This silly comparison never dies ! I won't even try to explain it anymore.
-
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.
EDGE: "Power struggle: the real differences between PS4 and Xbox One performance"
- Thread starter Musiol
- Start date
This silly comparison never dies ! I won't even try to explain it anymore.
that's not true. The whole point panello is making and the architects expect to share is that the system is designed in ways that you can not compare apples to oranges and that we don't know yet how all the fruit works in xbone so that's false conclusion no matter how many times "we keep explaining" that it is
Obviously not enough....
It's probably going to be something along the lines of "12 CU's =/= 50% more performance", and "Our peak on paper is 272gb/sec. (68gb/sec DDR3 + 204gb/sec on ESRAM)".
Penello has already tried to float this argument, I believe.
edit:
The only major difference between the two is the eSRAM and its pretty obvious what its point is. Theres no hidden secrets or fruits we have missed
SwiftDeath
Member
Besides the different approaches to memory, there is very little that could be revealed that would truly impact performance
The audio chip among other special processors will simply not have as much impact as the raw performance bonus the PS4's GPU will have
The memory debate is always well worth discussing (and has been) due to the accesiblity and bandwidth capabilities
This generation comparing architectures is far more sensible than previous
I don't understand how anyone could argue otherwise
Edit: Beaten by beta
It is true. We know more than enough about both systems to make this conclusion. No amount of Data Move Engine or ESRAM fuzzy math is going to take away from the simple fact that you've got AMD APUs powering both machines, except one's got a meatier GPU than the other.
well I guess we will find out when the details that we do not yet know about how the memory subsystem, helper chips etc are actually designed to work.
I find it interesting that people claiming to have scientific basis for "proving" that xbone is x% weaker in games are so resistant to learning new information upon which to form conclusions.
I suspect the helper chips are just what we already know about (move engines, display planes, etc) and we already know how they work and there performance rather well.
Alrighty then. I'll let you fellas get back to it in here then since there obviously is nothing more to learn about these machines from the engineers who actually spent years of their lives designing the system.
If only they could have known that all they needed to do was grab some parts off a shelf with numbers on them and slap them in a box.
Carry on
If only they could have known that all they needed to do was grab some parts off a shelf with numbers on them and slap them in a box.
Carry on
Something said at the TGS Keynote made me think about this a little more. Sony said they ORIGINALLY shipped their units with 4GB of DDR3 and after some rather forceful suggestions by developers, they went with 8GB of GDDR5.
This made me wonder if some of those developers weren't working with some of the Xbox hardware already and simply wanted it on par and made a few suggestions. If this is the case, the surprise of MS would have been legitimate. I do think this also makes sense when you consider the 4.5 GB of OS dedication on Sony which seems to imply it hasn't been optimized and possibly thrown last minute.
Granted this is all conjecture but it would flow with what has been historically true. Xbox usually does what Devs want first and foremost (larger devs generally). Sony, perhaps thinking 4GB was good enough gets a surprise and doesn't want a repeat of last gen. They didn't get "inside info" but possibly through reaction got a system that was superior.
Either way, this also leads me to believe the statements that we need to see results first.
A friend of mine said something I agree with, "In a few years when they write the book of backroom decisions, I'm going to LOVE to read about it."
I totally agree. -Adam
I didn't say that, but we already know a large amount about this system to think that there will be massive reveal from that will boost its performance or suddenly show the light to everyone about how powerful is will only leave you disappointed.
I think they are really going to go deep into explaining the enhancements they made to the SoC this time.
I think they get the point now that they are going to have to lay it all on the line & put everything out there this time.
PS3 and Xbox have strengths and weaknesses against one another, and the reason the PS3 ports generally perform worse is because the SPUs take a lot more time to properly use and the memory setup is generally worse.
Beerman462
Member
We know one thing...esram and move engines are there to mitigate DDR3s bandwidth disadvantage. The GPU is weaker, because MS needed that extra silicon for esram. Lets not pretend any decisions they have made stem from a performance stand point.
He brought out the fellow card, finally.
I think they are both different in a lot of ways.
eSRAM + DDR3 could have other advantages lets not forget that, just that they aren't going to be applicable in a large amount of cases.
No it was 4GB of GDDR5 not DDR3.
Did they specifically say DDR3 at the keynote? I missed that, and my impression was that it went more like "2GB GDDR5 (very early prototype thing) -> 4GB GDDR5 (before devs convinced Sony to put in MOAR RAM) -> 8GB GDDR5." Would be kinda surprised if they were really thinking about using DDR3 as well, though I suppose Cerny's comments about the consideration of EDRAM at one point would make that perspective not too outlandish.
Same here, heh.
useyourfist
Member
I never understand why multi plats are used for comparison, when looking at a machines power and performance surely it is more relevant to compare exclusives that have been coded specifically for each platform, and of a comparable genre and design of course.
Beerman462
Member
True, but that's not why they use it.
SolidSnakex
Member
The leaked specs always pointed toward the console having 4gb of GDDR5. It was Randy Pitchford from Gearbox that told Adam Boyes that the console would be done if they didn't push it to 8gb.
Because exclusives are each their entirely own games with different design goals and optimization preferences depending on what the developers of each game want. We've already got a great example between Forza 5 vs. Driveclub - the former stresses 1080p60fps while the other goes for more visual goodies. Sure, when you boil things down to the absolute basics, "they're both just racing games lol," but delving deeper in what the makers of those respective games are actually trying to deliver reveals there's more to what specifically they each want to offer. This in turn muddies any direct comparisons because it leaves you trying to arbitrarily evaluate each others' strengths and weaknesses against one-another, and you get to a point where you'll probably just be arguing subjective preferences with others in the discussion. (Some value prettiness, others performance)
That's why multiplats make the more logical comparison - the developers trying to offer the same basic product across all systems the game is published on, so you don't have to jump through hoops and subjective bickering to make the comparison (or at least, not nearly as much); you just point out what the basic differences are and then come to a conclusion which is better - differences in resolution, anti-aliasing, particle density, texture resolution, etc. are much more straightforward points of comparison compared to entirely different games.
Yeah, I knew about all that; I just figured if there was a DDR3-based PS4 at one point, it would've been some earlier prototype or even just a consideration like the EDRAM + 128-bit GDDR5 setup that Cerny mentioned.
useyourfist
Member
Whilst I agree completely with where you're coming from I also think this argument is flawed due to there being a lead platform with a multiplat then being ported to other platforms, which we know is where ps3 suffered. Wouldnt it be more fair to say only multiplats built separately and coded from the start from each platform to get the best from each platform should be used for comparisons in this sense? Obviously in the coming generation this should be less of an issue but then the argument of wanting to keep parity between consoles could be brought up as a factor that limits the more powerful console...so exclusives appears to me to be the better tool for comparison..unless multiplats are optimised for each system which I cant see happening
The problem with using exclusives is theres a lot more variable in play, less so with multi plat games even though some still exist.
I do expect a lot more of that this gen.
Gemüsepizza
Member
...like the PS4? The PS4 is also heavily customised, they did not just put unmodified off-the-shelf parts in it. But of course this information is in conflict with your "X1 design = efficiency / PS4 design = raw power" nonsense.
What do you mean heavily customized? But Penello said:Gemüsepizza;83102557 said:
"As a matter of fact, they actually go out and they talk about how proud they are about their off-the-shelf parts. Our guys’ll say, we touched every single component in the box and everything there is tweaked for optimum performance."
And Penello wouldn't say something like that if he wasn't 100% confident in that.
And the XBox One is already holding them back according to Ubisoft.
Source?
Engineers (especially MS engineers) never ever make a bad product? MSSQL is the best data base ever? IIS wow no problems ever right? Exchange server clearly works without maintenance or downtime right? Windows ME the best OS? RROD?
It's a pretty silly appeal to authority, the engineers made it. Hurrah. But it's not likely that there is some unnoticed feature that will drastically improve performance. Things like that tend to get noticed and trumpetted. Like the potential in the Cell SPU's or the practicality of universal shaders in the 360.
We page 100 now.
As you said, the whole "lead platform" and "least common denominator" factors should play much less of a role this time around. PS3 multiplats had more than that going against them - basically, the PS3 was its own worst enemy in some ways. Without that working against either system this time (unless ESRAM remains a hurdle throughout the gen, but I still expect MS to sort out issues with that by sometime next year at the latest), it should be much easier for devs to get more out of each system in multiplat games, even if it isn't the most.
Of course, ideally you'd have some multiplat game where the devs do their best to maximize performance across both machines, sure. Unfortunately, outside of something like GTAV that just doesn't really happen. But it still gives you a stronger direct comparison than exclusives. I suppose an angle you could go with in comparing exclusives is to say, "which system does [X] thing best." (i.e. some visual feature like lighting or whatever) Outside of that though, there's just too much in the way to make direct comparisons, and when you're trying to find software to analyze the results of the hardware differences between each machine, it's simpler to work with software found on both machines.
This just in, MS engineers never fuck up.
As you said, the whole "lead platform" and "least common denominator" factors should play much less of a role this time around. PS3 multiplats had more than that going against them - basically, the PS3 was its own worst enemy in some ways. Without that working against either system this time (unless ESRAM remains a hurdle throughout the gen, but I still expect MS to sort out issues with that by sometime next year at the latest), it should be much easier for devs to get more out of each system in multiplat games, even if it isn't the most.
Of course, ideally you'd have some multiplat game where the devs do their best to maximize performance across both machines, sure. Unfortunately, outside of something like GTAV that just doesn't really happen. But it still gives you a stronger direct comparison than exclusives. I suppose an angle you could go with in comparing exclusives is to say, "which system does [X] thing best." (i.e. some visual feature like lighting or whatever) Outside of that though, there's just too much in the way to make direct comparisons, and when you're trying to find software to analyze the results of the hardware differences between each machine, it's simpler to work with software found on both machines.
This just in, MS engineers never fuck up.
I think that blogger koojopanda is reading more into it that than what was intended by the statement.
It's a multiplatform game, so of course they can't go all out on PS4-specific stuff.
Nobody expects them to, even though it would be nice if they did PS4 and Xbone specific stuff in the different builds to get the most out of either system. It's just not feasible for that title that was developed along side an ever-changing SDK.
Chobel
Member
And I thought I couldn't hate Penello more.
The main gaps as I see it are
1) memory bandwidth
2) 12CU vs 18 CU
3) 2ACE vs 8 ACE (longer term benefits perhaps)
For (1), the crux IMO will be how much of a game's rendering time can be spent solely using the ESRAM. If they can get a significant percentage using that, then at least the memory bandwidth can be significantly mitigated. I think intel said 32MB of edram gave then a 90-95% hit rate when used as a cache? But the tools sound a bit manual at the moment, and the ESRAM doesn't behave as a cache, so they need to improve there.
For (2), I don't see how they bridge that gap. Maybe there are some stats that show what amount of bandwidth the GCN architecture can leverage, and maybe there is an optimal number of CUs, and 18 is actually too many for that bandwidth and so cannot be fully utilised. I doubt it, but I'm just throwing it out there. I'd like to see data on the relationship between memory bandwidth and number of CUs in GCN
For (3) this is another area that I think PS4 will just pull further away from Xbox with. You have 50% more CUs *plus* have mechanisms in place to utilise them more efficiently. That sounds like win-win to me. Then add in Sony first party familiarity with SPE coding for CELL, and you have teams perhaps better set up for exploiting GPGPU
I'm sure there is more to Xbox one than the simple stats. It'll be interesting to hear more about the architecture too. But I don't see how they close the performance gap to PS4 in a meaningful way.
1) memory bandwidth
2) 12CU vs 18 CU
3) 2ACE vs 8 ACE (longer term benefits perhaps)
For (1), the crux IMO will be how much of a game's rendering time can be spent solely using the ESRAM. If they can get a significant percentage using that, then at least the memory bandwidth can be significantly mitigated. I think intel said 32MB of edram gave then a 90-95% hit rate when used as a cache? But the tools sound a bit manual at the moment, and the ESRAM doesn't behave as a cache, so they need to improve there.
For (2), I don't see how they bridge that gap. Maybe there are some stats that show what amount of bandwidth the GCN architecture can leverage, and maybe there is an optimal number of CUs, and 18 is actually too many for that bandwidth and so cannot be fully utilised. I doubt it, but I'm just throwing it out there. I'd like to see data on the relationship between memory bandwidth and number of CUs in GCN
For (3) this is another area that I think PS4 will just pull further away from Xbox with. You have 50% more CUs *plus* have mechanisms in place to utilise them more efficiently. That sounds like win-win to me. Then add in Sony first party familiarity with SPE coding for CELL, and you have teams perhaps better set up for exploiting GPGPU
I'm sure there is more to Xbox one than the simple stats. It'll be interesting to hear more about the architecture too. But I don't see how they close the performance gap to PS4 in a meaningful way.
The sarcasm is making my eyes water.
Well Kaz Hirai himself said they are using off the shelf parts:
http://www.psxextreme.com/ps4-news/590.html
Edit: and yes - I've seen the second part of the sentence.
Oh dear. Is this all they have left? To accuse the competition of having off-the-shelf parts? Blatantly lying? They will reach the acceptance stage by next year. They are still on denial or bargaining at the moment.
Edit: And in reply to the post above: a + b does not equal a unless b is 0.
I suppose I can't accuse anyone of selective quoting. Just selective reading.
shinra-bansho
Member
I really don't get why the secret sauce won't die. Reading the Wii U thread and all of these Xbox One threads, there's an almost obsessive constant quest.
It's quite clear looking at the situation the machines we're getting are the product of differing overarching business priorities, competing for space, for power envelopes, for BoM.
Nintendo seemingly prioritised low power draw, a small form factor, backwards compatibility and a touch screen, and were satisfied with performance around or slightly above the current gen systems, and we got the Wii U.
Microsoft seemingly prioritised large amounts of RAM in order to run their OS functions and Kinect in every box, consequently they needed to implement various features to improve available bandwidth meaning a large amount of die space for ESRAM, meaning a larger die but with less CUs. It may have certain benefits in certain situations, as some of the technical GAF folk have alluded to, but it's the result of overarching business drivers not some sort of magical efficiency that will level the playing field. Silicon budget is also competing with the inclusion of Kinect. Microsoft also seem to want to break even or profit on hardware at launch. So we end up with the Xbox One, ultimately weaker, more complicated to develop for, and at the higher $499.
Sony seemingly prioritised ease of development, so they opted for a unified pool of fast RAM. Their refrain seems to be powerful, customized but familiar to developers. Had they not been fortunate enough for densities to increase in time they would have ended up with significantly less RAM in the PS4 than the Xbox One. They didn't need ESRAM, have more room in their silicon budget for compute units at a similar BoM etc. They also seem to think that the future is in off-loading compute tasks, so presumably that's also why they wanted to have a larger number of CUs. We end up with a more powerful, easier to develop for system at a lower consumer cost.
It's quite clear looking at the situation the machines we're getting are the product of differing overarching business priorities, competing for space, for power envelopes, for BoM.
Nintendo seemingly prioritised low power draw, a small form factor, backwards compatibility and a touch screen, and were satisfied with performance around or slightly above the current gen systems, and we got the Wii U.
Microsoft seemingly prioritised large amounts of RAM in order to run their OS functions and Kinect in every box, consequently they needed to implement various features to improve available bandwidth meaning a large amount of die space for ESRAM, meaning a larger die but with less CUs. It may have certain benefits in certain situations, as some of the technical GAF folk have alluded to, but it's the result of overarching business drivers not some sort of magical efficiency that will level the playing field. Silicon budget is also competing with the inclusion of Kinect. Microsoft also seem to want to break even or profit on hardware at launch. So we end up with the Xbox One, ultimately weaker, more complicated to develop for, and at the higher $499.
Sony seemingly prioritised ease of development, so they opted for a unified pool of fast RAM. Their refrain seems to be powerful, customized but familiar to developers. Had they not been fortunate enough for densities to increase in time they would have ended up with significantly less RAM in the PS4 than the Xbox One. They didn't need ESRAM, have more room in their silicon budget for compute units at a similar BoM etc. They also seem to think that the future is in off-loading compute tasks, so presumably that's also why they wanted to have a larger number of CUs. We end up with a more powerful, easier to develop for system at a lower consumer cost.
If you read the sentence in full then I don't know how you reduce what Hirai said to saying that they're just using off-the shelf parts. Unless you're just trying to be facetious.Well Kaz Hirai himself said they are using off the shelf parts:
http://www.psxextreme.com/ps4-news/590.html
Edit: and yes - I've seen the second part of the sentence.
brain_stew
Member
this. half the ms secret sauce argements and our engineers are better arguments are very silly. implying your engineers can do better cause they made direct x is like say google saying we made the best search engine so our engineers are the best and can achieve xbone one graphics on the galaxy s4. ... im obviously overextending ... but the point is unless some ms engineer reinvents algorithms and programming on a basic level no amount of skill will overcome a significant power gap. and the ms pr ppl need to realize that
brain_stew
Member
Well Kaz Hirai himself said they are using off the shelf parts:
http://www.psxextreme.com/ps4-news/590.html
Edit: and yes - I've seen the second part of the sentence.
Microsoft are using "off the shelf" technology as well. Nether are using off the shelf parts.
brain_stew
Member
This is all despite the fact that we have developer quotes stating that Sony are way ahead on tools. If experience writing DirectX can't even get you an advantage in tools then why would anyone believe it can gain you an advantage in hardware engineering (where
Sony has more experience in, as it happens)? It's a dumb false equivalency, a smokescreen of a nothing argument.