Predict: The Next Generation Console Tech

[XpiderMX]

If there existed tech to make a 1.2 TF GPU perform like a 3TF one, why the hell wouldn't AMD be using it already in PC?

And pseduo-GPU? Why not just build a bigger regular old GPU? It begs the question...

Seems dubious, though I'm ever irrationally hopeful

I'd bet any customizations are more subtle help than huge, but again I hope I'm wrong.

I don't know if it is plausible, but I guess that with a configuration like this, the system can work in "low power" when we are not gaming.

 

[Proelite]

If there existed tech to make a 1.2 TF GPU perform like a 3TF one, why the hell wouldn't AMD be using it already in PC?

MS money needed to design and build it perhaps? Maybe MS just wants it because they know what they need.

New hardware also migh break backwards compatability and perf. AMD won't be releasing a card that doesn't perform signifcicantly better for old titles.

 

[Inuhanyou]

Heavily customized hardware like this is probably much easier to provide in a console tailored to developers using those exact specifications and eccentricities compared to a normal "off the shelf" gpu that can be purchased and must work with any configuration.

That is of course assuming this is all true.

 

[McHuj]

If there existed tech to make a 1.2 TF GPU perform like a 3TF one, why the hell wouldn't AMD be using it already in PC?

And pseduo-GPU? Why not just build a bigger regular old GPU? It begs the question...

Seems dubious, though I'm ever irrationally hopeful

I'd bet any customizations are more subtle help than huge, but again I hope I'm wrong.

Maybe it's something that isn't transparent to the programmer and thus doesn't fit in nicely within current PC programming models. Therefore it wouldn't be utilitized effectively until something like a DX12 comes along and AMD is waiting to introduce it in the next architecture, but Durango is just the first product where it first appears.

 

[Inuhanyou]

It does make sense though, it jives with the rumors that have been continuously saying that AMD considers the 720 the "higher priority" product and that Durango would be a pretty specialized unit.

 

[ERP]

If there existed tech to make a 1.2 TF GPU perform like a 3TF one, why the hell wouldn't AMD be using it already in PC?

And pseduo-GPU? Why not just build a bigger regular old GPU? It begs the question...

Seems dubious, though I'm ever irrationally hopeful

I'd bet any customizations are more subtle help than huge, but again I hope I'm wrong.

I could imagine a GPU with no ROPS changing the ALU design in such a way to be more amenable to compute work loads. Possibly smaller wavefronts etc etc.

Having said that it doesn't match the rumors I've heard, but I don't have any definitive information about Durango.

 

[anexanhume]

I could imagine a GPU with no ROPS changing the ALU design in such a way to be more amenable to compute work loads. Possibly smaller wavefronts etc etc.

Having said that it doesn't match the rumors I've heard, but I don't have any definitive information about Durango.

But where's the need for that compute specific power?

 

[babybumb]

It does make sense though, it jives with the rumors that have been continuously saying that AMD considers the 720 the "higher priority" product and that Durango would be a pretty specialized unit.

Just like the Xenon/Cell stuff.. MS needs more from a 3rd party than Sony. This high bandwidth APU is in house Sony presumably

 

[Inuhanyou]

I doubt Sony wants another Cell fiasco, i just don't see it.

 

[XpiderMX]

MS needs more from a 3rd party than Sony. This high bandwidth APU is in house Sony presumably

No AMD?

 

[Inuhanyou]

I really doubt its in house. All rumors have been pointing to an AMD joint venture.

 

[ERP]

But where's the need for that compute specific power?

Depends on your view on GPGPU.
But even just for graphics, for various reasons Vertex processing is extremely inefficient when it comes to ALU usage on PC GPU's, wavefronts are only filled in unusual cases.
As I said in the original post, the existence of such a device doesn't match what I've heard.

 

[Blazkowicz]

Sony could make an in house GPU in 1999 for its PS2, but not today. The complexity of a current, big GPU makes me puke. Even the drivers are a nightmare (it should be a ton easier on a console. But you still need that shader compiler and stuff)
It would have been a billion dollar effort launched many years ago.

Even for the PS Vita, they simply used a PowerVR.

 

[Docwiz]

One thing I was thinking about is that many of the same people that worked on the 360 design are also working on the Next Xbox (Durango) design. We know that the Xbox 360 was a very balanced design, so I would expect the Durango to be the same.

So, the entire memory speed slowness part on the Durango doesn't make sense.

 

[Inuhanyou]

what "memory speed slowness"?

 

[3dcgi]

About "higher priority project for AMD",i saw a may not important but pretty interesting respond by #1 AMD china guy,from yesterday
"倒是接微软的这个活让大家都能活的滋润些，所有费用都是微软出，包括工资。。。感觉是个微软的外包了。。。农企也觉得这样很好，打算成立个新的BU了。。。

一句话，农企彻底变打工卖苦力的了"

He said "Microsoft pay for them everything,even including wages,he feel like he is(or AMD Durango team is) Microsoft's outsourcing team now,but AMD think that's good,and and intends to set up a new BU(what is BU?)

In short,AMD become a company working for someone"

This is the case for any console project. If a customer wants customizations they pay NRE (Non-Recurring Engineering) costs.

EDRAM in the APU might be a possible cause of yield problems, since it would be a first time for AMD.

Did you forget there was EDRAM in Xenos?

 

[JasonLD]

More I read the rumors about Orbis and Durango, I think they will be even closer than it had been for PS3/360, especially when both GPU seems to be heavily based on Sea Island for the most part.
I don't expect any miracles lol.

 

[Lucid_Dreamer]

Just like the Xenon/Cell stuff.. MS needs more from a 3rd party than Sony. This high bandwidth APU is in house Sony presumably

These theoretical questions are for anyone. If the "GPU" portion of the APU is to be used for GPGPU functionality (in Sony's solution without AMD), how many SPEs with crossbar could fit in that area? It would still be a shared project with AMD, because of the steamroller cores and discrete GPU, right?

Could the 1.8 TFLOPs number be just for the discrete GPU? Could this be related to the 3.2GHz clock rumored for the Steamroller cores? Would that be able to make backwards compatibility simple for a next-gen PS console?

Finally, what would be the primary differences (from a function and performance standpoint) between full HSA and a Steamroller/SPE/discrete GPU setup?

 

[Rangers]

Sony could make an in house GPU in 1999 for its PS2, but not today. The complexity of a current, big GPU makes me puke. Even the drivers are a nightmare (it should be a ton easier on a console. But you still need that shader compiler and stuff)
It would have been a billion dollar effort launched many years ago.

Even for the PS Vita, they simply used a PowerVR.

Yup, probably not even to mention the patent situation. Patents alone would probably make it impossible I'm guessing.

 

[Docwiz]

what "memory speed slowness"?

The rumors point to DDR3 and slow access speeds to the ram.