News & Rumors: Xbox One (codename Durango)

[BoardBonobo]

He's just digging himself, and the Xbone, in deeper with each statement, and putting numbers to it is just plain stupid. At some point MS PR has to get a hold of this and gag him.

On paper the PS4 has an advantage, in real terms this advantage will still be there. It's just the expressed objective amount will be different but whether that will affect the subjective experience is entirely down to the developer and how they utilise the different systems.

 

[BRiT]

*AHEM*

Reminder # 9001:


Do any of you see Sony or PS4 or Versus in the topic? No. So do your best to keep non-topic items out of this, especially the direct versus talk.

 

[DaveNagy]

The two points are related. When you add cores, you always add some level of overhead to handle the extra core. Two cores can't do twice the work of one core, even in a highly parallel problem, since the are always places where a bottleneck occurs. Let's say each added core makes all your cores 99% as efficient as they were before you added the new core. At 12 cores, you're looking at each core being about 88% as efficient as it would be by itself. At 18 cores each core would be 83% of its best. This is the same reasoning behind why beefing up all your cores could be better than adding a new core.

Wow, that makes a lot of sense. Thanks. Al should have said that.

 

[3dcgi]

Out of curiosity, and I'm not a graphics programmer, is it not possible to pipeline the graphics work such that vertex processing for frame "n+1" is happening while ROP's are processing pixels for frame "n"?

Are there in effect two sets of "efficiencies": 1. Of the time running shader code, how much of the shader's processing is utilized using useful work. and 2. Over the course of frame, how often is each unit (shader/rop/etc) active/inactive.

It's possible to have multiple frames active on the GPU like you say, but so much work goes into a single frame that you'll usually only have a single frame active as the overlap between frames will be small.

You can take this down a level to overlapping draw calls within a frame. Draw 1 might be rendering to the color buffer with a complicated pixel shader while draw 2 is rendering to a shadow map. Draw 2 is more likely to stress vertex processing as there are no pixel shaders for it and could have good overlap with draw 1. There might also be a bunch of shadow map draws after draw 2, but at some point you can't buffer up any more vertices in the GPU until the pixel shaders from draw 1 finish. How much buffering there is depends on the GPU implementation.

 

[ERP]

Out of curiosity, and I'm not a graphics programmer, is it not possible to pipeline the graphics work such that vertex processing for frame "n+1" is happening while ROP's are processing pixels for frame "n"?

Are there in effect two sets of "efficiencies": 1. Of the time running shader code, how much of the shader's processing is utilized using useful work. and 2. Over the course of frame, how often is each unit (shader/rop/etc) active/inactive.

One thing Sony has discussed that tries to address this is the idea of using compute to do the bulk of the transform work. Compute shaders don't have the same issue, and you can reduce the data sent into the vertex shader which is the bulk of the issue.
It does however use additional bandwidth, and can increase latency depending on how it's done.

 

[runwave87]

Are there any launch day multiplatform games targeting less than 60 FPS at the moment? The only ones I'm aware of that have mentioned FPS are all targeting 60 FPS.

Regards,
SB

I thought I read or saw in a video interview that Dead Rising 3 will be 30fps. Granted with that many zombies and the level of quality graphics shown in game I am perfectly happy with a locked 30fps

 

[Jwm]

I thought I read or saw in a video interview that Dead Rising 3 will be 30fps. Granted with that many zombies and the level of quality graphics shown in game I am perfectly happy with a locked 30fps

That seems to be best info I read. MS is also working closely with them to get it running at a proper fps.

 

[bagofsuck]

That seems to be best info I read. MS is also working closely with them to get it running at a proper fps.

Panello said that after they update their tools + specs bump the performance from DR3 PAX demo was much better than the demo they showed at Gamescom.

 

[XpiderMX]

Panello said that after they update their tools + specs bump the performance from DR3 PAX demo was much better than the demo they showed at Gamescom.

The framerate looks better than GC demo: https://www.youtube.com/watch?v=z579CJwFjYI

 

[Rangers]

I thought I read or saw in a video interview that Dead Rising 3 will be 30fps. Granted with that many zombies and the level of quality graphics shown in game I am perfectly happy with a locked 30fps

He said multiplatform though

From this list http://www.ign.com/wikis/xbox-one/Xbox_One_Launch_Titles

I dunno. A lot of the multiplats are sports franchises which are probably 60. Also BF has now gone to 60 and of course COD. AC4 i think is 60 too? Although I think that's just a legacy of being a current gen port.

As the gen goes on I expect most multiplats to be 30 though. Not the sports games, but many others.

 

[warb]

Not only that, but going deeply it's also a confession of inferior performance.
Which he shouldn't be putting himself in the position to confess in the first place.

He didn't say which way.

 

[Strange]

He didn't say which way.

Oh we all know which way it is...

 

[expletive]

CU's are MASSIVELY underutilized on vertex heavy workloads and plenty of the frame will be ROP or bandwidth limited.

Are the Depth/Color blocks the nomenclature for the ROPs on the XBO or are they something different?

 

[AlNom]

Are the Depth/Color blocks the nomenclature for the ROPs on the XBO or are they something different?

Yes.

quad ROP -> RBE -> CB + DB

 

[XpiderMX]

Ars Technica talinkg about Penello statements:

http://arstechnica.com/gaming/2013/...s-xbox-one-from-accusations-its-underpowered/

 

[AlphaWolf]

Ars Technica talinkg about Penello statements:

http://arstechnica.com/gaming/2013/...s-xbox-one-from-accusations-its-underpowered/

I wouldn't call that a response.

 

[Cjail]

Did Panello double the original 102Gb/s bandwidth of the ERSAM because he says that ERSAM can read write simultaneously?!

 

[3dilettante]

Did Panello double the original 102Gb/s bandwidth of the ERSAM because he says that ERSAM can read write simultaneously?!

The saga of the eSRAM isn't as straightforward. There is a near doubling from concurrent operations and an upclock included in those numbers.
The Ars snap judgement on that statement is more off-base.

 

[Shifty Geezer]

CU's are MASSIVELY underutilized on vertex heavy workloads...

Going OT, but doesn't that somewhat suggest a return to discrete vertex shaders would be better silicon utilisation? Or can we adjust vertex shader workloads to do more stuff? Or can compute fill in the blanks?

 

[blakjedi]

I wouldn't call that a response.

Not at all.