/new...source-calls-the-handheld-a-developers-dream/

With some SPUs as co-processor/stream processor helping...maybe ARM(and GPU Tile-based deferred render on die...) could handle very well backwards compatibility for ps4.

*
(Developers really needs a "very powerfull cpu" doing everything and low range GPU or extremelly eficiently and "reasonable-powerfull-GPU-for-five-years-cycle" + cpu "medium range like/enough power" extremely user-friendly?... Yeah.. we know the answer... )

But what's the advantage of ARM over sticking with even the current Power based PPU? You trade backwards compatibility, the expense of integrating a new CPU design, possibly even performance and what do you get? Maybe some marginal thermal/power advantages which won't help you very much in a home console setting, and a nebulous sense of "synergy" with NGP which evaporates once you're adding SPEs anyway? It's just a lot of trouble with no tangible benefit. Look at it this way, being able to easily target both the NGP and PS3 with the same game is pretty cool, but if PS4 games are easy to port to the NGP, I would consider that a huge fucking failure for Sony.

I know ARM seems like the new hotness, but it's easy to forget ARM is optimized for a very different usage scenario than a home console. They're very impressive for mobile applications, but you can't expect a 1 watt design to magically scale up well to 50-75 watts for a PS4.

 

[Arwin]

So basically it all comes down to how much time you spend on optimizing for your target hardware. Raw performance-wise PCs were way ahead of consoles by the time PS3 launched, just no one did any serious optimizations for them.

Yes, but my point is different. I am not ready to argue the actual potential of various hardware components in 2005, but PC architecture with DirectX9 simply did not allow that type of optimisation on PC. Considering the importance of PC as a development platform for new (multi-platform) engines crucially, next generation hardware is going to have an interesting relationship with DirectX11 and 12. I thought it was very interesting how D.I.C.E. pointed towards some serious limitations still in DirectX11 in that regard.

Whatever the next-gen hardware is going to be, the graphics pipeline to employ them is going to be even more crucial than ever before. I wouldn't want to be in hardware designer's shoes at the moment! And if I thought I could get away with it, I'd postpone the next generation as long as possible while a team
of software and hardware architects figure it out together.

 

[Lazy8s]

PowerVR has been the obvious choice for many console designs they didn't win, painfully so in some cases (PSP1 oversight -- ignorance); console manufacturers frequently haven't had their priorities straight.

While they may have favored aggressive discounts, supplier politics, and/or pre-existing relationships when making their GPU choices in the past, competition from companies like Apple who deliver cutting edge hardware and make software to exploit it will require other companies to use efficient solutions more and more.

 

[rpg.314]

PowerVR has been the obvious choice for many console designs they didn't win, painfully so in some cases (PSP1 oversight -- ignorance); console manufacturers frequently haven't had their priorities straight.

While they may have favored aggressive discounts, supplier politics, and/or pre-existing relationships when making their GPU choices in the past, competition from companies like Apple who deliver cutting edge hardware and make software to exploit it will require other companies to use efficient solutions more and more.

Are you arguing that competitive pressure from Apple in mobile device space will lead to some kind of reevaluation of suppliers amongst immobile console vendors? I doubt it, but I wouldn't say never.

 

[Heinrich4]

But what's the advantage of ARM over sticking with even the current Power based PPU? You trade backwards compatibility, the expense of integrating a new CPU design, possibly even performance and what do you get? Maybe some marginal thermal/power advantages which won't help you very much in a home console setting, and a nebulous sense of "synergy" with NGP which evaporates once you're adding SPEs anyway? It's just a lot of trouble with no tangible benefit. Look at it this way, being able to easily target both the NGP and PS3 with the same game is pretty cool, but if PS4 games are easy to port to the NGP, I would consider that a huge fucking failure for Sony.

I know ARM seems like the new hotness, but it's easy to forget ARM is optimized for a very different usage scenario than a home console. They're very impressive for mobile applications, but you can't expect a 1 watt design to magically scale up well to 50-75 watts for a PS4.

I agree with your statements, but it seems the reports coming out that sony will follow a user-friendly model for developers and lower costs (BOM - Bill of Materials) and perhaps the ARM architecture is the solution to some the field expenses of licensed IP and customization coupled with very low TDP,we must remember that high wattage were the great enemies of the consoles 'HD' causing RRods, Ylods and other flaws and certainly sony not want to spend large resources with consumer and extended guarantees(justices) or constant refreshs MBs that result in spending and killing profits with software sales.

And indeed, perhaps the perfect combination of ARM as "PPU" + SPUs and PowerVR 6 rogue could be complex and an attempt at synergy that may not result in acceptable performance for a console set for 5 or 6 years, but ARM architecture is so well known or more than powerpc and setting in so many companys,hardwares archtectures ( http://en.wikipedia.org/wiki/ARM_architecture even Intell! ) and increasingly accepted by developers (Microsoft in their windows 8) they (sony) may have enjoyed great results experienced with NGP/PsVita and are aimed at achieving a synergy with market cell phones(Apple changed all game market paradign) that seens to grow quickly... and still have to think about compatibility with ps3,because even though it co-exists with PS4 would be a big attraction there improvements in games (research indicates that 10% of gamers use backward compatibility enjoy a lot "smooth" features).


And I confess my dream machine is a Power 7 w/8 core 3.35GHz/32MB L3(at 45nm reachs 240 watts) cache@ 22nm + Maxwell 4TFlops GPU at 20nm with 8 GB RAM (VRAM and system), but unfortunately for a console that you spend no more than 250-Watt and costing up to US$400 I think not possible even for 2014/2015.