• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
  • The Politics forum has been nuked. Please do not bring political discussion to the rest of the site, or you will be removed. Thanks.

Digital Foundry: the complete Xbox One architects interview

Chobel

Member
Mar 26, 2013
15,673
3
515
Balance it if old

www.eurogamer.net/articles/digitalfoundry-the-complete-xbox-one-interview

So this is the complete article that discuss Xbox One architecture, it contains all the old stuff that we saw before like balance, 10% GPU reserve...etc. But also includes some new information like MS GPGPU approach, NAND memory and some other stuff.

CPU have access to eSRAM
Digital Foundry: And you have CPU read access to the ESRAM, right? This wasn't available on Xbox 360 eDRAM.

Nick Baker: We do but it's very slow.

GPGPU approach
Digital Foundry: So what is your general approach to GPGPU? Sony has made a big deal about its wider compute pipelines in order to get more utilisation of the ALU. What is your philosophy for GPGPU on Xbox One?

Andrew Goossen: Our philosophy is that ALU is really, really important going forward but like I said we did take a different tack on things. Again, on Xbox One our Kinect workloads are running on the GPU with asynchronous compute for all of our GPGPU workloads and we have all the requirements for efficient GPGPU in terms of fast coherent memory, we have our operating system - that takes us back to our system design. Our memory manager on game title side is completely rewritten. We did that to ensure that our virtual addressing for the CPU and GPU are actually the same when you're on that side. Keeping the virtual addresses the same for both CPU and GPU allows the GPU and CPU to share pointers. For example, a shared virtual address space along with coherent memory along with eliminating demand paging means the GPU can directly traverse CPU data structures such as linked lists.

On the system side we're running in a complete generic Windows memory manager but on the game side we don't have to worry about back-compat or any of these nasty issues. It's very easy for us to rewrite the memory manager and so we've got coherent memory, the same virtual addressing between the two, we have synchronisation mechanisms to coordinate between the CPU and GPU that we can run on there. I mean, we invented DirectCompute - and then we've also got things like AMP that we're making big investments on for Xbox One to actually make use of the GPU hardware and the GPGPU workloads.

The other thing I will point out is that also on the internet I see people adding up the number of ALUs and the CPU and adding that onto the GPU and saying, "Ah, you know, Microsoft's CPU boost doesn't make much of a difference." But there still are quite a number of workloads that do not run efficiently on GPGPU. You need to have data parallel workloads to run efficiently on the GPU. The CPU nowadays can run non-data parallel workloads but you're throwing away massive amounts of performance. And for us, getting back to the balance and being able to go back and tweak our performance with the overhead in the margin that we had in the thermals and the silicon design, it kind of enabled us to go back and look at things. We looked at our launch titles and saw that - hey we didn't make the balance between CPU and GPU in terms of our launch titles - we probably under-tweaked it when we designed it two or three years ago. And so it was very beneficial to go back and do that clock raise on the CPU because that's a big benefit to your workloads that can't be running data parallel.

Digital Foundry: The GPU compute comparison seems to be about Xbox One's high coherent read bandwidth vs. raw ALU on PS4. But don't the additional ACEs added to PS4 aim to address that issue?

Andrew Goossen: The number of asynchronous compute queues provided by the ACEs doesn't affect the amount of bandwidth or number of effective FLOPs or any other performance metrics of the GPU. Rather, it dictates the number of simultaneous hardware "contexts" that the GPU's hardware scheduler can operate on any one time. You can think of these as analogous to CPU software threads - they are logical threads of execution that share the GPU hardware. Having more of them doesn't necessarily improve the actual throughput of the system - indeed, just like a program running on the CPU, too many concurrent threads can make aggregate effective performance worse due to thrashing. We believe that the 16 queues afforded by our two ACEs are quite sufficient.

eSRAM Latency doesn't really matter?
Digital Foundry: [...] Does low latency here materially affect GPU performance?

Nick Baker: You're right. GPUs are less latency sensitive. We've not really made any statements about latency.
 

PdotMichael

Banned
Feb 9, 2011
20,731
0
540
I'm impresses how they are trying to sell the CPU upclock as big game changer. It's like I'm back in the 90s.
 

Sbraik79

Banned
May 12, 2013
42
0
0
I wish them luck balancing with a roller blade on one foot, and an ice skate on the other.
 

mrklaw

MrArseFace
Jun 10, 2004
59,891
2
0
Windsor, UK
The ESRAM stuff about addressing is interesting. Seems to suggest it is seamlessly integrated into the memory map - they talk about having some of your render targets in ESRAM and others in DDR3 for instance.

As I said in the other thread. Balance appears 25 times in this article.

I'm about halfway through reading at the moment.

To be fair, it does sound like a nicely balanced architecture, just with lower ultimate power than Ps4
 
Jun 11, 2011
3,851
0
0
England
Just like our friends we're based on the Sea Islands family. We've made quite a number of changes in different parts of the areas. The biggest thing in terms of the number of compute units, that's been something that's been very easy to focus on. It's like, hey, let's count up the number of CUs, count up the gigaflops and declare the winner based on that. My take on it is that when you buy a graphics card, do you go by the specs or do you actually run some benchmarks? Firstly though, we don't have any games out. You can't see the games. When you see the games you'll be saying, "What is the performance difference between them?" The games are the benchmarks
Of course I buy them based on specs, they are both the same family and that one has x number of cores more, more rops, more tmus, etc.. etc.. why would I buy the slower one?
 

Sword Of Doom

Member
Aug 18, 2013
11,594
3
0
The ESRAM stuff about addressing is interesting. Seems to suggest it is seamlessly integrated into the memory map - they talk about having some of your render targets in ESRAM and others in DDR3 for instance.



To be fair, it does sound like a nicely balanced architecture, just with lower ultimate power than Ps4


So with the components they have, they've made good design choices
 

inherendo

Member
Jun 20, 2013
336
0
0
I thought that since tasks given to the gpu for compute are the type that benefit from the ability of it to do parallel workloads, so additional threads would always be good;whereas, not all work done on a cpu can be done in parallel efficiently.
 

PdotMichael

Banned
Feb 9, 2011
20,731
0
540
To be fair, it does sound like a nicely balanced architecture, just with lower ultimate power than Ps4

Not sure how you can call it a "balanced" system if basicaly everything is about to deal with the slow DDR3 ram.
And the ESRAM is clearly not big enough for 1080p output in mind.
 
May 8, 2006
5,620
245
1,485
Spain
steamcommunity.com
Of course I buy them based on specs, they are both the same family and that one has x number of cores more, more rops, more tmus, etc.. etc.. why would I buy the slower one?

well I gotta say I bought my 7950 because of benchmarks and reviews. I could've bought a GTX660Ti but after reading some analisys, reviews and bechmarks I ended up going for the 7950
 

Sword Of Doom

Member
Aug 18, 2013
11,594
3
0
When you see the games you'll be saying, "What is the performance difference between them?" The games are the benchmarks


When I read this, I can't help but think multiplatform games will suffer from "console parity"
 

benny_a

extra source of jiggaflops
Apr 25, 2009
17,350
1
0
So with the components they have, they've made good design choices
The majority of posts I read on this topic about people that are interested in the tech never doubted the technical abilities of the Microsoft engineers. It's just that the person calling the shots in the Xbox division didn't ask his engineers to go with the most powerful games console at X price point, but rather try to have all these auxiliary features be present.
 
Feb 17, 2010
6,723
0
710
CT
www.doddscientifics.com
To be fair, it does sound like a nicely balanced architecture, just with lower ultimate power than Ps4



What am I missing about the "balance" argument? I admit that I'm not the most tech savvy person in the world but basically by balance they mean that there's no real bottlenecks but also nothing particularly fast... like it's all stuff in the same class. Which sounds fine. But why would that be better than something that has all that same shit except for one piece of the puzzle that is much faster. Would throwing the "balance" off make it worse somehow?


Basically when i hear balance what I hear is "everything is mediocre so there's no bottlenecks... but there's also nothing exceptional."


Is that wrong? Serious question.... i'm not trolling.
 

goomba

Banned
Jul 16, 2004
7,865
0
0
New Zealand
balance is not overrated in tech design, the gamecube was considerbly behind the xbox on paper but kept up and arguably even outclassed it.
 

shinra-bansho

Member
Nov 13, 2011
16,594
0
0
To be fair, it does sound like a nicely balanced architecture, just with lower ultimate power than Ps4
It was pointed out by phosphor in the other thread I think, but doesn't some of this suggest that they're seeing "bottlenecks" in other parts of the rendering pipeline and that's why they don't see any benefit of unlocking the additional compute units.

I.e. they may not be able to fully utilize the 12 CUs they have given the main RAM bandwidth and 16 ROPs? An extra two functional CUs then obviously wouldn't provide linear scaling in performance?

I may be completely wrong and defer to those more technical to either affirm or correct?

EDIT: also doesn't the PS4 also have coherent memory bandwidth?
 

Sword Of Doom

Member
Aug 18, 2013
11,594
3
0
The majority of posts I read on this topic about people that are interested in the tech never doubted the technical abilities of the Microsoft engineers. It's just that the person calling the shots in the Xbox division didn't ask his engineers to go with the most powerful games console at X price point, but rather try to have all these auxiliary features be present.

Yes definitely
 

Sentenza

Member
Dec 3, 2011
15,124
2,845
970
Leadbetter again?
Let me guess: the final conclusion of this interview is that XB1 is a bold, remarkable, marvelous piece of groundbreaking tech. Or something.
 

StuBurns

Banned
Jan 9, 2008
69,534
0
0
When I read this, I can't help but think multiplatform games will suffer from "console parity"
That's not going to happen. I don't know why this constantly comes up, but never has this been a factor before. Because Skyrim is broke on PS3, they didn't break it on 360 to make it fair. Bayonetta isn't at 30fps on 360 because it is on PS3. They didn't lower the resolution of FFXIII and muddy the cutscenes to make it 'parity'.

It's a competitive industry, and no one is going to leave huge performance on the table when the first parties won't, just to appease MS. DICE are coming out against KZ and CoD, they will put their best foot forward.
 
Jun 11, 2011
3,851
0
0
England
well I gotta say I bought my 7950 because of benchmarks and reviews. I could've bought a GTX660Ti but after reading some analisys, reviews and bechmarks I ended up going for the 7950

You are comparing different architectures though, XB1 and PS4 are from the same family so its easy to do a comparison based on specs no matter how much Microsoft protest otherwise.
 

mrklaw

MrArseFace
Jun 10, 2004
59,891
2
0
Windsor, UK
Not sure how you can call it a "balanced" system if basicaly everything is about to deal with the slow DDR3 ram.
And the ESRAM is clearly not big enough for 1080p output in mind.

The 8Gb was clearly central to their plans for Xbox. Those plans (media stuffs) might not be what you want, but it doesn't stop it being balanced

And the ESRAM being able to hold bandwidth intensive render targets and the DDR holding simpler ones should be good enough
 

PetriP-TNT

Member
Jul 13, 2007
21,465
0
0
Finland
What am I missing about the "balance" argument? I admit that I'm not the most tech savvy person in the world but basically by balance they mean that there's no real bottlenecks but also nothing particularly fast... like it's all stuff in the same class. Which sounds fine. But why would that be better than something that has all that same shit except for one piece of the puzzle that is much faster. Would throwing the "balance" off make it worse somehow?


Basically when i hear balance what I hear is "everything is mediocre so there's no bottlenecks... but there's also nothing exceptional."


Is that wrong? Serious question.... i'm not trolling.

Well, people are reading into it as "yeah we have inferior hardware to PS4 but..."
 

TronLight

Everybody is Mikkelsexual
Oct 2, 2011
3,317
0
775
Is this the first time they explain what these 15 co-processors can do?

Beside SHAPE (that looks awesome) the other are pretty usual I think (audio de-enconder and the upscaler).
 

PdotMichael

Banned
Feb 9, 2011
20,731
0
540
Leadbetter again?
Let me guess: the final conclusion of this interview is that XB1 is a bold, remarkable, marvelous piece of groundbreaking tech. Or something.

Just read the introduction of the interview.

He is implying that the PS4 power adventage is more effective PR than anything else.

The 8Gb was clearly central to their plans for Xbox. Those plans (media stuffs) might not be what you want, but it doesn't stop it being balanced

And the ESRAM being able to hold bandwidth intensive render targets and the DDR holding simpler ones should be good enough

Well, it does if the system is completly way less effective than the PS4 solution - while being the more expensive system.

There is not a single thing in the real world, which can be done better on the Xbox One than on the PS4.
 

mrklaw

MrArseFace
Jun 10, 2004
59,891
2
0
Windsor, UK
What am I missing about the "balance" argument? I admit that I'm not the most tech savvy person in the world but basically by balance they mean that there's no real bottlenecks but also nothing particularly fast... like it's all stuff in the same class. Which sounds fine. But why would that be better than something that has all that same shit except for one piece of the puzzle that is much faster. Would throwing the "balance" off make it worse somehow?


Basically when i hear balance what I hear is "everything is mediocre so there's no bottlenecks... but there's also nothing exceptional."


Is that wrong? Serious question.... i'm not trolling.


Balance to me would suggest that there aren't any major bottlenecks. Eg they talk about why they don't have 32 ROPs - they do some obfuscation but then basically say there is no point having 32 because the system doesn't have the bandwidth to feed them.

No point having exceptional areas if they are going to be held back by the rest of the system. You are better off spending that silicon elsewhere
 

Guymelef

Member
Sep 19, 2007
12,900
3
1,125
Spain
What am I missing about the "balance" argument? I admit that I'm not the most tech savvy person in the world but basically by balance they mean that there's no real bottlenecks but also nothing particularly fast... like it's all stuff in the same class. Which sounds fine. But why would that be better than something that has all that same shit except for one piece of the puzzle that is much faster. Would throwing the "balance" off make it worse somehow?


Basically when i hear balance what I hear is "everything is mediocre so there's no bottlenecks... but there's also nothing exceptional."


Is that wrong? Serious question.... i'm not trolling.

It's my impression, people using the balance argument as if PS4 Sony choose components on the bazaar and simply put them together.
 

fatgamecat

Member
Oct 24, 2012
1,167
0
670
This is from the last big thread.

X1 GPU:
1.18 TF GPU (12 CUs) for games
768 Shaders
48 Texture units
16 ROPS
2 ACE/ 16 queues

PS4 GPU:
1.84TF GPU ( 18 CUs) for games + 56%
1152 Shaders +50%
72 Texture units +50%
32 ROPS + 100%
8 ACE/64 queues +300%
 

shinra-bansho

Member
Nov 13, 2011
16,594
0
0
Is this the first time they explain what these 15 co-processors can do?

Beside SHAPE (that looks awesome) the other are pretty usual I think (audio de-enconder and the upscaler).
Besides SHAPE and the additional move engines, I think they're essentially standard AMD GPU features and/or things that the PS4 has as well I believe, although I think the PS4's video compositor only has two planes.
 

DeFiBkIlLeR

Banned
Dec 27, 2010
1,435
0
0
Just read the introduction of the interview.

He is implying that the PS4 power adventage is more effective PR than anything else.

So...do you think when he comes to doing his face off's in the coming years, he's going to fake videos and screen shots and basically lie through his teeth in order to keep MS happy?

..because that's what all haters of RL are implying with their pathetic character assassination, who's been in the industry for nearly 30yrs and knows more about this stuff than 99% of the inane fan boys that come out with this shit.
 

stryke

Member
Oct 1, 2011
10,922
1
760
It's my impression, people using the balance argument as if PS4 Sony choose components on the bazaar and simply put them together.

Luckily we have a handy reply ready for that

 

Chobel

Member
Mar 26, 2013
15,673
3
515
In this article DF is implying that Xbox One have higher bandwidth for GPGPU than PS4, but we know that Both have 30 GB/s for coherent memory access.
 
Feb 17, 2010
6,723
0
710
CT
www.doddscientifics.com
Balance to me would suggest that there aren't any major bottlenecks. Eg they talk about why they don't have 32 ROPs - they do some obfuscation but then basically say there is no point having 32 because the system doesn't have the bandwidth to feed them.

No point having exceptional areas if they are going to be held back by the rest of the system. You are better off spending that silicon elsewhere



That makes sense. I guess the next logical question is... are there any known bottlenecks on the PS4 that make it "unbalanced?"
 
May 8, 2006
5,620
245
1,485
Spain
steamcommunity.com
You are comparing different architectures though, XB1 and PS4 are from the same family so its easy to do a comparison based on specs no matter how much Microsoft protest otherwise.

well I like to read about every gadget I want to buy, I don't get carried on because of specs mainly, but the conclusion of some of the articles I read (depending on the source, of course).

I think PS4 will look better, but I don't think they will look that MUCH as many expect, or give for granted, and no, I don't think about the "console parity" many start to complain by now. I'm sure XB1 can do 1080p as PS4 will do, and I can help but feel we'll see 720p titles as well in both platforms, though I'd love to be wrong because I'm used to play almost every PC game on my rig downsampled from 2560x1440 on my 55" TV, so for me both platforms won't be as visually impressive as for other fellows
 

PdotMichael

Banned
Feb 9, 2011
20,731
0
540
So...do you think when he comes to doing his face off's in the coming years, he's going to fake videos and screen shots and basically lie through his teeth in order to keep MS happy?

..because that's what all haters of RL are implying with their pathetic character assassination, who's been in the industry for nearly 30yrs and knows more about this stuff than 99% of the inane fan boys that come out with this shit.

Pretty sure, if he stops writing shitty PR articles for Microsoft. He will get way less "hate" in the future.
 

mrklaw

MrArseFace
Jun 10, 2004
59,891
2
0
Windsor, UK
It was pointed out by phosphor in the other thread I think, but doesn't some of this suggest that they're seeing "bottlenecks" in other parts of the rendering pipeline and that's why they don't see any benefit of unlocking the additional compute units.

I.e. they may not be able to fully utilize the 12 CUs they have given the main RAM bandwidth and 16 ROPs? An extra two functional CUs then obviously wouldn't provide linear scaling in performance?

I may be completely wrong and defer to those more technical to either affirm or correct?

EDIT: also doesn't the PS4 also have coherent memory bandwidth?

I don't think so. I don't think they were ever going to unlock those extra compute units - they went with 14 on the silicon to allow for failures so they have 12 good ones to use.
 

Zen

Banned
Jun 24, 2005
13,050
0
0
Explain away the GPUs he uses for his XB1 and PS4 comparisons, go ahead and try.