Xbox One (Durango) Technical hardware investigation

[Cyan]

http://www.anandtech.com/show/6972/xbox-one-hardware-compared-to-playstation-4/2

anand has a look at xbone hardware analysis, and it's written by anand who i typically find insightful. still looking, but already found this











Although necessarily superficial, looks like another excellently well done article by Anand.

That's a very interesting read indeed. The best technical article I've read on the hardware of the console. Thanks for sharing.

 

[szymku]

I'm thinking that the case is too big for 100W - its like mini ATX form factor, which I personally use for i5/HD7970/630W power supply. Do you think it is possible that they showed the final design yesterday, that will be downsized for the launch? They will integrate motherboard some more, which should be possible, having in mind power consumption.

 

[Scott_Arm]

I'm thinking that the case is too big for 100W - its like mini ATX form factor, which I personally use for i5/HD7970/630W power supply. Do you think it is possible that they showed the final design yesterday, that will be downsized for the launch? They will integrate motherboard some more, which should be possible, having in mind power consumption.

The SoC is 100W, which is going to be the biggest draw, but it will be more than that for the whole system. I don't know. 150W? I have no idea how much power the memory, fan, Bluray drive would use.

The picture is deceiving. I've been reading it's just slightly wider than Xbox 360S, maybe around the width of PS3 (non-slim). I think it's shorter than PS3 because of PS3's hump. Depth is probably around the same in proportion. The brick is external though. They seem to have left a lot of breathing room inside, looking at the teardowns. Seems like an incredibly simple design for a manufacture and repair point of view. Tons of venting and that big fan too. Interesting to see how cooling works when you don't have a controlled airflow. Hopefully they deliver on the cool and quiet factor.

 

[betan]

Hmmm...this quote from Anandtech seems interesting:


"Although 32MB doesn't sound like much, if it is indeed used as a cache (with the frame buffer kept in main memory) it’s actually enough to have a substantial hit rate in current workloads (although there’s not much room for growth)........If it’s used as a cache, the embedded SRAM should significantly cut down on GPU memory bandwidth requests which will give the GPU much more bandwidth than the 256-bit DDR3-2133 memory interface would otherwise imply. Depending on how the eSRAM is managed, it’s very possible that the Xbox One could have comparable effective memory bandwidth to the PlayStation 4. If the eSRAM isn't managed as a cache however, this all gets much more complicated."

From the way I understand it, the eSRAM in the XO isn't software managed but a true cache. That is, it is transparent to the developer unlike the eDRAM in the 360. Anand seems to be hitting the point ERP and Gubbi were making about the flexibility of the eSRAM and its potential benefits. My question is; how does this cut down on the GPU memory bandwidth request?

I think you read it wrong, it's not a developer transparent hardware cache (nor should it be), but most likely software cache (maybe hardware with high degree control like tagging etc.)

Still, calling it a software cache is a little misleading though I wouldn't necessarily call it inaccurate.

As for the GPU memory bw question, it's the same with any local storage/cache, as in you don't have to read/write to/from main mem when you have the data locally.

 

[sebbbi]

Anand seems to be hitting the point ERP and Gubbi were making about the flexibility of the eSRAM and its potential benefits. My question is; how does this cut down on the GPU memory bandwidth request?

On Xbox 360, the EDRAM helps a lot with backbuffer bandwidth. For example in our last Xbox 360 game we had a 2 MRT g-buffer (deferred rendering, depth + 2x8888 buffers, same bit depth as in CryEngine 3). The g-buffer writes require 12 bytes of bandwidth per pixel, and all that bandwidth is fully provided by EDRAM. For each rendered pixel we sample three textures. Textures are block compressed (2xDXT5+1xDXN), so they take a total 3 bytes per sampled texel. Assuming a coherent access pattern and trilinear filtering, we multiply that cost by 1.25 (25% extra memory touched by trilinear), and we get a texture bandwidth requirement of 3.75 bytes per rendered pixel. Without EDRAM the external memory bandwidth requirement is 12+3.75 bytes = 15.75 bytes per pixel. With EDRAM it is only 3.75 bytes. That is a 76% saving (over 4x external memory bandwidth cost without EDRAM). Deferred rendering is a widely used technique in high end AAA games. It is often criticized to be bandwidth inefficient, but developers still love to use it because it has lots of benefits. On Xbox 360, the EDRAM enables efficient usage of deferred rendering.

Also a fast read/write on chip memory scratchpad (or a big cache) would help a lot with image post processing. Most of the image post process algorithms need no (or just a little) extra memory in addition to the processed backbuffer. With large enough on chip memory (or cache), most post processing algorithms become completely free of external memory bandwidth. Examples: HDR bloom, lens flares/streaks, bokeh/DOF, motion blur (per pixel motion vectors), SSAO/SSDO, post AA filters, color correction, etc, etc. The screen space local reflection (SSLR) algorithm (in Killzone Shadow Fall) would benefit the most from fast on chip local memory, since tracing those secondary rays from the min/max quadtree acceleration structure has quite an incoherent memory access pattern. Incoherent accesses are latency sensitive (lots of cache misses) and the on chip memories tend to have smaller latencies (of course it's implementation specific, but that is usually true, since the memory is closer to the execution units, for example Haswell's 128 MB L4 should be lower latency than the external memory). I would expect to see a lot more post process effects in the future as developers are targeting cinematic rendering with their new engines. Fast on chip memory scratchpad (or a big cache) would reduce bandwidth requirement a lot.

 

[DuckThor Evil]

The SoC is 100W, which is going to be the biggest draw, but it will be more than that for the whole system. I don't know. 150W? I have no idea how much power the memory, fan, Bluray drive would use.

The picture is deceiving. I've been reading it's just slightly wider than Xbox 360S, maybe around the width of PS3 (non-slim). I think it's shorter than PS3 because of PS3's hump. Depth is probably around the same in proportion. The brick is external though. They seem to have left a lot of breathing room inside, looking at the teardowns. Seems like an incredibly simple design for a manufacture and repair point of view. Tons of venting and that big fan too. Interesting to see how cooling works when you don't have a controlled airflow. Hopefully they deliver on the cool and quiet factor.

Maybe 120W for the whole system max.

What do you mean by "don't have a controlled airflow"? The case sucks air from the vents, through the heatsink and the fan exhaust it through the top. There is no doubt in my mind that it will be quiet.

 

[Scott_Arm]

Maybe 120W for the whole system max.

What do you mean by "don't have a controlled airflow"? The case sucks air from the vents, through the heatsink and the fan exhaust it through the top. There is no doubt in my mind that it will be quiet.

Just doesn't look like there's a clear ducted path. There seem to be vents just about everywhere. That's all.

 

[scently]

I think you read it wrong, it's not a developer transparent hardware cache (nor should it be), but most likely software cache (maybe hardware with high degree control like tagging etc.)

Still, calling it a software cache is a little misleading though I wouldn't necessarily call it inaccurate.

As for the GPU memory bw question, it's the same with any local storage/cache, as in you don't have to read/write to/from main mem when you have the data locally.

No I am sure I am not reading it wrong. By software managed, he is describing an approach more like the 360 where the framebuffer must go through the 10mb eDRAM and also they cannot manipulate the data in there. Essentially, it cannot be used as a scratch pad. The implementation on the ONE is different in that it can be used as a framebuffer but it can also be used as a cache/scratch pad.

Thanks sebbbi for your reply, much appreciated.

 

[french toast]

Maybe 120W for the whole system max.

What do you mean by "don't have a controlled airflow"? The case sucks air from the vents, through the heatsink and the fan exhaust it through the top. There is no doubt in my mind that it will be quiet.

I don't get it, this thing has all the right ingredients for a power house, space, cooling, process node, tdp headroom, technology just waiting on the table.. Yet they have been very very conservative with the raw performance and the over all design of the thing.

It can't even be stood up on its side like 360, the controller is bound to be the best out there, but would have been nice if we had a bit of lighting on there or even a small lcd display, hell even a small one on the console it's self, blue lcd or something matching the controller, displaying some basic information, it would look like an expensive pioneer amplifier.

I can only hope they have found the yields to up clocks and include psu before release.

I want to like xbox one I really do, so far it just seems a bit underwhelming too be honest.

 

[DJ12]

You need to look at this gen, and in particular at the Wii, won the generation hands down, made a profit from day one and was a generation behind PS3 and 360 in terms of graphics. At least MS haven't fully adapted that model and xbox one can still function as a "next gen" console

 

[MfA]

Just doesn't look like there's a clear ducted path. There seem to be vents just about everywhere. That's all.

CPU and main memory are not spread out and there is no chance of internal fans recycling stale air ... with a single fan and single significant heatsink there is no point in any ducting (other than from the fan to the fan grill).

 

[DuckThor Evil]

I don't get it, this thing has all the right ingredients for a power house, space, cooling, process node, tdp headroom, technology just waiting on the table.. Yet they have been very very conservative with the raw performance and the over all design of the thing.

They went for the quiet route instead. As a multipurpose living room device, a little added sophistication will help.

 

[anexanhume]

I know this is the strong consensus, but another voice to back it up:

https://twitter.com/anandshimpi/status/337296343208972288

(on the subject of 1T versus 6T eSRAM)

1T I believe must be licensed through MoSys, manufacturing (and potential portability) are both concerns for the SoC

 

[Cyan]

On Xbox 360, the EDRAM helps a lot with backbuffer bandwidth. For example in our last Xbox 360 game we had a 2 MRT g-buffer (deferred rendering, depth + 2x8888 buffers, same bit depth as in CryEngine 3). The g-buffer writes require 12 bytes of bandwidth per pixel, and all that bandwidth is fully provided by EDRAM. For each rendered pixel we sample three textures. Textures are block compressed (2xDXT5+1xDXN), so they take a total 3 bytes per sampled texel. Assuming a coherent access pattern and trilinear filtering, we multiply that cost by 1.25 (25% extra memory touched by trilinear), and we get a texture bandwidth requirement of 3.75 bytes per rendered pixel. Without EDRAM the external memory bandwidth requirement is 12+3.75 bytes = 15.75 bytes per pixel. With EDRAM it is only 3.75 bytes. That is a 76% saving (over 4x external memory bandwidth cost without EDRAM). Deferred rendering is a widely used technique in high end AAA games. It is often criticized to be bandwidth inefficient, but developers still love to use it because it has lots of benefits. On Xbox 360, the EDRAM enables efficient usage of deferred rendering.

Also a fast read/write on chip memory scratchpad (or a big cache) would help a lot with image post processing. Most of the image post process algorithms need no (or just a little) extra memory in addition to the processed backbuffer. With large enough on chip memory (or cache), most post processing algorithms become completely free of external memory bandwidth. Examples: HDR bloom, lens flares/streaks, bokeh/DOF, motion blur (per pixel motion vectors), SSAO/SSDO, post AA filters, color correction, etc, etc. The screen space local reflection (SSLR) algorithm (in Killzone Shadow Fall) would benefit the most from fast on chip local memory, since tracing those secondary rays from the min/max quadtree acceleration structure has quite an incoherent memory access pattern. Incoherent accesses are latency sensitive (lots of cache misses) and the on chip memories tend to have smaller latencies (of course it's implementation specific, but that is usually true, since the memory is closer to the execution units, for example Haswell's 128 MB L4 should be lower latency than the external memory). I would expect to see a lot more post process effects in the future as developers are targeting cinematic rendering with their new engines. Fast on chip memory scratchpad (or a big cache) would reduce bandwidth requirement a lot.

Thanks for the input -as usual- and for sharing your knowledge sebbbi!! Excuse me if you can't reply to this for whatever reason, but I am just curious.

Does the eSRAM have similar or additional benefits compared to the original X360 EDRAM?

This is not a direct question for you, but another curiosity.

I heard time ago -here on B3D iirc- that there is an "special mode" where developers can use more than 5GB of RAM on the Xbox One. Is that true? :smile:

 

[scently]

Thanks for the input -as usual- and for sharing your knowledge sebbbi!! Excuse me if you can't reply to this for whatever reason, but I am just curious.

Does the eSRAM have similar or additional benefits compared to the original X360 EDRAM?

This is not a direct question for you, but another curiosity.

I heard time ago -here on B3D iirc- that there is an "special mode" where developers can use more than 5GB of RAM on the Xbox One. Is that true? :smile:

The first part of his reply deals with the first part of your question while the second part deals with the additional benefits.

 

[Cyan]

The first part of his reply deals with the first part of your question while the second part deals with the additional benefits.

I re-read it like three times and ultimately I can't get around the fact that sebbbi never mentions the ESRAM nor the Xbox One but the potential benefits of the EDRAM on the X360.

Since the bandwidth on the Xbox 360's EDRAM was much, much higher, I wonder if those benefits apply to the eSRAM too.

 

[Jwm]

Maybe 120W for the whole system max.

What do you mean by "don't have a controlled airflow"? The case sucks air from the vents, through the heatsink and the fan exhaust it through the top. There is no doubt in my mind that it will be quiet.

It is three times more quiet than the slim, and it has custom designed internal components like the Blu-ray drive that all aide airflow allowing them to run the cooling solution. At least that is what I caught from the tech panel on twitch.

I don't get it, this thing has all the right ingredients for a power house, space, cooling, process node, tdp headroom, technology just waiting on the table.. Yet they have been very very conservative with the raw performance and the over all design of the thing.

It can't even be stood up on its side like 360, the controller is bound to be the best out there, but would have been nice if we had a bit of lighting on there or even a small lcd display, hell even a small one on the console it's self, blue lcd or something matching the controller, displaying some basic information, it would look like an expensive pioneer amplifier.

I can only hope they have found the yields to up clocks and include psu before release.

I want to like xbox one I really do, so far it just seems a bit underwhelming too be honest.

They did have controllers with screens, but after testing they discovered that the gamers did not like having to look down and back up at the screen. It now seems the controllers have IR emitters to let Kinect know who is holding what controller. If you swap controllers it will recognize this and switch you, which could be a negative - my son likes to give me the controller to help him get around things. If we are playing co-op I can see this becoming frustrating, hopefully there is an option to disable this.

The screen thing makes sense, using smartglass for maps I have not really been fond of looking down and having to focus on where my dot is, looking back up to the game and refocusing. But I am old!

 

[bkilian]

bkilian,

Any idea if the video output will follow video standards or are we looking at funky gamma curves and other such oddities agian? If I use it as my BluRay player, I don't want to the signal altered.

No, there is no piecewise linear approximation for the gamma curve this time.

Sound is not a system seller and both consoles will sound great.

Maybe bkilian can clarify (maybe not lol) but shape is more about kinect than actual sound we hear. Shape includes all the kinect stuff first leaked by vgleaks as seperate items I think

Probably can't say too much. As far as I can tell, MS has not announced anything about any audio hardware.

Any tip about what do MS refers to when talking about Kinect natural language processing?.Seing the conference it still seems similar to spoken commands ala old version of google search and not Siri alike in which you can say "i would like to see a film" instead of explicit command "Xbox watch film".I supposee Bkillian could know about whether there is truly natural language processing as well as speech recognition processing.
By the way the speech recognition of the conference was astounding, the improvements MS has done in this with deep learning make it similar to google's now assistant and way better than Siri's nuance.

I imagine it will have full natural language support if you're connected to the internet, and possibly a still good, but slightly less dynamic support if you are not on the internet.

 

[3dilettante]

I re-read it like three times and ultimately I can't get around the fact that sebbbi never mentions the ESRAM nor the Xbox One but the potential benefits of the EDRAM on the X360.

The second half deals with a read-write scratchpad memory, which the 360's eDRAM is not.

Since the bandwidth on the Xbox 360's EDRAM was much, much higher, I wonder if those benefits apply to the eSRAM too.

Bandwidth to the eDRAM daughter die is significantly lower. It's ROP operations on the eDRAM that can take full advantage of the internal bandwidth, but the analysis here is dependent on the evolution of the memory hierarchy and functionality of ROP hardware since Xenos.

 

[Cyan]

The second half deals with a read-write scratchpad memory, which the 360's eDRAM is not.


Bandwidth to the eDRAM daughter die is significantly lower. It's ROP operations on the eDRAM that can take full advantage of the internal bandwidth, but the analysis here is dependent on the evolution of the memory hierarchy and functionality of ROP hardware since Xenos.

My brain and my face have lit up with the comment.

Now I understand everything, that's a key difference for me that emerged from your words. My confusion came from the fact that I thought the EDRAM is also a scratchpad memory -why is the eSRAM a scratchpad memory and the EDRAM isn't, by the way? (a scratchpad is a notebook to me)-.

And more than anything else, it makes sebbbi's words the more interesting. That's a fabulous post 'cos it explains key differences in the architecture and that's going to make for a very interesting generation once again.

Alternatively, I thought that the Xbox One was basically a less powerful version of the PlayStation 4 hardware, but there is more to it.