• 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.

EuroGamer: More details on the BALANCE of XB1

This is obviously game dependent, but if the games they looked at scaled better with a 6.6% upclock than a 16.6% increase in ALU, then it suggests they were being limited more by other parts of pipeline. ROPs is the easy one to finger.

It's interesting that in article where Microsoft want to promote the "balance" of their systems, the numbers that they spit out would suggest otherwise. If you're so fill limited that adding extra functional units has barely any increase in performance then how can you argue that you're balanced at all? All that would suggest is that you have a serious problem with a bottleneck elsewhere in the pipeline.
 

IN&OUT

Banned
Does that mean MS thinks that PS4 will have more bottlenecks?



For CPU maybe, but having a better GPU is always better.
for most of the console games

Killzone devs said that PS4 has no performance bottlenecks which is an ubsurd statment IMHO but you can derive from it that PS4 is very efficient system.

The PlayStation 4 has been so carefully designed that developers won't find a performance bottleneck, Michiel Van Der Leeuw, technical director at Guerrilla Games, has told Edge.

http://www.videogamer.com/ps4/killz...ce_bottlenecks_claims_killzone_developer.html
 
It's kind of funny observing Microsoft's constant re-juggling of their PR message hoping something sticks. If at first you don't succeed, try, try again...

"Cloud is going to give you more computational power!" - No, it's not.
"We never targeted the best specifications on our machine." - Fair enough, but...
"We up-clocked the CPU, this makes us better!" - Well, it is a good start but...
"Everything is going to be even, 50% is not going to happen!" - Yeah, 50% is probably not what we'll see but...
"... this shit is balanced. It's been designed to work together. Basically, it's secret sauce." - Okay but... the whole initial push of the PS4 was how there is no bottlenecks and how it's perfectly balanced.


And yet every single step of the way they pick up some stragglers who have been waiting to be wrapped in the warm bosom of Microsoft once again.
 
"Yeah, I think that's right. In terms of getting the best possible combination of performance, memory size, power, the GDDR5 takes you into a little bit of an uncomfortable place"
GDDR5, so good it's uncomfortable.

I :lol'd

1315770935271.gif
 

REDRZA

Banned
Thanks to Sporran.

On the ESRAM:


On balancing the GPU:


http://www.eurogamer.net/articles/digitalfoundry-vs-the-xbox-one-architects

If you guys find something more interesting, throw it up and I'll add to the OP, obviously keeping the word count of quoted text to a minimum.

Last sentence makes me think there are yield issues. Says they toyed with opening up the two redundant CUs (which would much harder on yield %) vs. the GPU over lock of 6.6%. For them to have gone with the over clock instead of 14 CUs means there was no way they would have enough yields on there chip production which would have effected launch.
 

benny_a

extra source of jiggaflops
It's interesting that in article where Microsoft want to promote the "balance" of their systems, the numbers that they spit out would suggest otherwise. If you're so fill limited that adding extra functional units has barely any increase in performance then how can you argue that you're balanced at all? All that would suggest is that you have a serious problem with a bottleneck elsewhere in the pipeline.
If I remember correctly back when Orbis was still at 4GB and 192GB/s bandwidth for its memory people because people said the GPU is not strong enough to take full advantage of it and that was one of the reasons why the move to 8GB with 172GB/s bandwidth was unanimously seen as a great trade-off.

I mean ROPs probably play a big part, but what about the main memory bandwidth in the Xbox One that is lower than the PS4 that might mean the GPU can't take advantage of that?
 

gofreak

GAF's Bob Woodward
Absolutely but launch titles are hardly representative of future workloads.

Agreed. Especially with so many cross-gen games, the ALU:fill or ALU:tex burden of these games might be quite low.

As shader complexity increases the bottleneck may well go off ROPs to CUs. Which won't be a good story for XB1 either but anyway...
 
Yeah.

This isn't to say a PS4 would be 90% or whatever better on these workloads - the bottleneck would move up the pipe to somewhere else before you got to that in most cases - but it does also reveal something about the ALU talk.

12 CUs might be 'balanced' - for a GPU with that ROP configuration. Maybe in typical (early-gen) shader workloads more ALUs would not help framerate with that output capacity.

But if you had 32 ROPs, and your bottleneck moved up the pipeline, you could certainly do with more CUs to hold on to performance gains. A discussion about CUs and balance without touching on these contexts is an incomplete one to say the least.

You might even say it's deliberately misleading. Hmm. Imagine that!
 
Yeah.

This isn't to say a PS4 would be 90% or whatever better on these workloads - the bottleneck would move up the pipe to somewhere else before you got to that in most cases - but it does also reveal something about the ALU talk.

12 CUs might be 'balanced' - for a GPU with that ROP configuration. Maybe in typical loads more ALUs would not help framerate with that output capacity.

But if you had 32 ROPs, and your bottleneck moved up the pipeline, you could certainly do with more CUs to hold on to performance gains. A discussion about CUs and balance without touching on these contexts is an incomplete one to say the least.

What's so great about "balance" in this context anyway. I'd take an unbalanced system that is never fill or bandwidth limited that allows me to get closer to 100% utilisation of my compute units than a "balanced" system that is bottlenecked by fillrate and bandwidth as much as it is compute.

The latter may be more "balanced" but I've got three separate bottlenecks to contend with and my compute utilisation is ultimately lower.
 

Ushae

Banned
I have no idea what any of this means. I'll just enjoy my consoles and games when they get here. Good day gentlemen.
 

IN&OUT

Banned
Microsoft technical fellow Andrew Goosen inner thoughts looking at PS4 architecture design:

" You are a goddamn artist Cerny!"
 

TheD

The Detective
The reason that an upclock helped them more than 2 more CU's is not because of CU's not scaling "in a linear fashion" in what they do, it is because the upclock not only increased the clock speed of the CUs, but also other parts like the ROPs and TMUs!
The ROP stage and the TMU stage in the PS4 are already more powerful off the bat and thus the CUs are not going to be bottlenecked by them as soon as in the XB1 GPU.
 

sirap

Member
I'm even more confused after reading this. They've tossed out so many numbers and words I don't know who to believe anymore.

That said, we could all use a little Balance in life.
 
I have no idea what any of this means. I'll just enjoy my consoles and games when they get here. Good day gentlemen.
I want to avatar quote you so bad lol :p..but im pretty much same thought as you. Still its interesting to try and learn and follow some of these discussions before launch since we still got 2 months left :/
 

gofreak

GAF's Bob Woodward
What's so great about "balance" in this context anyway. I'd take an unbalanced system that is never fill or bandwidth limited that allows me to get closer to 100% utilisation of my compute units than a "balanced" system that is bottlenecked by fillrate and bandwidth as much as it is compute.

The latter may be more "balanced" but I've got three separate bottlenecks to contend with and my compute utilisation is ultimately lower.

Yeah, the question of utilisation of parts relative to each other vs what is better for absolute performance is also sidestepped if we focus purely on the former.

Just pointing out how they neatly avoid discussing why these decisions are 'balanced' or produce positive effects. They're ignoring the less flattering contexts although are inadvertently revealing them.
 

Perkel

Banned
Ladies and gentlemen:

"Yeah, I think that's right. In terms of getting the best possible combination of performance, memory size, power, the GDDR5 takes you into a little bit of an uncomfortable place," says Nick Baker

(-;-)_oO
 

Chobel

Member
I read the the article many times but still I have no idea how did they get 204GB/s number.

Why not 218GB/s? Heck why not another random number like 206GB/s?
 
The reason that an upclock helped them more than 2 more CU's is not because of CU's not scaling "in a linear fashion", it is because the upclock not only increased the clock speed of the CUs, but also other parts like the ROPs and TMUs!
The ROP stage and the TMU stage in the PS4 are already more powerful off the bat and thus the CUs are not going to be bottlenecked by them as soon as the XB1 GPU.

Actually, it's mostly because getting 6% more performance for free looked way better versus ~15% more performance at the enormous cost of having to throw away every single chip that had any defective CUs.

I read the the article many times but still I have no idea how did they get 204GB/s number.

Why not 218GB/s? Heck why not another random number like 206GB/s?

Because it's not really a doubling of performance. It's creative accounting based on access patterns that can take advantage of the way the ESRAM memory is banked. The 204 figure is itself never actually achievable. Favorable scenarios get you like 2/3rds of the way there.
 

benny_a

extra source of jiggaflops
I read the the article many times but still I have no idea how did they get 204GB/s number.

Why not 218GB/s? Heck why not another random number like 206GB/s?
I also didn't see it the first time. Check the side-bar. It's not repeated information as it often is on Eurogamer but unique.

Alternatively search for "bubble".

Actually, it's mostly because getting 6% more performance for free looked way better versus ~15% more performance at the enormous cost of having to throw away every single chip that had any defective CUs.
But every consumer Xbox One is a devkit according to Penello. So that means the yield still needs to account for 14 CU's all working fine.
 

TheD

The Detective
Actually, it's mostly because getting 6% more performance for free looked way better versus ~15% more performance at the enormous cost of having to throw away every single chip that had any defective CUs.

I was basing what I said on the idea that they did get better performance from the upclock, it is very much possible that what you said is right and that it is just a yield thing.
 

repeater

Member
And yet every single step of the way they pick up some stragglers who have been waiting to be wrapped in the warm bosom of Microsoft once again.
Bingo, I think this is the actual point rather than any honest belief in the relevance of what they're themselves saying. They know that there are large groups of people that are either 1) completely unskilled in or uninterested in the kind of technical discussions needed to see that MS aren't actually answering the worries that have been raised, or 2) already sufficiently emotionally invested in MS to "have faith" in their message despite the obvious concerns. Neither of these two groups will be affected by the fact that there seems to be large gaps in MS' argumentation here - the first group because they can't be bothered to delve into the discussion (which is really fair enough), the second because they will more or less disregard that discussion completely.
 
"Balance" can mean "it's ok having an under-performing gpu because our ram and cpu under-performs in the same quantity!". Everything have the same poor level, no bottlenecks!
 

TheD

The Detective
Oh I misread. My initial point from the first page still stands then: All Xbox Ones have 14 CUs and they could unlock those. ;-)

No, you could not unlock them unless you want a large number of systems to output corrupt graphics.
 
It's good to finally hear the facts from the people that designed the machine instead of unsubstantiated rumours. I feel confident that the Xbox One will produce graphically impressive games and that difference between the One and the PS4 will be small. There's no doubt that the PS4 is stronger, but in my opinion not appreciably so.
 
LOL far more interesting info in the article, but the OP ignores it in favor of poking fun at Microsoft having the nerve to say their design is balanced. You already know the purpose of this thread just from the title. :)

Yeah... from a mainboard layout perspective GDDR5 is actually more space-saving and less complex (and maybe even cheaper) than DDR3.

Also utilizes quite a bit more power than ESRAM, too... They aren't focusing on any one thing. They're focusing on the overall implications of choosing GDDR5. You don't have to agree with them. In fact, most people won't, and are justified, but they gave their reasons, and those reasons were clearly valid enough in their minds to make the decision they made. Forget about space saving. Power is a major concern also, and ESRAM and DDR3 as a combo uses less power, believe it or not.
 

Chobel

Member
I also didn't see it the first time. Check the side-bar. It's not repeated information as it often is on Eurogamer but unique.

Alternatively search for "bubble".

Thank you! I totally missed that
The same discussion with ESRAM as well - the 204GB/s number that was presented at Hot Chips is taking known limitations of the logic around the ESRAM into account. You can't sustain writes for absolutely every single cycle. The writes is known to insert a bubble [a dead cycle] occasionally... one out of every eight cycles is a bubble so that's how you get the combined 204GB/s as the raw peak that we can really achieve over the ESRAM. And then if you say what can you achieve out of an application - we've measured about 140-150GB/s for ESRAM
 

nib95

Banned
It's good to finally hear the facts from the people that designed the machine instead of unsubstantiated rumours. I feel confident that the Xbox One will produce graphically impressive games and that difference between the One and the PS4 will be small. There's no doubt that the PS4 is stronger, but in my opinion not appreciably so.

All the evidence and numbers point to the contrary, so not sure why you have that opinion besides the fact that it's the one Microsoft's PR is touting and wants you to believe, for obvious reasons.
 

stryke

Member
I have no idea what any of this means. I'll just enjoy my consoles and games when they get here. Good day gentlemen.

Ahh the inevitable "I'll enjoy the games and not worry about the specs" type comment in every tech thread.
 

Chobel

Member
LOL far more interesting info in the article, but the OP ignores it in favor of poking fun at Microsoft having the nerve to say their design is balanced. You already know the purpose of this thread just from the title. :)

maybe you should read the OP, this part:

If you guys find something more interesting, throw it up and I'll add to the OP, obviously keeping the word count of quoted text to a minimum.
 

artist

Banned
Oh I misread. My initial point from the first page still stands then: All Xbox Ones have 14 CUs and they could unlock those. ;-)
Couple that with a upclock to 1.29GHz and suddenly they secure a lead with exactly 768 flops. Coincidence?
 
All the evidence and numbers point to the contrary, so not sure why you have that opinion besides the fact that it's the one Microsoft's PR is touting and wants you to believe, for obvious reasons.

Games have to show that. So far they aren't. Everybody knows the PS4 is more powerful, but people are in valid and justified disagreement about how drastically that difference will show itself in actual games on the two systems. If it will be nothing more than a higher resolution with games more or less looking just as incredible on both, then I think the people who said the building likely wouldn't be on fire over at MS were in the right. If there this dramatic and suitably massive difference that people are expecting shows up, then you guys will be correct.

Thankfully, whichever side is right, I'll benefit both ways, since I'll have both systems :p
 

Mung

Member
feel confident that the Xbox One will produce graphically impressive games and that difference between the One and the PS4 will be small. There's no doubt that the PS4 is stronger, but in my opinion not appreciably so.

Good that the console that costs less only has a small performance handicap. Oh wait...

image.php
 

benny_a

extra source of jiggaflops
Games have to show that. So far they aren't.
You have footage of multiplatform games where we can analyze the differences on a common engine and game with the same workload?

Penello said the same thing that nobody can see a difference and when I asked him he never replied. I hope you do.
 

StuBurns

Banned
It's good to finally hear the facts from the people that designed the machine instead of unsubstantiated rumours. I feel confident that the Xbox One will produce graphically impressive games and that difference between the One and the PS4 will be small. There's no doubt that the PS4 is stronger, but in my opinion not appreciably so.
What is 'appreciably' though? Is 900p versus 1080p appreciable?
 
I just posted this to the "PS4 and Xbox One performance" thread before noticing this one.

It seemed like from the article that one of Microsoft's main defenses of their less powerful GPU was that most games are CPU limited anyways so that the extra power of the GPU would not matter. My question is "Is this true? Are most console games really CPU limited?"

I have a suspicion that this might be true for Microsoft because they want to be able to run other apps while a game is running.​
 
You have footage of multiplatform games where we can analyze the differences on a common engine and game with the same workload?

Penello said the same thing that nobody can see a difference and when I asked him he never replied. I hope you do.

Dude, Ryse stands right now as, imo, the most amazing looking game on either system. I don't need to focus on all these multi-platforms that I haven't seen yet. Dude, I've seen Fifa on both PS4 and Xbox One. Do you see a huge difference there, either? That's one game we can talk about that's multi, right? Oh, but there's a catch, I know. It's FIFA :p

Whatever the case, I'm not living and dying on the numbers. We already know sony's kicking MS ass on the raw numbers. Nobody denies. Nobody is even denying that it will have an impact. Sony's console is stronger, period. But just how much will that show up in the real world is the question. That's all I'm saying. Personally, I would love if the PS4's advantage shows up every bit as much as the raw numbers indicates, because I'm going to own the system. However, realistically, I don't think it's going to happen. Call me one of those stubborn people that have to see it before I believe it. Is that so wrong?
 

chris0701

Member
Assuming level scaling of compute power with the addition of two extra CUs, the maths may not sound right here, but as our recent analysis - not to mention PC benchmarks - reveals, AMD compute units don't scale in a linear fashion. There's a law of diminishing returns.

So power up to 14CU doesn't peform better than raising 7% GPU clock.

Of course you can say the X1 ROP limits it, but they said PC benchmarks as well.
 

benny_a

extra source of jiggaflops
Whatever the case, I'm not living and dying on the numbers. We already know sony's kicking MS ass on the raw numbers. Nobody denies. Nobody is even denying that it will have an impact. Sony's console is stronger, period. But just how much will that show up in the real world is the question. That's all I'm saying.
But multiplats are the only way we can judge them, as everything else is subjective. Artstyle and genre and stuff all color perception.

Also I haven't seen FIFA on PS4 yet, only Xbox One and that was off-screen too. But yeah, that's not a game I would expect a difference except for a little slowdown when it comes to replays or something.
 

nib95

Banned
Games have to show that. So far they aren't. Everybody knows the PS4 is more powerful, but people are in valid and justified disagreement about how drastically that difference will show itself in actual games on the two systems. If it will be nothing more than a higher resolution with games more or less looking just as incredible on both, then I think the people who said the building likely wouldn't be on fire over at MS were in the right. If there this dramatic and suitably massive difference that people are expecting shows up, then you guys will be correct.

Thankfully, whichever side is right, I'll benefit both ways, since I'll have both systems :p

Even at this early stage, I'd say the games are indeed showing the difference. Not just in terms of graphical features etc they're pushing, but also in the different resolutions and frame rates being achieved by the platforms. All of Sony's exclusives being 1080p, whilst exclusives on the Xbox One are struggling to keep that resolution, already with games releasing at 720p, 900p and one at a potential dynamic resolution.

Not that it matters much since launch games aren't indicative of much. Both consoles will have games that look far better than the launch stuff.
 
Dude, Ryse stands right now as, imo, the most amazing looking game on either system.

It is a 900p title. You are honestly saying it looks better than Killzone / Infamous both which are doing incredible things on screen while rendering 44% more pixels? When we have high quality footage of (several!) multiplatform titles then we can make some proper, first comparisons.
 
Top Bottom