Xbox One (Durango) Technical hardware investigation

[brasnial]

Hi guys
I'm totally new......just a interested party with a question for the more informed out there ......why would Microsoft put 8 gigs of ram in there machine if they didn't need it and were not able to use it .
Would that not be a case of over kill and a waste of resources........
So what I'm asking is this if we assume neither company Sony or Microsoft are fools and no what there doing ...what in your opinion would Microsoft be looking to do with 8 gigs of slow ram...?

Sorry if my question seems dumb but no one puts stuff into a machine that the don't think they need or don't intend using ...

 

[McHuj]

1.22TF is suspiciously the exact number of cape verde with 12CU@800MHz
I think that MS will not take an off the shelf architecture without making some enhancing, or picking something from the next architecture
It looks like it was made for inspiring "look! the numbers add up! It must be real!"

Flops are really only a measure of the number of multiply-adds that the GPU does, they say nothing else about the rest of the ISA. It would not surprise me if MS decided to add instructions to the ISA that they have in mind and would help with future directX releases.

Perhaps there are common instruction patterns in GPU code and some instructions can be combined into one. That's not always possible, but when it is it makes the GPU more efficient. I would expect MS research to be pretty forward thinking about future GPU usage and how rendering algorithms may evolve over the next 5 years or so, whereas AMD only needs to provide the best solution for the next year or 2.

It has to be something stronger but downclock.

Cape Verde 7770 GHZ has a performance of 1.2TF which is the reason why people are comparing it,but it also has a higher clock speed than Durango,so my guess is something higher down clocked..

What do you think.?

On paper they seem very similar in terms of raw performance, I would speculate that 12 CU's at 800MHz will be more power efficient than 10CU's @ 1GHz and that's why they are going with that configuration.

 

[brasnial]

Hi me again
I just want to add to my earlier question ....so if we believe Microsoft is using a 1.2 Tf gpu like the leaks state why would you pair it with 8 gigs of ram ........?
I have never heard of any PC gpu of the same spec being paired with this much memory ........have any of you .
So my question still stands let's forget about the power of the gpu and think about why Microsoft a software company would to pair a 1.2TF gpu with 8 gigs of memory .
It just seems such a unusual set up unlike any think I've read about or heard of before on PC .

 

[fehu]

Cape Verde has 10 CUs.

In fact I specified a cape verde with 12CU@800MHz

 

[V3]

Hi me again
I just want to add to my earlier question ....so if we believe Microsoft is using a 1.2 Tf gpu like the leaks state why would you pair it with 8 gigs of ram ........?
I have never heard of any PC gpu of the same spec being paired with this much memory ........have any of you .
So my question still stands let's forget about the power of the gpu and think about why Microsoft a software company would to pair a 1.2TF gpu with 8 gigs of memory .
It just seems such a unusual set up unlike any think I've read about or heard of before on PC .

On PC you don't have access to a 32MB of eSRAM. No dedicated PC GPU have that. In the future, when RAM chips reached certain densities, you will see 1.2TF PC GPU paired with 8GB of RAM, just for marketing purposes.

 

[AlNom]

Which is better - Orbis or Durango

 

[bgassassin]

I definitely look forward to learning more about the DMEs. To be honest when looking at the name and some of the explanations, they sound to me right now more like something to maximize the performance of the CPU, GPU, and memory, not something that adds to whatever performance those components may lack if that makes sense.

 

[anexanhume]

I definitely look forward to learning more about the DMEs. To be honest when looking at the name and some of the explanations, they sound to me right now more like something to maximize the performance of the CPU, GPU, and memory, not something that adds to whatever performance those components may lack if that makes sense.

Would it possibly be correct to think of it as a scheduler (at least in part) for the entire system that is coherent to main memory?

 

[Prophecy2k]

Would it possibly be correct to think of it as a scheduler (at least in part) for the entire system that is coherent to main memory?

I doubt it would be called a "Data Move Engine" if its as fancy as some kind of HW scheduler.

Imho its just a DMA engine, and becuase of second hand rumours of people crying "secret sauce", together with people like Aegis on NeoGaf who even admits he's not a HW engineer and has no clue about what he's seeing in the docs in his possession, people are trying to see too much into what are essentially quick-fix bandwidth saving features to mitigate the slow main memory bandwidth.

In short MS wanted 8GB main RAM. DDR3 was their only option (perhaps they evaluation HMC and stacking with TSVs/Interposers but realised it wouldn't be ready in time). They went with DDR3, but needed a high-bandwidth scratchpad and memexport engine in order to ensure their main components weren't bandwidth starved. Joe public and Joe gaming journalist catches wind after months of being drip-fed false rumours of MS's nextbox being "beast" by overexcited devs and internet trolls, and now start trying to read too much into and rationalise some "secret sauce" and magic voodoo out of what is effectively a relatively low-cost/low-perf console design. Acert93 and Interference are both right.

 

[bgassassin]

Would it possibly be correct to think of it as a scheduler (at least in part) for the entire system that is coherent to main memory?

Yeah. I think it was mentioned a little while back (working off of a few hours of sleep so I'm not all there) in comparing the design to HSA as maybe MS' take on it or something like that. A quick peak at an AMD HSA doc talks about how the scheduler can be handled in software, hardware, or both. Maybe MS decided on a full hardware approach. Speculation obviously.

 

[3dilettante]

If the fast memory pool isn't treated as a cache, I would lean towards it being a software (at least partly) controlled scratchpad. Filling and committing data to and from the primary address space would be handled by the CPU/GPU threads using it.

The scratchpad has some unknown granularity and banking. If not kept coherent, it could be composed of a number of large banks, possibly at the granularity of a memory page.
DMA engine(s) could offload most of the grunt work of arbitrating access and moving data back and forth to this pool of memory.
In this scenario, the SRAM/eDRAM pool is on a parallel portion of the non-coherent bus used by the GPU.

The data engines would have ancestry in the DMA engines discrete GPUs have had for quite a while, or the DMA engines used in a number of server and HPC designs. It would save time and power compared to having a CPU or GPU sending commands and reading in scratchpad memory back into coherent memory space before writing it back out, then reading in new data and then exporting it back to the scratchpad.
A DMA engine for the GPU, a DMA engine for the CPUs, and maybe a DMA engine for everything else.

It's not known how many independent accesses the memory pool can support in that sceneario. Even if it's not three, accellerating and offloading all the little moves and access negotiations implied in managing such a large memory space might make the scratchpad more easily utilized.

 

[brasnial]

If we assume both Sony and Microsoft's next Gen spec's are roughly correct what type of game will benefit from the differences in memory size .....?

As a gamer I don't view kinect as a bad for the industry nor the wii for that matter any think that brings in non gamers is good for the industry especial if some of these new gamers turn into more hardcore gamers .

It seems to me that to stick with the old idea of more power better graphics more sales is to live in the past .we now live in a world where doing every think we'll at a price your customers are willing to pay is the key to more sales .

To lose sight of this and get engaged in a pissing contest over who can push the most Pixels and there for base all you business decision's around this is a recipe to lose market share in my opinion .

 

[eastmen]

Good thing for you is that if these specs tell the whole story, MS will launch dirt cheap. The PPT clearly stated the goal was $299 with positive GM. Those specs fit that bill nicely.

$300 is also the target for the retail occulus rift. I rather get that. I rather get a new video card or a new tablet. $300 is the same price if the 360 back in 2005 and that brought a whole lot of kit comparied to this

 

[liquidboy]

What an interesting patent pic...

 

[Pugger]

What an interesting patent pic...

Looks like my fuse box. Who's the leak?

 

[3dilettante]

It's a patent, those don't need to be leaked.

 

[jbq.junior]

So, if Durango GPU has 12 CUs, given the GCN architecture, it may be able to output only 1 primitive/clock, like Cape Verde, right?

 

[ERP]

So, if Durango GPU has 12 CUs, given the GCN architecture, it may be able to output only 1 primitive/clock, like Cape Verde, right?

It would depend on how the input is configured.
it "could" certainly be limited to 1prim/clk, or it could be more these things aren't off the shelf parts jammed onto a piece of silicon. If MS thought they need more I can't imagine that level of customization would be out of order.

 

[ROG27]

aeges just talked more about sauce

http://www.neogaf.com/forum/showpost.php?p=46709207&postcount=1150

True. I'm just looking at the block diagram and how much silicon Microsoft is dedicating to that and the memory movers (and the audio DSP), all of which is custom to the system and expensive, and wondering why that, instead of more space dedicated to more GPU resources. They're telling developers it brings a number of advantages and frees them up to do "things," which I don't understand in an appreciable way. They spend a lot of time in their documentation talking about them.

Perhaps the "Data Move Engines" are some sort of advanced (semi?)programmable schedulers that ensure the GPU is running at or near 100% efficiency at all times.

If MS could get near 100% efficiency out of the GPU rather than say a more typical 60-70% efficiency at any given moment, they could save even more money (and heat) on the silicon in the future after subsequent die shrinks while still getting similar real world performance that rivals a GPU with greater brute strength but a less efficient architecture.

 

[Xenio]

Looks like my fuse box. Who's the leak?

it' a patent from microsoft from 3 years ago, when they start to develop the durango project