News & Rumors: Xbox One (codename Durango)

[Rangers]

For what it's worth one of my mystery GAF pmers said he hasn't heard of any dual APU.

I'm not sure I'd even like that idea, a couple thoughts:

would two APU's, then you have to split the RAM right (though I guess DDDR3 is cheap enough you could think about doubling?) at 2.4 TF, really even be better than a single 1.8?

I'm going to say yes simply because in a console you'll really tap the hardware.

But SLI adds a frame of latancy AFAIK, and I dont really want that in my console. Not sure I wouldn't prefer a nice simple upclock even if it wouldn't melt the internet down. Joe GAF Casual would freak out over two APU's as it's simple enough for them to understand, it'd be wild

They would need an external multi-processor bus, so they would have planned to have more than one APU from the start.

Crossfire isn't more than glomming two completely discrete GPU's together (however they do it), is it? You would assume Durango being an AMD GPU would have that built in?

 

[Rangers]

spotted on GAF. Unknown by me if real. Found on twitter I guess. Could we be seeing inside the famously reclusive tent?

Something to get hyped I guess. Though those gfx look bad if you can see them.

Edit: It actually looks like a placard over the screen?

Edit 2: And are those even TV's? Lol. Some guy could have mocked this up in his bedroom.

 

[Xenio]

But SLI adds a frame of latancy AFAIK, and I dont really want that in my console. Not sure I wouldn't prefer a nice simple upclock even if it wouldn't melt the internet down. Joe GAF

SLI and crossfire are solutions that interconnect GPU's, not APU's.. Jokes aside and just for the fun of talk, doubling the APU I think means doubling eSram too, this will be too expensive;
but in wet dreams maybe some sort of external blitter/move engine can create a virtual "64 MB shared Memory" with two 32 MB eSram blocks, in which both APU's can read and write.

I don't think that this would means to double the assets as in SLI/XFire uses

 

[AlphaWolf]

Edit 2: And are those even TV's? Lol. Some guy could have mocked this up in his bedroom.

I don't think those are tv's. The couch is facing the other way.

 

[Xenio]

I don't think those are tv's. The couch is facing the other way.

Agreed, but the couch is facing the other way to the projection wall too
maybe they are yet working on the room layout

 

[MTd2]

I'm not sure I'd even like that idea, a couple thoughts:

There's a solution. Get an "off the shelf" 8XXX, with a few more Tflops, connect to the APU, do a show on 21st with some demos, and pray that it can work and mass produced under 300W until the end of the year.

 

[Shifty Geezer]

Microsoft does have a track record of making huge turnarounds at the last minute, like they did in windows 95. If it were me, and I were in this position, and I wanted to make sure my console were the most powerful, and I was pressed for time, I would double up on the APU. Changing the motherboard design is a much smaller and well understood problem than redesigning an APU at the last minute. It also allows game developers, if they have designed their game well, to update to the new design with minimal issues.

This would be a serious, knee-jerk reaction though, Dual APUs would come at dual cost for less than twice the performance. Devs will have 16 CPU cores and two pools of eSRAM, and all sharing the same 60 GB/s DDR3 bus. It would be a panic response with severe consequences and questionable returns.

If MS do want to adjust their original gameplan, the easiest improvement would be clocks. They may get another 10%+ on speed without having to redesign their cooling solutions much. I don't think more RAM would gain them any benefit other than a marketing number. One area not touched upon is storage. What if Durango has an SSD? Maybe IO performance for virtual texturing/meshes could be enhanced, which wouldn't be in opposition to any existing rumours. Other than that, I'm seeing little room to manoeuvre regards silicon, and any major changes would be dependent on MS executing the most incredible misinformation campaign in history, to the detriment of Durango game developers.

 

[Brad Grenz]

For what it's worth one of my mystery GAF pmers said he hasn't heard of any dual APU.

I'm not sure I'd even like that idea, a couple thoughts:

would two APU's, then you have to split the RAM right (though I guess DDDR3 is cheap enough you could think about doubling?) at 2.4 TF, really even be better than a single 1.8?

I think trying to string together two Durango APUs that can barely talk to each other would be worse is basically every conceivable way than the announced PS4. It would just be a development nightmare that still can't match PS4 performance and now costs twice as much.

 

[MrFox]

Seriously I think that, as improbable as it is, a dual-mode DDR3/GDDR5 memory controller is less far fetched than a last-minute dual APU or APU+GPU.

 

[Xenio]

still can't match PS4 performance

why 24 CU's and 64 MB of eSRAM + 16 Core's Jaguar + SHADE DSP can't?
maybe at the first start 2 apu's don't means double performances of single apu, but I think this would be however a lot faster than a 18 CU, no eSRAM and 8 core Jaguar, even with unoptimized driver layer, too much brute force, well beyond 2.5 TF ( 1.2x2 gpu's + 100x2 CPU's +100 TF SHADE DSP= 2.7 TF am I right?)

 

[Brad Grenz]

why 24 CU's and 64 MB of eSRAM + 16 Core's Jaguar + SHADE DSP can't?
maybe at the first start 2 apu's don't means double performances of single apu, but I think this would be however a lot faster than a 18 CU, no eSRAM and 8 core Jaguar, even with unoptimized driver layer, too much brute force, well beyond 2.5 TF ( 1.2x2 gpu's + 100x2 CPU's +100 TF SHADE DSP= 2.7 TF am I right?)

Because it's not 24 CUs, 64MB of eSRAM and 16 Jaguar cores. It's half of each of those things spread across two chips with no easy way to either share data or spread the work across both. It would be so inefficient to try and juggle 3 or 4 memory pools, 2 CPU banks and 2 sets of move engines and 2 seperate GPUs that at a certain point you'd probably get better performance by simply turning one of the APUs off and concentrating on the other chip, using it the way it was designed to be used.

 

[MrFox]

why 24 CU's and 64 MB of eSRAM + 16 Core's Jaguar + SHADE DSP can't?
maybe at the first start 2 apu's don't means double performances of single apu, but I think this would be however a lot faster than a 18 CU, no eSRAM and 8 core Jaguar, even with unoptimized driver layer, too much brute force, well beyond 2.5 TF ( 1.2x2 gpu's + 100x2 CPU's +100 TF SHADE DSP= 2.7 TF am I right?)

Look at the schematics we have of Durango, duplicate it... and try to connect them together with a slow external bus.

The problem is that it's a clusterf^&* of bottlenecks. For that amount of silicon they would be better off with off the shelf CPU and GPU.

 

[bkilian]

Like I said, just speculation. It does introduce a ton of issues, so it is not too likely. Any inter-APU communication would have to be kept to a minimum, so keep one 4 core module for system, one 4 core module for, say, compute, or audio effects not possible in hardware, or physics, maybe, and then the other two modules (on one apu) for the game use. GPU interaction would also be complicated, so probably a bad idea in the long run.

Another option would be APU + discrete GPU, keeping the 12 CUs on the APU for compute and physics, but again, cost here will be huge. I'm betting no change at all. If they do change, it'll be a redesign, with a corresponding slip into 2014. Not something anyone wants to see, methinks.

 

[James Car]

One thing I think you guys are overlooking is the fact that this is x86 hardware with a common/standard GPU.

The simplest (last minute) solution would be to simply pull off-the-shelf components. The silicon budget would allow for a fairly aggressive component selection and given the msrp last gen, I wouldn't be surprised to see $400-$500 as the staggered offerings.

The fact that MS has instructed developers to to code to api instead of "to the metal" leaves them completely open to changing the spec last minute (within reason).

Another option/possibility is that they've had two designs all along. Depending on which way Sony went and which way the market went in general, they could decide to go low spec, or high.

This would not be unheard of (Heck, Sega did this back in the day for Dreamcast) and again, they've had their developers coding to api's not to the metal. So any subsequent "upgrade" from the baseline Durango spec would not render their efforts useless. It would just open up the headroom for developers to push their efforts further in subsequent releases past the initial wave of launch software which is typically not pushing the hardware anyway (see every past console launch).

Just something to consider.

 

[Brad Grenz]

There doesn't even seem to be any offchip communication path fast enough for an APU + GPU bandaid solution. I mean, that was the whole point of using an APU in the first place, to keep that highspeed connection on a single chip. Are you gonna hang the discrete GPU off the south bridge? That will be really slow. Even throwing away the APU and using a 4 core Bulldozer and a 7870 would arguably be worse than the PS4 design in many respects.

 

[Rangers]

I mean, that was the whole point of using an APU in the first place,

Pretty sure the point was "reduce cost"...

 

[expletive]

What if they had an alternate, more traditional, non-SOC design where they could more easily swap out the key components like CPU and GPU and then move to an SOC in the first hardware revision?

 

[James Car]

Even throwing away the APU and using a 4 core Bulldozer and a 7870 would arguably be worse than the PS4 design in many respects.

It would be worse in many respects, but it would also be superior in many respects as well.

The best thing about choosing x86 hardware in general is the flexibility in design time to go with whatever design seems necessary at the last minute. With the world and the game market being the way it is, it doesn't surprise me that they would head in this direction of x86 if for nothing else, the flexibility it brings to last minute spec changes.

Granted, not as nice/efficient as an apu designed from the ground up, but if their research is telling them that the durango apu spec as we know it is riskier than going with a last minute change-up consisting of off-the-shelf components, then they could choose to go with off-the-shelf components.

R&D costs are not cheap, but neither is a losing proposition at retail.

 

[Xenio]

Pretty sure the point was "reduce cost"...

the xbox 360 came in my mind, remember when it change last minute from 256 MB of GDDR3 to 512? this was very costly at its time, as microsoft admits, but they were force by EPIC
they used to make those kind of things, maybe they have done another last minute change

what if the mainboard was flexible enough to permit DDR3 or GDDR5, and one/two APU?
it's reasonable that they're gone with this first type of solution to be able to such last minute change "reaction"? future revisions can shrink component and cut useless features to save the cost

 

[eastmen]

wouldn't the easiest thing be just to add another dedicated gpu and its own ram pool. 4gddr 5 added to 8 gigs of ddr 3 = 12 gigs as rumored