• 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.

Digital Foundry: the complete Xbox One architects interview

angelic

Banned
ps4xbonebalancesrjm1.png

ps4xbonebalancesrjm11.png
 

Vizzeh

Banned
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.

This seems an attempt to downplay PS4's advantage in compute/compute queing? - Can anyone apart from Cerny, Comment on this portion of the interview, if MS is accurate or not in its positivity towards their on setup and downplay of PS4s? (apart from common sense, I mean from a tech standpoint)
 
Did that idiot just compare ACE:s to soft threads and said too much is bad... All credibility lost. This is pr bs to the MAX.

And funnily Butter said its high coherent read bw vs alu:s in ps4. wtf. Butter has no idea what the shit he writes about.
 
prove what exactly? that 1.84 TFLOPS> 1.18 TFLOPS? have you have ever heard of "math" and how it works? the PS4 is objectively is more powerful, i don't need shills like RL to tell my otherwise.

This isn't about which console is more powerful, that's clearly the PS4, it's about the likes of you making baseless accusations that he's deliberately pushing an agenda of Microsoft's liking, whether formally or informally.

Honestly the way he presents Microsoft's strengths, and misrepresents the PS4's (to the point of implying they are in fact weaknesses) has been consistent since the console reveals some months ago.

Guess you missed the articles where he clearly states the PS4 is the more powerful machine when it comes to GPU power.

But the most obvious is when he tried to downplay the difference in "compute" performance with wildly irrespective PC GPUs in a tapestry of misrepresentation, goalpost dancing, and either utter incompetence or an unshakable agenda.

No, that's just him reporting what the Xbox engineers are telling him, he's made no such decisions as to whether it will accurately reflect what we see on our screens over the coming years.


People have said he is single-handedly dragging down Digital Foundry's reputation, and I have to agree.

He IS digital foundry, he's the one who founded it.
 

Vito

Banned
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.

I think the digital foundry bias toward Microsoft is evident by now, they've basically become a PR outlet for them, Microsoft pushes their spins, Eurogamer gets the clicks, both parties happy.

Meanwhile Sony keeps quiet because they have nothing to address. And I doubt that Sony engineers are completely clueless at to not balance their system as well.
 

EvB

Member
I think the digital foundry bias toward Microsoft is evident by now, they've basically become a PR outlet for them, Microsoft pushes their spins, Eurogamer gets the clicks, both parties happy


I've never got this, why are they bias? Because 360 games almost always came out top in the console face offs?

Salt?
 

gofreak

GAF's Bob Woodward
I rather wish they'd just posted the full interview from the beginning.

It's an interesting read, but not a huge amount of new interesting stuff vs the articles published in advance of this.

However, the big surprise? They were upfront and decided not to play the eSRAM latency card at all in either a rendering or GPGPU context. I'm surprised, I thought they would try to fudge around that by amplifying some corner case benefits. It also - again - clarifies the only real purpose for the eSRAM, bandwidth.

The clarification that the CPU only has 'very slow' access to eSRAM is also interesting given Leadbetter's earlier comments about Microsoft viewing 'low latency access to shared memory' as key to good GPGPU performance. I guess Richard misinterpreted something.

The GPGPU commentary was interesting because they seemed to acknowledge the importance of more ALU into the future, but the only answer they seemed to have that was different to tweaks Sony is also making is the amount of coherent bandwidth. I'm not sure how XB1 actually compares here, I think the numbers we've seen from Sony are split out a little differently, but my underlying thought while reading that is that while GPGPU might be able to saturate 30GB/s of DDR3 bandwidth in pure GPGPU benchmarks, I really wonder if there'd be that kind of bandwidth going to spare in a mixed compute/graphics scenario in a real game on XB1. It would be much easier for PS4 to fully utilise whatever coherent and incoherent bandwidth it has available during GPGPU tasks. Dithering around the ACE question was unconvincing also, the lack of performance of context switching is something AMD has flagged too as to be addressed in future architectures.
 
This isn't about which console is more powerful, that's clearly the PS4, it's about the likes of you making baseless accusations that he's deliberately pushing an agenda of Microsoft's liking, whether formally or informally.



Guess you missed the articles where he clearly states the PS4 is the more powerful machine when it comes to GPU power.



No, that's just him reporting what the Xbox engineers are telling him, he's made no such decisions as to whether it will accurately reflect what we see on our screens over the coming years.




He IS digital foundry, he's the one who founded it.


are you honestly telling me you haven't noticed how RL is downplaying the PS4 power advantages? here just read this:

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?

notice how MS downplayed the PS4 large number of ACEs & Queues( i am 100% sure that the MS engineer has zero experience with the PS4 hardware or design) and he's soon to agree with them.
 

Perkel

Banned
We can finally put rest to:

- ESRAM LATENCY ! - No, confirmed by tech devs of console. ESRAM was used because of its production values not performance. ESRAM is simply bandwidth stitch.
They reworked their memory so they can fit buffer which is more than 32MB but still it isn't best sollution since they need to flipflap between DDR3 and ESRAM
- 15 custom hardware chips ! - Yes but those don't have mostly anything to do with power. DMA engines SHApe which is 4 chips and few other like compression. In summary they offload lot of things. PS4 also have most of those
- They can add write/read in ram getting 150GB/s real peak data. Thing is as we well know before it is either 109GB/s read or 109GB/s write. No 150GB/s read or write. 209GB/s is theoretical max where you write and read and the same time.
- reconfirmed 10% from GPU take for OS. They also mentioned some hit on GPU but very small (not Flop)

When they talked about CUs and hardware difference between PS4 and Xbone they have gone full defensive mode ignoring additional texture units for example which suggest everything is like we said. Also they used old 14+4 bullshit so they could preach their "balance"


Very honest interview imo and what is more important good one. They did some PR instead of science when they started to compare hardware relative power but most of what they said was actually interesting.


TLDR: They confirmed almost everything we knew already.
 

StuBurns

Banned
I've never got this, why are they bias? Because 360 games almost always came out top in the console face offs?

Salt?
Because when the 360 wins, it's because it's better, when the PS3 wins, it's because the developers are incompetent.

Look up his FFXIII coverage. He played the demo on PS3, decided the 360 version would be better, then when they came out, said they were close, then after people ripped him apart, he followed it up with another article saying they're not close, but it's because Square are morons.
 

Vizzeh

Banned
I rather wish they'd just posted the full interview from the beginning.

It's an interesting read, but not a huge amount of new interesting stuff vs the articles published in advance of this.

However, the big surprise? They were upfront and decided not to play the eSRAM latency card at all in either a rendering or GPGPU context. I'm surprised, I thought they would try to fudge around that by amplifying some corner case benefits. It also - again - clarifies the only real purpose for the eSRAM, bandwidth.

The clarification that the CPU only has 'very slow' access to eSRAM is also interesting given Leadbetter's earlier comments about Microsoft viewing 'low latency access to shared memory' as key to good GPGPU performance. I guess Richard misinterpreted something.

The GPGPU commentary was interesting because they seemed to acknowledge the importance of more ALU into the future, but the only answer they seemed to have that was different to tweaks Sony is also making is the amount of coherent bandwidth. I'm not sure how XB1 actually compares here, I think the numbers we've seen from Sony are split out a little differently, but my underlying thought while reading that is that while GPGPU might be able to saturate 30GB/s of DDR3 bandwidth in pure GPGPU benchmarks, I really wonder if there'd be that kind of bandwidth going to spare in a mixed compute/graphics scenario in a real game on XB1. It would be much easier for PS4 to fully utilise whatever coherent and incoherent bandwidth it has available during GPGPU tasks. Dithering around the ACE question was unconvincing also, the lack of performance of context switching is something AMD has flagged too as to be addressed in future architectures.

I seem to remember the EsRAM (before the whole GDDR5 vs DDR3 Latency debate) was quite a benifit over PS4 because of the memory latency to the CPU, the GPU access benfits of esRam was always downplayed because the GPU is bandwidth dependent and not latency. So that kind of kills off the esram > Cpu advantage notions when games are apparently CPU bound.... (possibly source engine/titanfall?)
 

EvB

Member
are you honestly telling me you haven't noticed how RL is downplaying the PS4 power advantages? here just read this:

Isn't that just a case of how you build rapport in an interview?

If you want to see examples of bad rapport to look at Angry Joe interviewing major Nelson. As soon as AJ tried to start asking awkward questions and disagreeing with the rep of the company he was interviewing , MN just went into shut down.

If I'm trying to get something from someone I don't make them feel awkward.
 

mrklaw

MrArseFace
MS (well, these guys in the interview) are not saying Xbox is more powerful than PS4. They carefully avoid specific comparisons. They are purely talking about the Xbox in isolation.

I realise this always ends up in a 'but which is better' but this article doesn't really set out to answer that - it is there to help us better understand how Xbox is setup. For me, it does that nicely, and raises some questions I'd like to know more about - like the comments about having your render targets split between ESRAM and DDR3.

It is probably a more detailed look than we've had for PS4 so far (although that architecture does seem more straightforward so perhaps a deeper look isn't as necessary)
 
Isn't that just a case of how you build rapport in an interview?

If you want to see examples of bad rapport to look at Angry Joe interviewing major Nelson. As soon as AJ tried to start asking awkward questions and disagreeing with the rep of the company he was interviewing , MN just went into shut down.

If I'm trying to get something from someone I don't make them feel awkward.

yeah, but he would at least hint that it's not accurate.

edit: and it's just one example. RL has the downplayed the PS4 power advantages several times before. the most infamous one is when he compared an HD 7850 to HD 7870 XT to show the GPU power advantage of the PS4, ignoring the ROPs, memory architecture and bandwidth, TMU...etc.
 
MS (well, these guys in the interview) are not saying Xbox is more powerful than PS4. They carefully avoid specific comparisons. They are purely talking about the Xbox in isolation.

I realise this always ends up in a 'but which is better' but this article doesn't really set out to answer that - it is there to help us better understand how Xbox is setup. For me, it does that nicely, and raises some questions I'd like to know more about - like the comments about having your render targets split between ESRAM and DDR3.

It is probably a more detailed look than we've had for PS4 so far (although that architecture does seem more straightforward so perhaps a deeper look isn't as necessary)

Well, you can't blame MS engineers for being proud of their system.

But a good journalist would ask some serious questions about the efficency of the technical solutions.
 

EvB

Member
Because when the 360 wins, it's because it's better, when the PS3 wins, it's because the developers are incompetent.

Surely that is the case on many occasions? If 9/10 versions see the 360 version performing better, then 1/10 games sees the 360 version falling on it's face wouldn't you ask the question, why can't these developers do what the other 9 did?
 

Chobel

Member
MS (well, these guys in the interview) are not saying Xbox is more powerful than PS4. They carefully avoid specific comparisons. They are purely talking about the Xbox in isolation.

I realise this always ends up in a 'but which is better' but this article doesn't really set out to answer that - it is there to help us better understand how Xbox is setup. For me, it does that nicely, and raises some questions I'd like to know more about - like the comments about having your render targets split between ESRAM and DDR3.

It is probably a more detailed look than we've had for PS4 so far (although that architecture does seem more straightforward so perhaps a deeper look isn't as necessary)

How about this part? I call this downplaying the competition

Andrew Goossen said:
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.
 
MS (well, these guys in the interview) are not saying Xbox is more powerful than PS4. They carefully avoid specific comparisons. They are purely talking about the Xbox in isolation.


It is probably a more detailed look than we've had for PS4 so far (although that architecture does seem more straightforward so perhaps a deeper look isn't as necessary)

There is direct coparasions and full on BS...

I feel like LB did not do these questions himself as they reek ignorance. there is clear narrative.

And we know almost everything there is to know about ps4 (apart from backroud chip) we kniow the bw numbers. We know gpu numbers. We have overview of custominations. What more would you like to know?
 

Horp

Member
Not really. The dude should be in the left as that's where the majority of the weight is, unless the rocks are different densities and the xbox side is in actuality heavier than the ps side.

SHHHHHhhhhhhh nothing to see here, that never happened.
 
Brown nose them in the interview then slag them off in the article?

Maybe burning bridges isn't the best thing to do.

read my edit, as for your point above i can't disagree with you, but if i were RL and avoiding "burning bridges" means dragging my credibility to the ground i would do the former.
 

StuBurns

Banned
Surely that is the case on many occasions? If 9/10 versions see the 360 version performing better, then 1/10 games sees the 360 version falling on it's face wouldn't you ask the question, why can't these developers do what the other 9 did?
That's cyclical logic. Essentially, the best version is just the lead SKU almost every time. Why is the decision to have the PS3 be the lead SKU so rare is a valid question, but the answer is so clean cut there's not much discussion to be warranted from it.

If you look at the more recent, bigger releases, Tomb Raider, BioShock, GTA5, Metro LL, the PS3 has been coming out better. Is it because those developers have suddenly become worse? No, obviously not. They're just focusing much more on the PS3 versions, maybe because it's market share is now more worthy of their time, maybe because the CELL architecture requires more asynchronous coding (or is it multithreading? Someone in the thread will know), which all their next-gen engines will utilize, maybe something else entirely. But no, it's not because they're incompetent.
 

i-Lo

Member
Pertaining to PS4, while it will still remain ahead of Xbone in terms of raw power, how do we know that a certain % of GPU power (as it is with CPU) is not locked away for non gaming purposes?
 

B_Boss

Member
The thing that puzzles me and that actually makes me laugh a bit is that.....in all of these fascinating articles regarding the XB1 specs when discussing its technical specs lol..........there seems to be absolutely no negative aspect of the console when clearly the PS4 is significantly superior from a technical standpoint.

You would think that both consoles are factually on technical 'even ground' when reading this ginormous article and others like it....its just positive spin. Please stop. Baker & Goosen et al. did not do what Cerny & Co. did. Bad thing? No but you get my point....This post perfectly sums up my point:




nM20wqH.gif


??
 

Chobel

Member
I have hard time believing this
"The biggest source of your frame-rate drops actually comes from the CPU, not the GPU"

Frame rate drop usually happens in busy scenes, so that make the GPU the main reason for drops not the CPU.
 

TheOddOne

Member
How about this part? I call this downplaying the competition
Difficult to avoid the comparison when the question asked itself implies the comparison.
DF: 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?
If they said nothing, people would have noticed. They did say something and people noticed. It's a catch 22.
 

Chobel

Member
Pertaining to PS4, while it will still remain ahead of Xbone in terms of raw power, how do we know that a certain % of GPU power (as it is with CPU) is locked away for non gaming purposes?

For Xbox One that percentage is from Kinect and snap, both not in PS4. So GPU reservation in PS4 will be minimal, my guess less than 2%
 
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 +400%
I've seen you post this a couple times now. I'm genuinely curious what your point/goal is?
 
While some of the article is certainly about how it compares to PS4, mostly it's just about the design itself. You can admire the thought process and technology of a device even if there is a superior product offered by the competition.
Yes, that is super interesting. Even though, as a programmer, I'm sure I'd much prefer the more "elegant" and straightforward PS4 architecture, the XB1 team went through more technological choices which makes more interesting stories to read about (that's not the programmer in me speaking, that's the engineer). It was the same with other "complicated" hardware architecture (Amiga, SNES, Saturn, PS2, PS3...), always hard to apprehend but often brilliant in their approach too.

Is Leadbetter a shill? I don't know, but I also don't know if it matters. It's not like he lies in his comparisons, he's sometimes wrong, but he tends to correct those things when they happen. When PS3 games are better, it's typically about how the developers are incompetent, which I think is obnoxious, but ultimately, who cares? If the PS4 versions of games are better, it's not like he's not going to report it.
I don't think he is a shill. Microsoft is just providing content to him as a part of their marketing push and Sony just doesn't probably because Sony's PR guys are satisfied with the public perception of the console.
 
Top Bottom