*spin-off* Has The PS3 Been Based On A EE/GS Evolution

[liolio]

I'm willing to have more info about the chip used in the 3DS, too bad the pdf on DMP won't load.
Is somebody managed to download it or my PC/network is the problem?

I would be interesting to see how it perform in perfs per Watt and per mm² vs more programmable devices provided by PowerVR and to some extend to our consoles GPUs.
EDIT
I've gone through every piece of information available in the handled forum dedicated thread without much sucess. Anyway I feel like something a super sized PICA200 may have been a good choice for the system.

 

[Carl B]

Sony should have that idea when they looked at their Eyetoy. Eyetoy was like one of the best selling peripheral. They just didn't know how to move forward with it. Sony Move controller tech was demonstrated long time ago in one of the dev conference. Eyetoy ability to track colour bulb thing look promising even back then for gaming. They sit on that idea and didn't do anything with it. Wii happened and Move is still not out. That's how clueless they were.

Oh my god seriously, I am with you 100%. Sony has all of this (fairly impressive) internal R&D... and then what do they do with it? Tech is left to languish, IMO for want of execs knowing how to angle/market it. Take that PSEye game they were working on for a while where you draw up tanks and other objects and things and they just sort of work and animate on screen... where the hell is that thing at? Did you cut the budget for both guys you had assigned to it? Ugh.

That change for PS3, where they pour much of the development effort into Cell and tacked RSX at the end. My guess is that Sony had a raster hardware for GPU ready to pair with Cell for quite sometimes. Why they went with RSX is said to be politic, as in devs pressure for familiar dev environment. Anyway RSX was definitely last minute, you can see it when NV didn't have RSX ready when it was announced and devkit had the 256 bit version and of course later they have to add the PS2 hardware in there too, subsequent delay, second revision still having GS for BC, and BC canned at the end.

Well exactly - by now we all know the legend of the RS, which was the direction Sony was going in with Toshiba rather than the on-die/package visualizer setup in the original patent; same sort of deal as GS spiritually in terms of uber-fillrate, supposedly crazy programmable when setup correctly, and obviously well suited to work in conjunction with Cell. Of course, when you are creating an even more exotic environment for devs, it's super large/hot/expensive... and Cell is running behind by a node itself... looks like progress and engineering excitement are the casualties. You always have NVidia with a dusty design to sell you when you run into those cases.

The rumors on Cell's fate seem to wax and wane for PS4, but whatever happens I do think it could have come into its own with the support for a true second generation - both in console and elsewhere. We'll see what we see I guess!

 

[liolio]

Oh my god seriously, I am with you 100%. Sony has all of this (fairly impressive) internal R&D... and then what do they do with it? Tech is left to languish, IMO for want of execs knowing how to angle/market it. Take that PSEye game they were working on for a while where you draw up tanks and other objects and things and they just sort of work and animate on screen... where the hell is that thing at? Did you cut the budget for both guys you had assigned to it? Ugh.

Indeed the sad point is that the one who paid the cost of those incompetencies, lack of vision and execution are the employees... that got fired during the crisis.

Well exactly - by now we all know the legend of the RS, which was the direction Sony was going in with Toshiba rather than the on-die/package visualizer setup in the original patent; same sort of deal as GS spiritually in terms of uber-fillrate, supposedly crazy programmable when setup correctly, and obviously well suited to work in conjunction with Cell. Of course, when you are creating an even more exotic environment for devs, it's super large/hot/expensive... and Cell is running behind by a node itself... looks like progress and engineering excitement are the casualties. You always have NVidia with a dusty design to sell you when you run into those cases.

The rumours on Cell's fate seem to wax and wane for PS4, but whatever happens I do think it could have come into its own with the support for a true second generation - both in console and elsewhere. We'll see what we see I guess!

Honestly now that it mostly all said and done I really think that Sony should not have betrayed the EE+GS philosophy. They should have stick with proprietary hardware or partner with/buy tiny companies. Sony spent a shit ton of money in R&D but in foundry too.
One has to wonder about why Sony wanted to break the memory wall, try to revolutionize the IT, etc. when they are selling consoles, tv, etc. Clearly their executives got a bit hot headed. They were close to a monopoly and if they really wanted get in people living room they needed to consolidate what they had, not try to make engineers wet dreams come true.

Think of Nintendo move with the 3DS, ok it's unlikely to reach perfs offered by ATI mini xenos, Nvidia Tegra line or powerVR SGX products but I'm willing to know more about the size of the thing and its power consumption. DMP is a tiny company I'm sure Nintendo got a great deal.
Sony may have bough them it would have not even been noticeable on their fiscal report.
I'm confident it could have been a great addition to their existing research and whatever they were working on for the PS3. Clearly the PICA200 is not up to the task but it's designed for the mobility market and conceptually I feel like it would have been a great match to could/should have been an evolved Emotion Engine.
While not programmable it offers most effects that opengl ES 2.0 can provide and while I'm still waiting for more informations about the 3DS I'm confident it does it at a consistent perf/mm² advantage.
Here an update description of what the system could have looked like (with less details):

*One pretty big chip (300mm²<< xxx <400mm²)
*UMA design.
*Lower power consumption and thermal dissipation.
*Better form factor.
*close programming model so developers would have been familiar to it from day one.
*Easily achieved backward compatibility.

*EE.2
Two MIPS cores @800MHz
Super Vu0 unit, 16 wide, accessible to both CPU (think of it as a primitive bulldozer) @800MHz
Vu0 unit linked to a super Vu1 unit,
Vu1 unit, 64 wide, @400MHz
DSP for decompression.
Various improvements across the board.

*GS.2
A potent but fixed function GPU.
Tile based rendering.
MSAA support.
Clock speed 400MHz.

*MISC
512MB of GDDR3.
E-dram accessible to both the GPU and the Vu1 unit.
A third more primitive CPU handling OS and security.
BRD player.
Good scaler.
Hardware multimedia decoder.
60GB HDD
399$/€ at launch, only one SKU.
Sony would have team up with NEC in regard to process technology.

Overall, I think it would have been enough of a match for the 360. Performances is not everything.
The PS3 would still be a more complete system: BRD player, standard HDD but they would have been there without the price premium. They would also have been in a position to start a price war against MS: way cheaper hardware, less R&D cost to cover, less investments in foundry, no royalties. Honestly considering Sony editing fire power I think that MS would have been obliterated and pushed out of the market.
Sony would have been in a better position to counter the Wii too. Without being bleeding so much money and focused on catching up they may have considered to leverage their R&D and would have the funds to buy some relevant company(ies (3DV, Primes sense come to mind).

 

[Arwin]

Oh my god seriously, I am with you 100%. Sony has all of this (fairly impressive) internal R&D... and then what do they do with it? Tech is left to languish, IMO for want of execs knowing how to angle/market it. Take that PSEye game they were working on for a while where you draw up tanks and other objects and things and they just sort of work and animate on screen... where the hell is that thing at? Did you cut the budget for both guys you had assigned to it? Ugh.

Actually, it's in EyePet since release, and in the latest patch for EyePet (and first US release at the launch of Move) you can now also draw the stuff with the Move controller on the screen, and with 3D support to boot. So EyePet has Augmented Reality in various shapes and forms, AI, 1-1 motion control, fur effects and 3D, all in one neat, casual friendly package. I'd say Sony is improving in bringing their tech out there rather than declining, these last few years.

 

[Carl B]

Actually, it's in EyePet since release, and in the latest patch for EyePet (and first US release at the launch of Move) you can now also draw the stuff with the Move controller on the screen, and with 3D support to boot. So EyePet has Augmented Reality in various shapes and forms, AI, 1-1 motion control, fur effects and 3D, all in one neat, casual friendly package. I'd say Sony is improving in bringing their tech out there rather than declining, these last few years.

Well if they're getting the tech out, they're not getting the word out. I think Europe has a greater awareness of the Sony side projects for whatever reason - here in the US the support for these innovations is nowhere that I can see. And in EyePet alone - which I'll be honest, does not compel me as a title - just seems a strange move. Maybe I'll purchase it to check it out though. There's still no rhyme or reason for Sony not to have had both a Kinect alternative and a Wii-mote alternative before either MS or Nintendo respectively, since they were already putting R&D into those fields.

 

[Shifty Geezer]

They could have thrown loads of funky little titles up on PSN to generate word-of-mouth interest. Sony as a business are utterly rubbish at handling their brilliant thinkers and creators.

 

[liolio]

They could have thrown loads of funky little titles up on PSN to generate word-of-mouth interest. Sony as a business are utterly rubbish at handling their brilliant thinkers and creators.

Indeed no matter the hard they lacked vision. They had an engineers in charge that wanted to create a mondial grid of PS3 and in the same time they didn't come with the proper support to connect gamers/users.
Sony basically is blessed to survive so many errors. To some extend Microsoft suffers from the same issue, they spend shit ton on money on R&D, they have an incredible human power and I'm sure innovations are dying in some labs, in the mean time Google...

 

[Arwin]

Well if they're getting the tech out, they're not getting the word out. I think Europe has a greater awareness of the Sony side projects for whatever reason - here in the US the support for these innovations is nowhere that I can see. And in EyePet alone - which I'll be honest, does not compel me as a title - just seems a strange move. Maybe I'll purchase it to check it out though. There's still no rhyme or reason for Sony not to have had both a Kinect alternative and a Wii-mote alternative before either MS or Nintendo respectively, since they were already putting R&D into those fields.

They haven't done it too well, but they did the right thing. EyePet was released in Europe last year and received middling scores - the interaction with the Pet just doesn't work that well with the PS Eye alone. I should know, because I have the game. For the U.S., they therefore decided to delay its release and get Move support in there, and as it turns out now they have had time to add 3D support as well.

Many of the Move tech demoes are making it into games pretty quickly now too. EyePet has the augmented reality bit for the Move controller with a blowdryer, shower and some other stuff, but of course Start-the-Party has it too for most of its games, and it works well.

All the face recognition and head-tracking and so forth is now coming together with the Move controller support and 3D. Gran Turismo 5 showed both 3D and head-tracking (though impressions were mixed, the E3 environment probably not great, and as other Sony devs suggested, head tracking might be better off with moving the head up/down/left/right rather than turning your head). So many games are picking up Move and 3D support, and some of the Move games don't just track movement of the controller, but they also take into account your upper-body movement combining information from the controller and the camera with some clever skeleton computations.

My take is that the main problem so far has been to integrate knowhow and tech development going on in the three regions and use them across the three regions effectively. Now with Sony worldwide studios (pushed by Phil Harrisson, whom however found everything going far too slow for his taste, and it probably came too late as well), this process is now finally becoming far more effective and efficient and we're actually seeing everything from the different regions coming together and being shared.

It's very promising to see this happening, and I hope it brings Sony success.

 

[Carl B]

Integrating Move and head-tracking and such isn't what I would want to see from them ideally though; I'm not one for integrating 'gimmicks' into existing well defined experiences, I'm more for creating altogether new experiences around said tech. EyePet counts, but GT5 adaptions, I don't know.

Like the existing games that will support Move, is it compelling? Honestly with EyeToy/PSEye, what I would have wanted to see and much earlier would have been the minority-report interfaces - simply for novelty's sake - more Play/Play II type games, maybe an FPS/Mirror's Edge/non-stop sort of hybrid thing where you are constantly moving forward (so no directional input required save for quick actions), but the game tracks your body movements in order to deal with the obstacles and enemies.

When the basic controller remains the best interface in a game, don't try to bring external interfaces in, rather build new games around them where the traditional controller is sub-optimal. I remember back to PS3's launch and an aspect that was supposed to revolutionize and become integral to core gaming on the system: the SIXAXIS. I don't think it made much of a mark.

 

[corduroygt]

Why not 1/2 G80 evolution instead of going with G71? Take half of 8800 GTX, released November 2006, and cut it in half, 384MB of 192-bit memory, 64 unified shaders, 12 ROPS. G80 is 680M transistors, so the cut down version would be definitely less than 400M. They should have pushed NV to make a half G80 in parallel, and yields, etc would favor the smaller chip. R600 would be a little too late.

Oh well, I'm sure they've learned their lesson and PS4 won't have any such cripplings...combined with the millions of hours of cell knowledge they had to learn to help out if needed as well.

 

[liolio]

Why not 1/2 G80 evolution instead of going with G71? Take half of 8800 GTX, released November 2006, and cut it in half, 384MB of 192-bit memory, 64 unified shaders, 12 ROPS. G80 is 680M transistors, so the cut down version would be definitely less than 400M. They should have pushed NV to make a half G80 in parallel, and yields, etc would favor the smaller chip. R600 would be a little too late.

Oh well, I'm sure they've learned their lesson and PS4 won't have any such cripplings...combined with the millions of hours of cell knowledge they had to learn to help out if needed as well.

Spoiler

because it would not be based on EE+GS

The answer is that Nvidia either could not or didn't want to modify their schedule for G80 derivative.

 

[jonabbey]

And Sony may have been shooting for an earlier release date than they managed to hit. Sony surely didn't want to come out a year after the 360.

 

[patsu]

Well if they're getting the tech out, they're not getting the word out. I think Europe has a greater awareness of the Sony side projects for whatever reason - here in the US the support for these innovations is nowhere that I can see. And in EyePet alone - which I'll be honest, does not compel me as a title - just seems a strange move. Maybe I'll purchase it to check it out though.

In US, they delayed the game at the last minute. The "special" PSEye features you're looking for are in the GameSpot interview:

Our favorite scenes are @ 5:34 (Teaching EyePet how to sing) and 7:45 (Sketch recognition). My son imitated the EyePet a few times after watching the video.

Kotaku has an article on EyePet after Move integration:
http://kotaku.com/5565368/eyepet-what-a-difference-the-playstation-move-makes

Should the game be delayed to work with Move, or was the tech they had installed good enough?

The answer became much easier after the EyePet's European launch in October of last year revealed a strong flaw in the game.

In order to spawn virtual objects for your pet to play with, Sony included a flat card that players would hold up to the camera, spawning items like toys, hair dryers, and other fun tools to use on your pet.

This worked great when the flat side of the card was facing the camera.

"The problem was that children didn't understand the need to keep the card facing the camera," Doucet explained, waving the Move controller he was using to show me the game off-screen. "As they turned, so did the card, making the objects disappear."

The Move, on the other hand, has a uniform shape no matter how you point it at the camera. Furthermore, it can even be detected when it's behind your back. Doucet moved the controller behind him to demonstrate, and the game still detected a slightly fainter blue circle of light.

Personally, I hope they make casual gamers deeper. e.g., for the sketch recognition, I think it's much better to have the recognized objects in a real interactive game like the "Tank war" demo. Now it seems that the pet can ride the recognized object and that's all.

There's still no rhyme or reason for Sony not to have had both a Kinect alternative and a Wii-mote alternative before either MS or Nintendo respectively, since they were already putting R&D into those fields.

Well, looking at the current Kinect issues, I think I understand Dr. Marks' concerns better. However, I think they could still market PSEye more aggressively instead of sidelining it now.

 

[patsu]

Integrating Move and head-tracking and such isn't what I would want to see from them ideally though; I'm not one for integrating 'gimmicks' into existing well defined experiences, I'm more for creating altogether new experiences around said tech. EyePet counts, but GT5 adaptions, I don't know. Like the existing games that will support Move, is it compelling?

That's the problem. It's hard to say without trying ourselves. GT5 should be played with a steering wheel. Flight games have flight stick. Music games use instruments. FPS have DS3 although some may prefer Move/Wiimote+. Heavy Rain, Folklore, LBP2, Sorcery, Golf, Tennis, and Baseball seem more natural with Move, but they have to make the tracking and recognition work flawlessly.

Basically, it's difficult to draw a conclusion. The desire to buy Move will be dampened.

Honestly with EyeToy/PSEye, what I would have wanted to see and much earlier would have been the minority-report interfaces - simply for novelty's sake - more Play/Play II type games, maybe an FPS/Mirror's Edge/non-stop sort of hybrid thing where you are constantly moving forward (so no directional input required save for quick actions), but the game tracks your body movements in order to deal with the obstacles and enemies.

When the basic controller remains the best interface in a game, don't try to bring external interfaces in, rather build new games around them where the traditional controller is sub-optimal. I remember back to PS3's launch and an aspect that was supposed to revolutionize and become integral to core gaming on the system: the SIXAXIS. I don't think it made much of a mark.

I agree in principle (although there are grey areas). Not sure if LBP2's Move editor will allow users to contribute new Move and PSEye ideas. Sony should try to define new type of games using both. I think Sony knows that but the challenge is these new ideas are not created by mythical man-month. They take a long time to gestate. The real problem is even when Sony found something compelling (like the Tank war demo), they left it to die an anonymous death. Perhaps Sony held back because of PS3's high price point, but it's a different situation now.

FWIW, the Sony team demoed a list of mechanics here:
http://kotaku.com/5569137/how-developers-see-the-playstation-move

I am also curious to see "whipping" as a mechanic too.

 

[liolio]

And Sony may have been shooting for an earlier release date than they managed to hit. Sony surely didn't want to come out a year after the 360.

Anyway this has been discussed quiet often, not sure I want to dig this to much actually, not uninteresting but off-topic.

The discussion on eye-toyPSeye while not exactely on topic is a bit more parallel as the ps2 was lacking the horse power to make anything sensible with it. The actual PS3 doesn't.

The thing I find interesting is that even if Sony went with something a bit like I described, they would have been in a much better situation to leverage the R&D. They would have a much better cash flow so they may have considered some buy out (prime sense / 3DV ). They would also have another advantage vs MS they could have built whatever needs a PS-eye more akin to kinect in house which may have allowed for a more flexible strategy in regard to pricing.

Integrating Move and head-tracking and such isn't what I would want to see from them ideally though; I'm not one for integrating 'gimmicks' into existing well defined experiences, I'm more for creating altogether new experiences around said tech. EyePet counts, but GT5 adaptions, I don't know.

I don't agree I never experienced it for my-self but Head-tracking would be a great addition to almost any game. The sense of depth highlighted by some demo/vids available on the web is impressive, I can only wonder about what would be the impact of this tech on a 3D display.
On top of it's non situational whereas Wiimote/move are, I just realize that now that I play at my desk their relevance as pointing device is close to zero... Head tracking on the other hand works in any situation, whether you stand play an "physically active" game, crumbled on the couch or sitting at a desk

 

[Arwin]

Personally, I hope they make casual gamers deeper. e.g., for the sketch recognition, I think it's much better to have the recognized objects in a real interactive game like the "Tank war" demo. Now it seems that the pet can ride the recognized object and that's all.

That's not right. The plane thing, yes, but there's also a game where you can draw a car in the EU version, and after the car has 'come to life', you can then remote control that car with the DualShock with which you can then play cat and mouse with the Pet. This is extremely close to the idea of the tank demo.

Well, looking at the current Kinect issues, I think I understand Dr. Marks' concerns better. However, I think they could still market PSEye more aggressively instead of sidelining it now.

From now on, PS Eye will be part of Move. It's a great side-effect that the camera is included. Anyway, we're strongly veering away from the topic.

 

[liolio]

The rumors on Cell's fate seem to wax and wane for PS4, but whatever happens I do think it could have come into its own with the support for a true second generation - both in console and elsewhere. We'll see what we see I guess!

A bit OT but anyway as far as the discussion is polite it's not a problem

You should listen

Spoiler

if you can tat's it sound quality is pretty bad

to this discussion between Sweeney and Richard, Andrew Richard has a really interesting take on Cell
http://www.semiaccurate.com/2010/06/24/epic-debate-part-5-weapons-unsheathed/

By the way part six (the last one) should but up soon, kuddos to Groo for those interview

 

[Carl B]

I'll leave my feelings on Groo out of it...

Anyway I watched part 5 you linked above and part 4 that preceded it; it's an interesting debate no doubt, one that honestly reflects a lot of what is pertinent around here at any given time in a number of threads (hardware vs software, fixed-function vs programmable). I'm not sure that what Andrew was saying spoke directly to what I might be envisioning or reflecting on in terms of a 2nd gen Cell, but from a landmark/benchmark perspective it was interesting to hear his thoughts. Frankly I do hope that the architecture is carried forward, even though my gut feelings are that at the moment, it will be in a B/C form or not at all. I'd love to be wrong, as I think the entire project was a botched effort with the musical chairs going on at Sony at the time - but the vision was sound.

 

[patsu]

That's not right. The plane thing, yes, but there's also a game where you can draw a car in the EU version, and after the car has 'come to life', you can then remote control that car with the DualShock with which you can then play cat and mouse with the Pet. This is extremely close to the idea of the tank demo.

Ah, I didn't know that. They should expand more on that idea alone in the same game. Looking forward to EyePet.

but the vision was sound.

There are some rough edges (e.g., Local Store too small, PPU is f*cked up slow), but overall I think Cell has turned out amazingly well; thanks mainly to relentless developer determination.

 

[Akumajou]

And Sony may have been shooting for an earlier release date than they managed to hit. Sony surely didn't want to come out a year after the 360.

Thats debatable the only time Sony came neck and neck with a direct hardware competitor was with Sega and the Saturn in 94 in Japan and 95 in the US while Jaguar and 3DO had over a yeand start.

2006 falls in line with what Sony was aiming at however yields and software learning curve and most importantly that E3 2005 was set as a place to announce a release date for an unrealistic spring 2006 to please shareholders and press expectations.

I think its possible for a EE evolution based PS3 but I think that Sony also went after IBM for the chip fabrication capabilities to meet demand hence Cell but for the EE evolution we would have to eliminate IBM and leave Sony and Toshiba to design and fab the chip.

for GS evolution I really would not mind seeing something like that since I feel that GS could have used higher frequency and more ram, maybe Sony made an arcade board with a beefed up EE and GS to test scaling, EE and GS where very capable chips but the millions of dollars/yen that it takes to dream up these architectures is staggering, I feel that PS2 would have had to last up to 2008 09 for replacement to give the engineers enough time to design chips for 65nm to 55nm or better yet 40nm, I feel in retrospect that 2005 was too early to start a new generation.

The other thing is that theoretically engineering sample prototype PS2 board was shown with running software demo in March 1999 so the design had been going for many years, I went back into old magazines, namely pre-US PSX (early 95) launch next generation magazine where Ken Kutaragi was interviewed and quoted that he believed a .250nm chip design would yield 10 million polygons per second

So years later after PS2 and after it was well known that rich kid Microsoft was in the mix in 2001 Cell design and vision was pretty much announced or talked about and the effects of that new era affected the timing of a GS follow up that would have to perform.

So from 2000 to 2005 or 2006 at 90nm engineering process to design a GSE that would still be on a 128bit bus would have to incorporate more cache and better comunication to ram and have a sophisticated pipeline evolution I think maybe 64 pipes to GS's 16 and cover full hardware t&L proprietary pixel, vertex shading evolution and incorporate some type of hardware culling or deferred rendering engine.