The Real Differences Between PS4 and Xbox One Performance

I would rather believe names sources vs "anon" (made up) ones. And john carmack is a good source as he has gone out of his way to bash MS over the years and refuse to use direct X. Yet he says there is not so much difference ... well except he goes on to scrub Kinect.
 
I would rather believe names sources vs "anon" (made up) ones. And john carmack is a good source as he has gone out of his way to bash MS over the years and refuse to use direct X. Yet he says there is not so much difference ... well except he goes on to scrub Kinect.

John is more concerned with the VR experience these days than console specs. John hasn't developed one decent game in quite a long time. I would rather hear from a well respected dev that has released at least one very popular title if not multiple titles on current generation systems or in recent times. Mega Texture tech was suppose to change the industry , instead it was the exact opposite.

Going off of specs alone is enough to true. The PS4 is a superior console in terms of graphics but third party developers don't care , they will put out whatever is the lowest common denominator and go from there. Down the line however we might start to see differences in games from third party developers but for the next few years , things will look largely the same.

I think the single best thing in these next generation consoles are Blu Ray discs. The PS3 showed the advantages of having higher optical storage versus much lower with standard DVD's. Now that its all on even ground developers won't have to create tons of new texture sets and thus increase the cost/production of the same game on different platforms. And with better texture sets comes better looking games even at lower resolutions.
 
John is more concerned with the VR experience these days than console specs. John hasn't developed one decent game in quite a long time. I would rather hear from a well respected dev that has released at least one very popular title if not multiple titles on current generation systems or in recent times. Mega Texture tech was suppose to change the industry , instead it was the exact opposite.

Going off of specs alone is enough to true. The PS4 is a superior console in terms of graphics but third party developers don't care , they will put out whatever is the lowest common denominator and go from there. Down the line however we might start to see differences in games from third party developers but for the next few years , things will look largely the same.

I think the single best thing in these next generation consoles are Blu Ray discs. The PS3 showed the advantages of having higher optical storage versus much lower with standard DVD's. Now that its all on even ground developers won't have to create tons of new texture sets and thus increase the cost/production of the same game on different platforms. And with better texture sets comes better looking games even at lower resolutions.

going off of hardware specs alone the difference is very very minimal and it is completely blown our of the water by fanboys, thats all there is to it
 
going off of hardware specs alone the difference is very very minimal and it is completely blown our of the water by fanboys, thats all there is to it

I agree...Plus MS has a TON of experience with DirectX, and I personally think that many are under estimating the powerful eDRAM..I remember when everyone was raging how cool it was that the Xbox 360 had what, 10MB of eDRAM..There was free 4XAA etc to go around..

Now in practice that turned out not to be enough to do 720P, but I think MS will have learned the lesson from that this time around. They bumped it up to 32MB this time, and it is much, much faster at that..Only time will tell, and I certainly don't have a dog in this fight since the last console I touched was playing Tennis with my ex's daughter on the Wii..
 
Going off of specs alone is enough to true. The PS4 is a superior console in terms of graphics but third party developers don't care , they will put out whatever is the lowest common denominator and go from there. Down the line however we might start to see differences in games from third party developers but for the next few years , things will look largely the same.
A spec sheet of hardware doesn't even tell half the story.

That 32MB of eSRAM can store up to 6GB worth of textures at a time using DX11.2, which is only on Xbox One (and Win 8.1). And since that 32MB of eSRAM is faster then Sony's GDDR5 that has 5GB towards games, it can give the Xbox One and advantage.

But again that's on paper. Its up to the devs to actually do it. None will bother with third party games, but first party games should.


Also, we know in desktop usage, a modern CPU has extremely little to do with helping game performance. But when dealing with CPUs that are so slow clock for clock, and one is 1.6Ghz, and the other 1.75Ghz, that may make some difference. Also, having things like audio offloaded to a dedicated card in the Xbox One, and AI calculations able to be done in the cloud, that clears up more CPU usage as well.


Plus the Xbox One on paper after the GPU overclock has about 25-30% less shading performance. But shading performance does not mean total performance. If you overclock your vid card 30%, you aren't going to get 1:1 gains.

Real world physical graphic capabilities will be more like 15-25% different depending on the situation, in favor of the PS4. But when it comes to AI, processing more things at once, etc... the Xbox One should have an advantage. Texture pop-in should be better on Xbox One as well because of DX11.2 and tiled textures. Its just when you get up close to those textures, they could look better on PS4.



So point is, paper specs mean nothing. There is a lot of software tweaks and differences between the two, and while PS4 may be better at some things, Xbox One will be better at others.

But the differences will only matter with first party titles anyway, so no way of comparing directly.
 
I think in this case considering Anandtech has done a once over and given their opinion I'm going to go with the PS4 being a superior console in terms of graphics. Yes , software plays a big part in getting the performance you want to harness but they both are using pretty uncomplicated designs. Sony isn't going for some overly complicated system design in some kind of exotic setup , they chose to go the more direct road in terms of performance.

As I said , both consoles likely won't look any different for the next 2 years or so but down the line the difference will be something you won't be able to ignore with the right kind of game and developer.
 
I have a source that tells me people will buy both quantities and there will be really good games for both consoles. This source also confirmed the internet will have a very small portion bickering about console specs while the world didnt care.
 
I have a source that tells me people will buy both quantities and there will be really good games for both consoles. This source also confirmed the internet will have a very small portion bickering about console specs while the world didnt care.
 
[QUOTE =Godmachine;1040201158]I think in this case considering Anandtech has done a once over and given their opinion I'm going to go with the PS4 being a superior console in terms of graphics. Yes , software plays a big part in getting the performance you want to harness but they both are using pretty uncomplicated designs. Sony isn't going for some overly complicated system design in some kind of exotic setup , they chose to go the more direct road in terms of performance.

As I said , both consoles likely won't look any different for the next 2 years or so but down the line the difference will be something you won't be able to ignore with the right kind of game and developer.[/QUOTE]

2 things are drastically wrong with what you said. The first being actually buying any of the BS that is spouted in this article. I can tell you exactly why 18 CUs is actually a disadvantage to the ps4. I can also tell you exactly how the cpu will bottleneck the gpu in the ps4 well before it gets anywhere near its specs. I can also tell you why GDDR5 in actual throughout when it is responsible for all data actually performs about the same as DDR3.

The second is actually believing these consoles will have a several year optimal time. Its x86 architecture. Engines are infinately better optimized on x86 architecture and the hardware design isnt unique enough for it to take years for developers to figure out how to use it. In fact if either console faces a long optimization period its the one. If sonys devs cant develop well enough on x86 architecture to get everything they can out of that box within a year theyre shit programmers using horrible gaming engines.

The offloading in the XBOX one has a significant advantage with its additional co processors. And offloading IS THE oonlyreason tlou exists
 
Nothing new here

Games designed/imited to Gamecube specs got ported to PS2
Games designed/imited to Xbox 360 specs get ported to PS3

Dev's are a lazy bunch. PC gamers have known this for a looooong time!!

I see where you're going with this but the Gamecube was actually more powerful than the PS2. Xbox -> GCN -> PS2 from fastest to slowest that generation.
 
this article is such complete unverified garbage. i can debunk 3/4's of it just based off my hardware architecture design experience alone

I can tell you exactly why 18 CUs is actually a disadvantage to the ps4. I can also tell you exactly how the cpu will bottleneck the gpu in the ps4 well before it gets anywhere near its specs. I can also tell you why GDDR5 in actual throughout when it is responsible for all data actually performs about the same as DDR3.

I'm game. Go ahead and actually tell us.
 
going off of hardware specs alone the difference is very very minimal and it is completely blown our of the water by fanboys, thats all there is to it

You consider the PS4 having 30% more shader cores and two times the number of ROPs to be "very very minimal"? Like, seriously?
 
You consider the PS4 having 30% more shader cores and two times the number of ROPs to be "very very minimal"? Like, seriously?

Exactly. I've seen this trend before with desktop GPU's and the same arguments as well.

And there will be optimization to be had over the next few years. Going from optimizing games for the PS3 to PS4 will take time for developers to get fully comfortable with. Perhaps optimizing will have a shorter achieve goal on the Xbox One but I can say that all the Xbox 360 only dev's will take likely a few years to really show off what they can do thanks to the advantages of not only a faster system but far more storage from Blu Ray.

There is always a general difference from the launch title graphics to the graphics in Winter time releases years later.

Despite our disagreements we'll see the truth at launch. If Microsoft has nothing to worry about why the last minute clock increase? If it was totally pointless and added very little to the actual performance of the console then why waste millions upon millions of dollars to change something last minute?
 
http://arstechnica.com/gaming/2013/...s-xbox-one-from-accusations-its-underpowered/ > Even Ars Technica had an issue with Penello's numbers.

Carmack on the same article admitted that he hadn't had time with the hardware and thus hadn't benchmarked it, he was going by paper specs alone.

Cecil, the tiled memory resources are exist on OpenGL 4.3 and in the version that Sony designed to be used with their own hardware, to the point that some developers consider it a verbatim copy of one another, with a small modifications to streamline C++ programming.

PS4 also has a dedicated audio DSP, but since it wasn't designed with something like the kinect in mind it is suppossed to be less powerful, but then again it is much MUCH more powerful for GPGPU and you can use GPGPU to for example audio raycast, which the One's DSP can't do.

Real world graphical details will be 35-40% difference, sorry hardware doesn't change and thus even accounting for possible inefficiency in adding extra cores, we know that graphical workloads are massively parallel (and besides the CU's the PS4 has 100% more ROP's that do matter just to deal with possible bottlenecks).

Again, even Ars Technica had issues on the FUD said by Penello, so don't try and keep spreading it around here.
 
Since i can't edit, let's add another fun link:
http://www.anandtech.com/bench/product/777?vs=778

7770 vs 7850 comparison, it has roughly less difference in gflops than what will exist in reality, it has the proper ROP difference tho of a 100% increase.

Sure, you can use a smaller resolution or slightly modify assets and get to a parity of sorts, all for 100$ more.
 
The ROP numbers that have been sited are complete garbage. Which is proven by the fact that the website that first leaked the numbers in the first place completely defined ROP wrong.

Also the PS4 does not have a dedicated audio DSP. The only thing Sony has alluded to is a dedicated compression/decompression chip which is a very very small part of audio processing. That means the CPU is still responsible for doing the actual DSP which is a lot more work than anyone thinks and will tie up a fair amount of processing power

As for the CUs. There is an extreme bottleneck on them after a certain amount, and its at the fabrication level. In fact this last month I sat in on a PhD dissertation where the candidate presented this very thin on current and past SoC chipsets. Its resistance. After 10 CUs the resistance of the chip increases dramatically making the return on additional CUs minimal in comparisson to the first 10. He didnt compare 12 to 18 but he did compare 10 vs 15 and there was an overall gain of 10%. And the only way around this was to use Tech that flat out can not exist in the amd APU because its quantum tech such as carbon nano tubes. I forget who but a chip authority a month back even directly stated the diminishing returns of additional CUs.
 
Also 30% shader cores does not equate to 30% performance increase by any dream. Not even remotely so
 
Wow... you speak like we don't have working graphics cards with way more CU's... i am just baffled... wow >.<

And the ROP numbers are correct, but well, the CU thing you said, that alone... the level of facepalm... it is like you haven't realized in which forums you are posting.
 
Find me a source that uses a correct definition of ROP and quotes the numbers. Along with a quote by microsoft. And yes I do realize what forum I am on which is why wouldnt expect people to know how semiconductor technology and fabrication works. I am a hardware architect and have spent time as a VLSI design engineer for intel. And no process and I mean no process is strictly linear. Period. An assumption is being made when they are saying 50% Increase in cus means 50% better which is false
 
Just to add salt to a wound:
Graphics related Computing Unit is a term used exclusively by AMD, it isn't a generally used term, you can read about it here:
http://www.anandtech.com/show/4455/amds-graphics-core-next-preview-amd-architects-for-compute/4

And if you would only know, the biggest single chip that is available to standard users from AMD is composed of 32 CU's in the form of the 7970.
Since each Compute Unit as defined by AMD, which are the only one that use them, is composed of 4 SIMD (Single Instruction Multiple Data) arrays, that which is composed of 128 of those arrays.
Each array is composed itself of 16 Stream Processors (as each SIMD has a vector width of 16) then you have a total of 2048 Stream Processors.

And due to the massively parallel data management of graphics it isn't perfect but as the link i showed before, it is very darned close.

And ROP's are also VERY defined for this architecture. Honestly they aren't using anything new whatsoever, sit down and get to read info that exists since 2011, it is called AMD's Graphics Core Next.

BTW if you wanna get laughed again, go to the videocard forums and say that thing about the suppossed lecture where going from 10 to 15 CU's (term that you used as if it was a broad definition but actually only refers to how AMD defines themselves), would only grant a 10% increase.

You can post that kind of crap on other forums, here won't fly, same as in most tech related forums... go to IGN or similarly.
 
Ironically im talking about APUs specifically. Of course dGPUs have higher levels of tolerance to increasing component counts. Thats the nature of them. If you want to get laughed at go onto an actual discrete component board and tell them in an APU scale process that you can infinitely add components without any decreased gain because as we all know especially on that scale that resistance is negligable right?
 
The number of CUs alone does not define performance. Period. End of story. The same with shader cores. The reasons you see that high of component counts in dGPUs is because they have the architecture to support them. Architecture that simply cannot exist in an APU with our current technology nodes.

When you downscale a components overall layout size you also decrease the width of the metals used in routing and that exponentially increases resistance. Which is why the added benefit of additional CUs diminishes
 
Your also completely ignoring what I said about the ROP reports in the first place. I said that the reports all use an incorrect definition for what an ROP is which nullifies their reported numbers. I never said amd doesnt have it defined but the reports clearly do not know what it even is so how can I trust their numbers?
 
Vladdracule you never broke down other posts in the Console forum. You're still saying you're a hardware engineer and say you can lay down some serious knowledge how about you actually do that.
 
So basically, does not matter that the PS4 has superior capabilities. Unless it's a super easy tweak to make the PS4 version look better ( if the game is both PS4 and Xbone ) then the PS4 version will be neutered to look just as bad as the Xbone version out of simplicity. That's what I got from the article.

True for 3rd party, cross-console games, but not necessarily for 1st party games -- and 1st party PS4 games is pretty much the only reason I'm interested in the console. If the game's cross-console, there's a high probability it's also coming to PC which is where I'll buy the game buy default. I'm pretty sure the next iterations of God of War and Uncharted will be pushing the envelope.
 
Find me a source that uses a correct definition of ROP and quotes the numbers. Along with a quote by microsoft. And yes I do realize what forum I am on which is why wouldnt expect people to know how semiconductor technology and fabrication works. I am a hardware architect and have spent time as a VLSI design engineer for intel. And no process and I mean no process is strictly linear. Period. An assumption is being made when they are saying 50% Increase in cus means 50% better which is false

Good god this post has e-peen writing all over. :rolleyes:
 
I think the big difference will be resolution. The PS3 has a crappy GPU so to meet their framerate targets developers often have the PS3 version render at a lower resolution than the Xbox 360.

This time around the Xbox One has a crappy GPU so it will be the one where games are rendered at a lower resolution.

You don't really notice that much when you play the games, but in a side by side comparison the difference is very apparent.

Of course, if you want resolution you buy the PC version :)

The PS3 doesn't have a "crappy" GPU, it's pretty much parity with the 360 (essencially 7800 GTX vs 1900XT). It's the lack of memory pool (256mb) on the PS3's GPU that kills it.

The divide between GPUs is much more straight-cut this generation: both consoles are using the same Southern Islands-based GPUs, but the Xbox One has 30% less shaders than the PS4. Than there's the faster GDDR5 on the PS4... And the 6% boost in core clock speed is in no way making up for these deficiencies.
 
I forget who but a chip authority a month back even directly stated the diminishing returns of additional CUs.

Many of your ideas seem to mirror the same talking points of Albert Penello. I have no idea if he's right or wrong, but after getting torn a new one over at Neogaf, he officially "tapped out" on the conversation today (refuting the Xbox One vs PS4 performance). Hopefully he'll have one of his engineers step in to continue the discussion.

The only thing I've gathered from the arguments over there is that the PS4 has an overall performance advantage (whether slight or significant is still open to debate); games on both systems should have parity at launch (any major differences should be apparent on multi-platform, unless intentionally crippled); several game developers have stated that the PS4 is more powerful and significantly faster to write for (so any crippling, if actually an issue, shouldn't last long imo); and $100 cheaper/but Kinect, is faster, smaller/but Titanfall, VitaTV/but secret dGPU, PC > everything =)

Dunno.. both should be alright as far as consoles go. I made up my mind a month ago that I would get a PS4 instead of the Xbox One. For many reasons, but mostly size, performance advantage/parity, price, and mainly because the PS4 appears to be swamping the One in pre-order sales, and I'm hoping if more successful, the PS4 will have the larger library over it's lifespan.

Unfortunately, I can't find a pre-order for the PS4 anywhere, just the One.
 
Just to add salt to a wound:
Graphics related Computing Unit is a term used exclusively by AMD, it isn't a generally used term, you can read about it here:
http://www.anandtech.com/show/4455/amds-graphics-core-next-preview-amd-architects-for-compute/4

And if you would only know, the biggest single chip that is available to standard users from AMD is composed of 32 CU's in the form of the 7970.
Since each Compute Unit as defined by AMD, which are the only one that use them, is composed of 4 SIMD (Single Instruction Multiple Data) arrays, that which is composed of 128 of those arrays.
Each array is composed itself of 16 Stream Processors (as each SIMD has a vector width of 16) then you have a total of 2048 Stream Processors.

And due to the massively parallel data management of graphics it isn't perfect but as the link i showed before, it is very darned close.

And ROP's are also VERY defined for this architecture. Honestly they aren't using anything new whatsoever, sit down and get to read info that exists since 2011, it is called AMD's Graphics Core Next.

BTW if you wanna get laughed again, go to the videocard forums and say that thing about the suppossed lecture where going from 10 to 15 CU's (term that you used as if it was a broad definition but actually only refers to how AMD defines themselves), would only grant a 10% increase.

You can post that kind of crap on other forums, here won't fly, same as in most tech related forums... go to IGN or similarly.

Um it exists since the 6970 aka Northern Islands that switched from 5 SIMD to 4 SIMD. GCN is just an improved version of that.
 
I'm buying a PS4 a few months after launch, Xbox dOne kinda killed it's interest for me after the initial announcements, and then the prompt BS spin, and back peddling at E3. Don't care if it's better by .01GHz or 1MB of RAM, I can't trust a system that can have those features originally announced switched back on without even telling anyone.

And Microsoft aren't exactly known for getting the system right for launch, they relied on someone figuring out the problem, and supplying an aftermarket fix, until the third revision of the console seemed to have resolved most of the trouble (what's with the noisy as f#@k fans when they speed up though?), and the original xbox had problems at launch as well.
 
Also 30% shader cores does not equate to 30% performance increase by any dream. Not even remotely so
It's not just shader cores. There is significant difference in pixel pushing ability (32 ROP on PS4 vs 16 ROP on Xbone) and the huge difference in typical gaming memory bandwidth (some Xbone developers should be able to exploit the eSRAM for certain tasks and situations). No offense, but there is a huge difference in performance based on specs, particularly when comparing configurations of the same hardware generation. It's not anti-MS FUD or whatever people lie to themselves about. That is just predictable performance based on how GPUs work.

But since I'm not a console developer, and neither are you, but here's someone (former creative director at People Can Fly) who has talked to AAA developers about the consoles' abilities:
https://twitter.com/adrianchm/status/375992980634480640

Adrian Chmielarz
&#8207;@adrianchm
I know that pure hardware specs are not the whole story, but man, all next-gen AAA devs I talk to say it's 50% speed difference...

Oh. Snap. He. Went. There. Sorry. Bro. Mythbusters: Confirmed.

MS didn't really do anything wrong when designing the Xbone. Seriously. It's just that Sony reach much higher and got a huge advantage this generation. In a way, it's funny how it worked out. Last gen MS stealthily benefited from work Sony was doing with IBM, and was able to squeak out a pretty decent triple core CPU, partially at Sony's expense. Nintendo also flubbed it's 1/2 gen update too. It happens, bro.
 
Just a click bait repost. A bunch of un named sources with no cited companies or people.
Could have been written by Sony for all we know.
nice repost, the click bait seemed to work.
 
I think the big difference will be resolution. The PS3 has a crappy GPU so to meet their framerate targets developers often have the PS3 version render at a lower resolution than the Xbox 360.

This time around the Xbox One has a crappy GPU so it will be the one where games are rendered at a lower resolution.

However, we are 7+ years on with a new GPU that's probably 4 times more powerful than the GPUs in the 360/PS3 so I would hope this wouldn't be an issue.

Glad to see however, that it's been mentioned that developers are in the business to make as much money as possible so to do that you produce multi platform titles that are broadly the same. Been saying this a lot to no effect.

However, as mentioned, the validity of this article reminds me of other such articles from 7+ years ago.
 
However, as mentioned, the validity of this article reminds me of other such articles from 7+ years ago.

Seven years ago people argued the 360 would have better multiplatform titles (based on specs and ease of programming), Sony fans revolted and declared Cell would work once developers figured it out. Now 7+ years later and the console that started with the better end of the stick (360) for multiplatform games still has the better multiplatform games.

I figure this will work out in Sony's favor this time, and maintain that slight but notable advantage throughout the gen.
 
Could it be that MS plans on a new console in say 4 years rather than the longer wait we had this time round?

Who knows.
 
As for the CUs. There is an extreme bottleneck on them after a certain amount, and its at the fabrication level. In fact this last month I sat in on a PhD dissertation where the candidate presented this very thin on current and past SoC chipsets. Its resistance. After 10 CUs the resistance of the chip increases dramatically making the return on additional CUs minimal in comparison to the first 10.
I'm incredibly confused by this statement. How was he defining "CU"? The "CU" in this context pertains specifically to the computational units in architectures designed to efficiently solve embarrassingly parallel problems. In other words, GPUs. In this context, an increase in computational units ("CU"s, "SP"s or whatever you wish to call them) yields almost perfect performance scaling in terms of computational throughput.

In what context was he describing "resistance"? In the electrical domain? In the parallel performance domain? How exactly was he using this term?

What was his name? Is the slide desk available?
 
Um it exists since the 6970 aka Northern Islands that switched from 5 SIMD to 4 SIMD. GCN is just an improved version of that.

Still, it is an AMD specific construct, and the Southern Islands model is the one used for these SoC's.

BTW i'm honestly interested, about how "it wouldn't work" in an APU, in an APU you still have the exact same structure as a discrete GPU with closer datapaths, but if according to this fella they can't have that, i wonder how are the current developers getting a proper performance of around 170GB/s (very close to the theoretical peak of 176) with the working system while the real world peak of the one has been of around 133GB/s... and yes i am talking bandwidth because that would be the first thing needed to keep the graphic units properly fed of data.

Also i gotta ask where have you seen that ROP's are improperly defined, because again, this is a widely known architecture and i am pretty sure that if ROP's weren't real Raster Operations Pipelines they would have yelled foul, same as they would have on beyond3d and the other forums were actual developers have posted "keeping it inside NDA".
 
Also to the designer, i always said 35-40%, which would be much more realistic than "around 10%".
 
Back
Top