News & Rumors: Xbox One (codename Durango)

[Nisaaru]

Sure.
I only noted it since it sounds eerily similar in concept to the idea of using display planes in Durango to layer 3 planes on top of one another to allow their fidelity to dynamically be adjusted by the game engine in order to keep the res, fps, etc of the target display plane at full fidelity (so they can offer QoS guarantees...aka "1080p, 30fps or 60fps locked gaming is standard for all Durango games"). That'd be a helluva bullet point to beat their chest over, even if the background planes actually render at dynamic res/fps/etc. So it'd be *only* slightly misleading depending on the implementation.

I see this QoS stuff mostly as a method to run all kinds of gimmick services in the background not affecting the foreground game experience. Services nobody but some hipster or marketing drone require to justify their job.

 

[Silent_Buddha]

Sure. http://research.microsoft.com/pubs/176610/foveated_final15.pdf

The basic idea is use an eye tracking camera device to track/calculate where your eyes are looking on the screen and render the circular "foveated region" surrounding that point on the screen at high res/fps/polygons/etc. Then surrounding that region with another layer that is at lower res/fps/polygons/etc. Then they iterate on that once more in a 3rd layer, likewise at lower fidelity than the 2nd layer. They then blend everything together smoothly. This all happens on the fly as your eye moves around the screen. So you only are rendering the stuff your eyes are actually focused on at high fidelity while everything else is gradually lower fidelity.

The human eye naturally does this anyhow, so the novel concept here is that computer games and whatnot are wasting resources by rendering an entire frame at high fidelity when viewers can't ever focus on anything more than small regions of it at a time anyhow. I won't post more on this since tehre is likely already a topic for it somewhere and by itself it's not rumored to be part of Durango's setup as it stands.

Anyhow...

I only noted it since it sounds eerily similar in concept to the idea of using display planes in Durango to layer 3 planes on top of one another to allow their fidelity to dynamically be adjusted by the game engine in order to keep the res, fps, etc of the target display plane at full fidelity (so they can offer QoS guarantees...aka "1080p, 30fps or 60fps locked gaming is standard for all Durango games"). That'd be a helluva bullet point to beat their chest over, even if the background planes actually render at dynamic res/fps/etc. So it'd be *only* slightly misleading depending on the implementation.

It is interesting research but not practical for a console currently for quite a number of reasons.

The largest obstacle is going to be the fact that this is very latency intolerant. Which means it requires very low latency between when the user's view shifts and when the display is updated. Which leads to the following issues...

[1] It requires an eye tracking system with very low latency. In this test they had to use a Tobii TX300 with an average eye tracking latency of 7-10 ms (image capture, processing, and then submitting to the PC). This isn't an inexpensive piece of hardware. The Tobii TX300 also has a limit of approximately 26". Fine for desktop work, not so much for a living room TV. A system that could not only accurately track a person's eyes AND do it quickly enough to provide the necessary low latency is going to be quite expensive.

[2] Rendering latency was 8-18 ms. As it approaches 18 ms, the quality of the rendering starts to suffer with the illusion breaking as you get to the point where the foveated center region falls out of the center of your view enough that it appears to "pop" in as the user's eyes move. That doesn't sound as bad as it could be since you gain some performance in saved rendering time. But not in saved gameplay (physics, AI, etc.) time. So it's "bad" but not that bad. 60 FPS allows less graphics rendering time savings allowed for increased graphics rendering than does 120 FPS.

[3] Scan out latency also varied from 2-8 ms.

[4] But the biggest potential deal breaker is that at the end of that chain you have the latency of the display itself. Considering that Microsoft won't be able to control what TV a user will use that latency can range anywhere from a low 3-6 ms (you basically need a relatively specialized TV for this) to as high as over 100 ms (TV video processing, A/V passthrough and potential processing there as well). The test used a PC monitor with a measured latency of 6 ms.

All of that means...

[1] is going to be prohibitive from a cost and implementation perspective.

[2] will put very hard constraints on the time a game can spend for rendering, AI, physics, etc. Basically it will require games to have a mandatory 60-120 FPS rendering performance. No more 30 FPS games.

[3] is the easiest to deal with and is just down to making sure the right graphics hardware is chosen for the console.

[4] is something that neither the game developer nor the console manufacturer will be able to control. And all that effort will go to waste if your average console player uses an average LCD TV.

This tech could do wonders on PC, however. Where gamers are used to picking monitors specifically with an eye towards a low latency display. It also makes eye tracking easier as the user is usually sitting within two to three feet of their display.

Very nice tech, but impractical in the living room currently and for the near (5-10 years) future.

Oh and forget multiplayer games on the same console with something like this. Forget having other people in the room watching you play as well, as their eyes are not likely to track the same as the user's eyes.

Regards,
SB

 

[pjbliverpool]

Actually no that is not more likely as it doesn't fit the fact pattern. To meet your goal, its seems unlikely that they would have chosen to go the Durango route.

How do you come to that conclusion? If you're suggesting they could have released a set top box without the gaming capabilities then no, they could not. Gaming and especially casual / party gaming will still be the main draws of the system. Its just that the other functions will more then counterbalance the lack of core gaming power.

 

[Arwin]

I put $10 on my Skype account 2 years ago, I use it to phone home when on business trips, I think I'm down to ~$3 left.
Now I'm not a heavy phone user, but that seems like a pretty good deal.

We use it at work, I get new credit assigned automatically. I can blow through 10 euro in an hour thanks to international calls in particular even if they are just within Europe. But of course lots of cheap calls from Skype to Skype, with more and more people having it available.

 

[Xovek]

So? Why would they add a fairly competitive gpu for a low end sku? Why would they go through the trouble of adding ESRAM wich is a sign of trying to make it more competitive rather than them aiming for low end performance. They have gone over budget if their goal was to design a low end SKU, and im going to assume Microsoft engineers arent stupid

Ockhams razor: Durango is Microsofts answer to core gaming and the upcoming threat from Apple/Google in the living room. Its a flexible console designed to be competitive in both areas

Excuse my intervention, but besides the fact I´m agree with you when you say durango is the answer to google and apple´s attempt to gain the living room, the fact is when the vast majority of resources of durango´s hardware are not aiming exclusively to games -I mean, there are evidences ponting other functions like display planes, bewtween others functions-, then is not a console... in the traditional meaning of term.

Maybe Durango is set to make a new segment in market.

Quite interesting this porpousal of MS in this age of entertainment. IMO I do believe this (Durango) is the beggining of death of the traditional consoles.

 

[astrograd]

I see this QoS stuff mostly as a method to run all kinds of gimmick services in the background not affecting the foreground game experience. Services nobody but some hipster or marketing drone require to justify their job.

Did you read the patent or are you only saying this based on the Yukon lea, which also seemed to have the concept shown in the target sketches? The patent goes out of its way to specifically not the gaming applications.

 

[astrograd]

It is interesting research but not practical for a console currently for quite a number of reasons.

The largest obstacle is going to be the fact that this is very latency intolerant. Which means it requires very low latency between when the user's view shifts and when the display is updated. Which leads to the following issues...

Just wanting to clarify for you and others reading...my comments there are less about using the foveated research stuff in Durango and more about trying to glean from that research some rough idea of what the display planes might bring to the table. I don't necessarily agree/disagree with your concerns there, but again, pretty much none of them are relevant to the display planes concept (which has nothing to do with the eye tracking end of things).

That foveated tech is just mentioned as a highly ambitious implementation of that same concept. I'm only meaning to use it as a broader reference to help us forge a useful context in discussing these display planes (alongside the patents for the display plane stuff).

With any luck VGLeaks will report about those this coming week. :smile:

 

[Alucardx23]

Interesting read for those with more understanding,it was some spanish article translate to english.

http://www.microsofttranslator.com/...013/02/08/durango-nos-hace-coger-el-delorean/

Excellent article, the guy is talking about the virtualized textures as a main focus for the 720 since January 5.

http://translate.google.com.do/tran...com/2013/01/05/dando-forma-a-durango/&act=url

 

[SlickShoesRUCrazy]

So? Why would they add a fairly competitive gpu for a low end sku? Why would they go through the trouble of adding ESRAM wich is a sign of trying to make it more competitive rather than them aiming for low end performance. They have gone over budget if their goal was to design a low end SKU, and im going to assume Microsoft engineers arent stupid

Ockhams razor: Durango is Microsofts answer to core gaming and the upcoming threat from Apple/Google in the living room. Its a flexible console designed to be competitive in both areas

'low end' sku is your terminology. I said consumer sku. Microsoft wants to compete with Apple/Google for the living room, you're right. Which means gaming takes a back seat, which to me, seems exactly what has happened with the recent specs if legitimate. To me, it seems like they are taking a 'good enough' approach. Granted again, my technology knowledge is basic at best.

Now the consumer sku is basically what was seen in Yukon pdf dated 2010, minus the multiple cpus/gpus. It's a $299, or less, box that is 'good enough' for the average consumer. The people that are going to spend a lot of time using the media features rather than the gaming features. Basically the audience being aimed at by Apple/Google. The people that would be happy that they can play call of duty or their sports games or what not. don't care what resolution, don't get pissed at frame drops, screen tearing etc.

The consumer sku will be aimed at the average consumer who are the majority of the market, which a $299 or lower price makes since.

But there are still millions of people out there that are not an average consumer. People you consider enthusiasts who do not want 'good enough' hardware, but want top of the line hardware. They might be pushed away by the general consumer/casual approach where gaming is the back seat. So that is where the enthusiast sku is. You start with the same base as the consumer sku + an additional gpu or something or other.

A modular console was exactly what the Yukon was suppose to be, my idea is simply a more permanent approach.

 

[Acert93]

It is interesting research but not practical for a console currently for quite a number of reasons.

The largest obstacle is going to be the fact that this is very latency intolerant. Which means it requires very low latency between when the user's view shifts and when the display is updated. Which leads to the following issues...

[1] It requires an eye tracking system with very low latency. In this test they had to use a Tobii TX300 with an average eye tracking latency of 7-10 ms (image capture, processing, and then submitting to the PC). This isn't an inexpensive piece of hardware. The Tobii TX300 also has a limit of approximately 26". Fine for desktop work, not so much for a living room TV. A system that could not only accurately track a person's eyes AND do it quickly enough to provide the necessary low latency is going to be quite expensive.

[2] Rendering latency was 8-18 ms. As it approaches 18 ms, the quality of the rendering starts to suffer with the illusion breaking as you get to the point where the foveated center region falls out of the center of your view enough that it appears to "pop" in as the user's eyes move. That doesn't sound as bad as it could be since you gain some performance in saved rendering time. But not in saved gameplay (physics, AI, etc.) time. So it's "bad" but not that bad. 60 FPS allows less graphics rendering time savings allowed for increased graphics rendering than does 120 FPS.

[3] Scan out latency also varied from 2-8 ms.

[4] But the biggest potential deal breaker is that at the end of that chain you have the latency of the display itself. Considering that Microsoft won't be able to control what TV a user will use that latency can range anywhere from a low 3-6 ms (you basically need a relatively specialized TV for this) to as high as over 100 ms (TV video processing, A/V passthrough and potential processing there as well). The test used a PC monitor with a measured latency of 6 ms.

All of that means...

[1] is going to be prohibitive from a cost and implementation perspective.

[2] will put very hard constraints on the time a game can spend for rendering, AI, physics, etc. Basically it will require games to have a mandatory 60-120 FPS rendering performance. No more 30 FPS games.

[3] is the easiest to deal with and is just down to making sure the right graphics hardware is chosen for the console.

[4] is something that neither the game developer nor the console manufacturer will be able to control. And all that effort will go to waste if your average console player uses an average LCD TV.

This tech could do wonders on PC, however. Where gamers are used to picking monitors specifically with an eye towards a low latency display. It also makes eye tracking easier as the user is usually sitting within two to three feet of their display.

Very nice tech, but impractical in the living room currently and for the near (5-10 years) future.

Oh and forget multiplayer games on the same console with something like this. Forget having other people in the room watching you play as well, as their eyes are not likely to track the same as the user's eyes.

Regards,
SB

So the next "best" option is go with a faux approach; 1080p HUD and then go with a 1280x720 pixel native resolution in the center of the 1080p framebuffer and then the pixels outside the centered frame would be 1/2 or 1/4 resolution.

PANE#1: 720p resolution image centered on screen
PANE#2: 1/2 or 1/4 resolution
PANE#3: 1080p HUD*

Note that the HUD need not just be text; it could be the inside of a helmet (why not?).

An alt. option would be to stair step 480p-720p-1080p blocks with 100%, 50%, and 25% resolution.

This becomes a reasonable approach when looking at a couple 720p vs. 1080p images:

http://lockergnome.net/upfiles/preview_large_1.png
http://tasel.files.wordpress.com/2012/04/720p_vs_1080p.jpg
http://images.hardwarezone.com/upload/files/2011/03/edc157c0d3.jpg
http://obamapacman.com/wp-content/uploads/2012/03/iTunes-720P-vs-1080P-video-quality.jpg
http://www.lcd-tv-ratings.net/wp-content/uploads/720p-vs-1080p.jpg

Taking it one layer deeper: why does it need to be a 1:1 scaling? Why not a 2:1 or 3:2 scaling (H:V) or whatnot.

http://hv20.info/yopu/NeoHDvsNeoHDV.jpg

I think the illumi-room (or whatever it is called) is clearly an attempt by MS to bring Foveated Rendering to the console (I personally thought the images produced look trashy, but could add to peripheral immersion). I don't see why they couldn't--or won't--take it one step further. For same games it won't work well (e.g. Madden) but I could see in a shooter like Halo where you often are focused on the center of the screen why you couldn't or shouldn't break the screen up into planes where the center 50% of the screen is a 1:1 rendering and the peripheral edge (the other 50% or so) is a 2:1 or 4:1. If there are 3 planes it could even have a natural step down.

The biggest problem would be aliasing; I wonder how 4xMSAA with a post-process AA with some DOF and motion blur would hide this? I also wonder how once you put in all the work, how much it actually saves you? 1080p may be 2.25x as many pixels as 720p (125%) but if you do 720p and cut the other pixels say down to 50% more the question is what overhead did you have getting those 2 extra panes?

 

[ramr]

'
Now the consumer sku is basically what was seen in Yukon pdf dated 2010, minus the multiple cpus/gpus. It's a $299, or less, box that is 'good enough' for the average consumer. The people that are going to spend a lot of time using the media features rather than the gaming features. Basically the audience being aimed at by Apple/Google. The people that would be happy that they can play call of duty or their sports games or what not. don't care what resolution, don't get pissed at frame drops, screen tearing etc.

The consumer sku will be aimed at the average consumer who are the majority of the market, which a $299 or lower price makes since.

.

According to Yukon, the 3 soc machine was the $299 machine. The set-top box would have been just the APU.

 

[Betanumerical]

So the next "best" option is go with a faux approach; 1080p HUD and then go with a 1280x720 pixel native resolution in the center of the 1080p framebuffer and then the pixels outside the centered frame would be 1/2 or 1/4 resolution.

PANE#1: 720p resolution image centered on screen
PANE#2: 1/2 or 1/4 resolution
PANE#3: 1080p HUD*

Note that the HUD need not just be text; it could be the inside of a helmet (why not?).

An alt. option would be to stair step 480p-720p-1080p blocks with 100%, 50%, and 25% resolution.

This becomes a reasonable approach when looking at a couple 720p vs. 1080p images:

http://lockergnome.net/upfiles/preview_large_1.png
http://tasel.files.wordpress.com/2012/04/720p_vs_1080p.jpg
http://images.hardwarezone.com/upload/files/2011/03/edc157c0d3.jpg
http://obamapacman.com/wp-content/uploads/2012/03/iTunes-720P-vs-1080P-video-quality.jpg
http://www.lcd-tv-ratings.net/wp-content/uploads/720p-vs-1080p.jpg

Taking it one layer deeper: why does it need to be a 1:1 scaling? Why not a 2:1 or 3:2 scaling (H:V) or whatnot.

http://hv20.info/yopu/NeoHDvsNeoHDV.jpg

I think the illumi-room (or whatever it is called) is clearly an attempt by MS to bring Foveated Rendering to the console (I personally thought the images produced look trashy, but could add to peripheral immersion). I don't see why they couldn't--or won't--take it one step further. For same games it won't work well (e.g. Madden) but I could see in a shooter like Halo where you often are focused on the center of the screen why you couldn't or shouldn't break the screen up into planes where the center 50% of the screen is a 1:1 rendering and the peripheral edge (the other 50% or so) is a 2:1 or 4:1. If there are 3 planes it could even have a natural step down.

The biggest problem would be aliasing; I wonder how 4xMSAA with a post-process AA with some DOF and motion blur would hide this? I also wonder how once you put in all the work, how much it actually saves you? 1080p may be 2.25x as many pixels as 720p (125%) but if you do 720p and cut the other pixels say down to 50% more the question is what overhead did you have getting those 2 extra panes?

Effects that cross the boundaries are going to look pretty jarring and I know that I don't constantly look at the centre of the screen even in a fps.

 

[Silent_Buddha]

'low end' sku is your terminology. I said consumer sku. Microsoft wants to compete with Apple/Google for the living room, you're right. Which means gaming takes a back seat, which to me, seems exactly what has happened with the recent specs if legitimate. To me, it seems like they are taking a 'good enough' approach. Granted again, my technology knowledge is basic at best.

That is exactly the thinking which has led to Apple TV and Google TV (the two mentioned set top box competitors) basically having little to no success. They are a set top box first with gaming taking a back seat.

Microsoft is approaching it from another direction entirely. Depending on how someone wants to think of it. It is either...

[1] Gaming and set top box functionality share equal importance.

or

[2] Gaming is still the top priority but set top box functionality is a very close second.

In either case, gaming doesn't take a backseat because gaming is the biggest potential game changer in attempting to create mass consumer adoption of a set top box ecosystem.

You can think of it in another way.

You use Xbox and console gaming combined with an advanced and well polished living room media and application experience. It's enough to get people to jump in on a set top box experience because they aren't limited to just a set top box experience with crappy games.

Now, those people, just like with Kinect, will want to show it off to their friends, assuming they did a good job with the non-gaming aspect. That might convince some more people to get the Xbox, but that isn't the point.

The point is that it'll hook some people that in the past have been unsatisfied with Apple TV, Google TV, or any other of the plentiful Linux/Windows set top boxes. Or have never tried a set top box. But they may not be gamers, and may not be interested in the gaming experience and think why pay so much money?

That's when 6 - 12 months later it'll be an opportune moment for Microsoft to release a set top box without the gaming functionality assuming they did a good job with that functionality in the Xbox.

IMO, that's a much more likely scenario for how Microsoft intends to approach the set top box market.

In many ways you can think of this as similar to the PS2 and PS3. In those cases the console was a medium to drive adoption of DVD and BluRay, but gaming never took a back seat to those things even though in the case of PS3 it likely impacted design decisions (cheaper RSX versus a more proficient GPU, for example). While Durango won't be trying to drive adoption of some optical media, it could be attempting to drive adoption of Microsoft backed set top boxes of the future.

That set top box without the console gaming bit (it'll likely still have the casual game app type stuff you can find in the Microsoft, iOS, Android marketplace) could bear the Xbox branding as well. Which fits in nicely with the rumors of multiple Xbox branded devices. Similar in many ways to how Zune was re-envisioned as Xbox Music. There's also the possibility, however small, of an Xbox branded Windows HTPC with the same set top box UI combined with the ability to play proper PC games. I find the last pretty unlikely but it would be neat if it happened, IMO.

Regards,
SB

 

[expletive]

So the next "best" option is go with a faux approach; 1080p HUD and then go with a 1280x720 pixel native resolution in the center of the 1080p framebuffer and then the pixels outside the centered frame would be 1/2 or 1/4 resolution.

PANE#1: 720p resolution image centered on screen
PANE#2: 1/2 or 1/4 resolution
PANE#3: 1080p HUD*

Note that the HUD need not just be text; it could be the inside of a helmet (why not?).

Are you saying this:

PANE#1: 480p DOF background 'painting' (cant think of a better word for it) scaled to a full 1080p frame. This pane is only visible when parts of PANE#2 and PANE #3 are transparent on the same pixel.

PANE#2: 720p Primary Game graphics PANE scaled to a full 1080p frame and overlaid on top PANE #1. Transparent pixels on this PANE show what is on PANE#1.

PANE#3 1080P HUD overlaid on top of PANES #1 and #2.

Was there some custom hardware for a robust scaling engine that could handle all these resolution modifications for the multiple panes?

 

[Shifty Geezer]

Here's a link for those interested: http://research.microsoft.com/pubs/176610/foveated_final15.pdf

MSR used a Tobii TX300 camera to track the eyes with 10ms latency.

Good info.

Not sure why you'd need 'microscopic' level precision here, just millimeter scale should be fine which iirc the patents suggest the new Kinect device can do.

I used the term 'microscopic' loosely. A more accurate term would be 'very high resolution'. That or a zoom camera. The Tobii system is a 'desktop' unit to be placed a few feet in front of the user, who remains stationary. Kinect's wide-angle camera wouldn't be up to the job. Potentially MS could add a zoom camera on a motorised mount and track the eyes, but that's a lot of cost. You also have to consider the value of a system designed around a one-player experience for its intrinsic optimisations, especially if you bundle in Kinect for its social aspects.

I do have a question for you though Shifty, you've mentioned something about TBDR. I'm a physicist and know some tech areas but very little when it comes to computer engineering. What is TBDR and how might it fit into the puzzle of Durango's strange design decisions?

Brit's covered that for me.

 

[(((interference)))]

I don't want to believe in #4...

The palatability of a belief does not determine it's truth value.

You have to ask yourself, since MS and Sony did not know what each other were doing - how likely is it that both machines will end up 'equal', besides the rare situation this gen it has usually been the case that consoles in a particular gen differ in capabilities.

MS wanted 6-8x the power of 360 and something that can run Win 8 apps, IE, Skype, DVR, Smartglass etc in the background and come with Kinect as a pack in, all while being affordable and not sold for a loss.

The rumoured Durango specs will achieve those goals perfectly.

 

[(((interference)))]

Welcome this faveated tech..sounds interesting. .and if im honest too good to be true...any info I can read? Thanks

Isn't it a problem for foveated rendering, that to everyone else watching, besides the player, the game might look crap?

 

[Jubei]

'low end' sku is your terminology. I said consumer sku. Microsoft wants to compete with Apple/Google for the living room, you're right. Which means gaming takes a back seat, which to me, seems exactly what has happened with the recent specs if legitimate. To me, it seems like they are taking a 'good enough' approach. Granted again, my technology knowledge is basic at best.

Now the consumer sku is basically what was seen in Yukon pdf dated 2010, minus the multiple cpus/gpus. It's a $299, or less, box that is 'good enough' for the average consumer. The people that are going to spend a lot of time using the media features rather than the gaming features. Basically the audience being aimed at by Apple/Google. The people that would be happy that they can play call of duty or their sports games or what not. don't care what resolution, don't get pissed at frame drops, screen tearing etc.

The consumer sku will be aimed at the average consumer who are the majority of the market, which a $299 or lower price makes since.

But there are still millions of people out there that are not an average consumer. People you consider enthusiasts who do not want 'good enough' hardware, but want top of the line hardware. They might be pushed away by the general consumer/casual approach where gaming is the back seat. So that is where the enthusiast sku is. You start with the same base as the consumer sku + an additional gpu or something or other.

A modular console was exactly what the Yukon was suppose to be, my idea is simply a more permanent approach.

You keep missing the point. The hardware they have chosen for Durango is too expensive for your scenario to be plausible, its not going to make them any profit at under 299 dollars, i have doubts it would make them profit even at 299 dollars given the esram and Kinect 2

It makes no sense for them to use custom hardware like this if they wanted a 200-249 dollar console, they could easily have chosen off the shelf hardware that would perform enough for their target audience and still make Microsoft profit from day 1

 

[SlickShoesRUCrazy]

You keep missing the point. The hardware they have chosen for Durango is too expensive for your scenario to be plausible, its not going to make them any profit at under 299 dollars, i have doubts it would make them profit even at 299 dollars given the esram and Kinect 2

It makes no sense for them to use custom hardware like this if they wanted a 200-249 dollar console, they could easily have chosen off the shelf hardware that would perform enough for their target audience and still make Microsoft profit from day 1

according to the Yukon pdf, the estimated cost was about $225 bucks, and that's a 3xSoc design.

They would be making a profit even at a $249 pricepoint. I think you are over-estimating the cost of the parts that is going to be in the Durango.

From what we have seen of the gpu, if legitimate, is very comparable to a HD7770. Very inexpensive GPU, especially if MS would be buying parts in bulk + the removal of unnecessary stuff like vram.

 

[Betanumerical]

according to the Yukon pdf, the estimated cost was about $225 bucks, and that's a 3xSoc design.

They would be making a profit even at a $249 pricepoint. I think you are over-estimating the cost of the parts that is going to be in the Durango.

From what we have seen of the gpu, if legitimate, is very comparable to a HD7770. Very inexpensive GPU, especially if MS would be buying parts in bulk + the removal of unnecessary stuff like vram.

225 for the 3 SoC is no where near realistic. How is 3 SoC cheaper then the ps3 and 360 it would be huge ?