The Real Differences Between PS4 and Xbox One Performance

ocellaris said:
We've had full sound processing done on the host CPU since the days of the Pentium 2, it doesn't take a lot of CPU power.
Click to expand...
Takes a little more then to output PCM in 8 channels, then it does processing the 2 channels from back in the day. But yeah it still doesn't take a ton of work but when people bitch about the probable 10-15% real difference we'll see in games why not bitch about the 2-5% of cpu cycles taken up by audio.

Really that % difference is such a shit way to do things. I'm sure weeks after the release we'll get a ton of side by side screencaps of 3rd party games between consoles and it be little differences.

Frankly if people can live without physX people can live without the little bit of eye candy the ps4 will offer. As unless it's a 1st party game the details wont be that drastic between the two. Frankly the ps3 to 360 the ps3 had a larger power gap to the 360 then the one that excists for the ps4->xb1 and yeah in the past 2 years 1st party titles have looked great on ps3 but people who own an xb360 don't care too much.

As xb360 focused more on social aspects and multiplayer games, it was more of a NA market. PS3 most popular in japan focused on single player for 1st party titles and just like japan didn't do other things like dlc as much as japan doesn't buy much online.

If you talk about the main problem with the ps3 and multiplayer it has been ps3 is slower, it's a slower system when you talk about overall input lag, ps3 has an extra frame in it's buffer and the controller to console response for that input buffer is also longer. So ps3 versions of games would be slower then the 360 mostly an issue in fighting games where your 1,2 frame links and going from 360 to ps3 is like playing underwater. Makes me wonder about the ps4's o/s and if they changed that.
 
Semantics said:
Frankly the ps3 to 360 the ps3 had a larger power gap to the 360 then the one that excists for the ps4->xb1 .
Click to expand...



http://www.neogaf.com/forum/showthread.php?p=81909537#post81909537


PS3 | RSX: 176 Gflops and Cell: 230 Glops, Total 406 Gflops
360 | Xenos: 240 Gflops and CPU: 77 Gflops, Total 317 Glops

+36% gpu wise in favor of the 360 vs +28% in favor of the ps3 when accounting the whole system (including optimizations)

PS4 | GPU: 1.84 Tflops and CPU: 100 Glops, Total 1.94 Tflops
Xbox One | GPU: 1.31 Tflops and CPU: 109 Gflops, Total 1.41 Tflops

+40% in favor of the ps4 gpu wise vs +37.6% also in favor of the ps4 when accounting the whole system.

Just saying, numbers this generation do not look nice, and the power gap is bigger in one direction.
 
Revdarian said:
http://www.neogaf.com/forum/showthread.php?p=81909537#post81909537


PS3 | RSX: 176 Gflops and Cell: 230 Glops, Total 406 Gflops
360 | Xenos: 240 Gflops and CPU: 77 Gflops, Total 317 Glops

+36% gpu wise in favor of the 360 vs +28% in favor of the ps3 when accounting the whole system (including optimizations)

PS4 | GPU: 1.84 Tflops and CPU: 100 Glops, Total 1.94 Tflops
Xbox One | GPU: 1.31 Tflops and CPU: 109 Gflops, Total 1.41 Tflops

+40% in favor of the ps4 gpu wise vs +37.6% also in favor of the ps4 when accounting the whole system.

Just saying, numbers this generation do not look nice, and the power gap is bigger in one direction.
Click to expand...

Ok where exactly did you pull these numbers out of? Sony hasnt even so much as released the clock speed on the cpu. Not to mention the information we have suggests the xbox ones cpu is able to do 6 operations per clock cycles as opposed to 4 on the ps4. Also not to mention that the overall system power cant exceed its strongest component. The sum of the parts simply cannot be greater than the individual components in hardware
 
Tbh I'm rather excited to see if we will be able to over clock all this heavily x86 based hardware...
 
VladDracule said:
Ok where exactly did you pull these numbers out of? Sony hasnt even so much as released the clock speed on the cpu. Not to mention the information we have suggests the xbox ones cpu is able to do 6 operations per clock cycles as opposed to 4 on the ps4. Also not to mention that the overall system power cant exceed its strongest component. The sum of the parts simply cannot be greater than the individual components in hardware
Click to expand...

How would this be possible if both processors are using the same Jaguar-based modules?

A single Jaguar module is:

  • 1.0 GHz to 2.0 GHz max
  • 4 cores
  • 2 MB shared L2 cache - "shared cache unit" (SCU)
  • called a "compute unit"
  • x86-64 processor
  • 13 cycle instruction pipeline

Both the Xbox ONE and PS4 are SoCs (system-on-a-chip):

  • 2x Jaguar modules (8 cores total)
  • integrated memory controller-- 2133MHz DDR3 (Xbox ONE); 5.5GHz GDDR5 (PS4)
  • integrated southbridge (SATA, USB, LPC, SPI, etc.)
  • Integrated Radeon-based GPU with possible UVD and HDMI out
Unique features to Xbox ONE SoC:

  • 32 MB eSRAM (102 GB/sec bandwidth total; approx. 50 GB/s per direction)
  • 853 MHz GPU
  • 1.75 GHz CPU
  • 2133MHz DDR3 256-bit memory controller (max 68 GB/sec bandwidth)
  • 12 CU (768 cores) at approx. 1.31 TFLOPS (approximately between a Radeon HD 7750 and 7770, single-precision)
  • DirectX 11.2 supported GPU
  • possible hUMA implementation
Unique features to PS4 SoC:

  • 5.5 GHz GDDR5 256-bit memory controller (max 176 GB/s bandwidth estimated)
  • 18 CU (1152 cores) at approx. 1.83 TFLOPS (approximately between a Radeon HD 7850 and 7870, single-precision)
  • OpenGL 4.4 support
  • hUMA support
We can surmise that the Playstation 4's GPU and CPU components are/were clocked higher than the Xbox ONE's based on two very obvious clues:

  1. Microsoft increased CPU speed to 1.75 GHz
  2. Microsoft increased GPU speed to 853 MHz
It leads me to believe that the PS4 is clocked higher, possibly closer to 2.0 Ghz CPU speed (maximum for Jaguar modules) and that the GPU is probably between 860 MHz (7850) to 1 GHz (7870 GHz Edition). A higher clocked GPU and CPU in the Xbox ONE means it's possibly compensating for lesser hardware performance compared to the PS4.

There is also no possible way that the Xbox ONE's SoC can do 6 operations per clock cycle compared to 4 operations per clock cycle of the PS4's SoC if the Jaguar modules are STATIC in design and structure. One Jaguar module won't do more operations per clock cycle than another unless AMD changed the Jaguar compute unit itself. Its instruction pipeline is still 13 cycles. Customization of the SoC is beyond the modules themselves-- memory controller, southbridge, integrated GPU, etc. And, that is up to Microsoft's and Sony's engineers to do.

AMD is providing the modules, or Jaguar compute units, to OEMs. It's up to the OEMs to design the SoC outside the modules (albeit with help from AMD).

Another thing is: eSRAM is not the end-all, be-all solution here. It's like trying to connect a highway at 80 MPH speed limit to a highway with a 60 MPH speed limit. There will be bottlenecks somewhere along the memory pipeline going from DDR3 to eSRAM to GPU/CPU. It's going to come down to software, again, on how the developers and the Xbox ONE's dev kit will utilize the higher speed, lower latency eSRAM and how to get 3D polygonal, texture, pixel/vertex data to-and-from the GPU and main memory within a 32 MB space. It is also a cache, so data has to be small enough or compressed to be small enough to utilize it and exploit its advantages. D3D tiled resources might help here but I don't see how large textures can be compressed to such a small size without loss of quality.

It's going to take some magical software engineering and programming to compensate for the lower performance of the GPU and the amalgamation of eSRAM, DDR3 RAM, and GPU/CPU to work together in the Xbox ONE.

Also, which handles memory better? Games? Software? Hardware?

Windows 8 kernel (apps, XBLA games, DVR function, cable TV overlay app, etc.) with 3 GB pre-allocated plus Xbox OS (games) with 5 GB pre-allocated on top of a custom Hyper-V implementation (Windows 8/8.1 kernel) or FreeBSD-based OS on top of a security hypervisor (similar to PS3's hypervisor) split between 4.5 GB (games), 3 GB (OS, apps, etc.), and 512 MB "flexible memory".

Regardless how Sony and Microsoft implemented and customized their SoCs, games will look very similar between the two consoles if developers keep the games in a one-to-one parity with the lowest common denominator out of the two-- Xbox ONE. But, the PS4 will be able to and should be able to showcase them better. Also, the PS4 is a more unified, standard PC setup compared to the Xbox ONE that tries to be everything more than a game console. We'll have to wait and see how this pans out in the next few years as the developers get more familiar with the hardware.
 
^ Best explanation on each console's design and system specs I've seen so far. :cool:
 
Damnit Octoberasian i was intentionally not answering to make his lazy ass to finally farking read up <.<

Still you are awesome dude.

BTW the current massive issue with 3rd parties is that MS forgot to include an easy way to manage the eSRAM in the dev tools, in the 360 this was done transparent to the dev because the eDRAM was used directly as a frame buffer, but on the one it is kinda programmable, thing is if you don't fine tune the useage you get the dismal performance that developers have reported with less than half of the total PS4's performance.

The numbers on neoGAF come from certified industry insiders and developers who can sorta bypass the NDA by keeping their real persona a secret to the world (secret that they have had to share with a Moderator of the neoGAF forum in order to not get perma banned). > this part i have explained so many times in this thread yet Vlad hasn't grasped it.
 
"Microsoft claimed that each CPU core can perform six operations per cycle. The CPU is believed to be using AMD's Jaguar core, but typically this would only be described as able to handle four operations per cycle; two each of integer and floating point (though even here counting operations is complicated; the floating point operations could use vector instructions such as SSE2, in which case one operation would result in four actual computations, potentially giving eight per cycle for floating point alone)."

NeoGaf itself has not found conclusive evidence that the PS4 can handle the same amount of operations per second. Yeah it may be the same base core, but that doesnt mean each company didnt dedicate a certain amount of time an resources to AMD to customize the chip further
 
Revdarian said:
Damnit Octoberasian i was intentionally not answering to make his lazy ass to finally farking read up <.<

Still you are awesome dude.

BTW the current massive issue with 3rd parties is that MS forgot to include an easy way to manage the eSRAM in the dev tools, in the 360 this was done transparent to the dev because the eDRAM was used directly as a frame buffer, but on the one it is kinda programmable, thing is if you don't fine tune the useage you get the dismal performance that developers have reported with less than half of the total PS4's performance.

The numbers on neoGAF come from certified industry insiders and developers who can sorta bypass the NDA by keeping their real persona a secret to the world (secret that they have had to share with a Moderator of the neoGAF forum in order to not get perma banned). > this part i have explained so many times in this thread yet Vlad hasn't grasped it.
Click to expand...

Being anonymous DOES NOT allow you to legally bypass an NDA. NDA's state that you will not disclose information under any circumstances or surname, which includes anonymous citation
 
Tell that to the fellas at neoGAF. Sure if MS gets to determine who did the leak in a certain way they could get some legal action going, but so far, they haven't done anything.
 
About the Operations number, if you take notice, they talk all the time in that thread PS4/Xbone... meaning that they are on the same boat in that regard... Even Penello only painted a ~10% increase in cpu power for the one in his first comparison.
 
Thanks for making me read, i knew that your assesment was wrong Vlad, here, from the original thread:

http://forum.beyond3d.com/showthread.php?t=62867&page=133

"A Jaguar core can issue 6 ops per cycle -- two memory, two int alu, and two fpu/simd ops. However, it can only decode two, and retire two, so the average rate cannot ever go above two.

The primary reason for the 6-wide issue is that between decode and retire the separate pipelines are effectively independent and there's no reason they shouldn't be able to issue in parallel. However, being able to issue more than you can sustain in the long term does have some advantages, notably, a bad data stall will cause your execution to stall until the data arrives, while the decode parts can happily run forward until all the queues and buffers are full. When the data arrives, being able to issue more ops means you can clear these queues, and the retire can again catch up the next time you stall.

So basically, having more issue resources means that you can get closer to the max of 2 in practice.

Of course, picking the number 6 in particular is just the kind of "xbox 360 gpu has x TFlops" we all know and love from the past gen. There are many numbers that made varying amount of sense, this one got picked because it's the biggest one."

So, the PS4 has the exact same limit of 6op/cycle, and again, that is why the difference is of roughly the overclock "if the PS4 ships at the 1.6 GHz target".
 
octoberasian said:
How would this be possible if both processors are using the same Jaguar-based modules?

A single Jaguar module is:

  • 1.0 GHz to 2.0 GHz max
  • 4 cores
  • 2 MB shared L2 cache - "shared cache unit" (SCU)
  • called a "compute unit"
  • x86-64 processor
  • 13 cycle instruction pipeline

Both the Xbox ONE and PS4 are SoCs (system-on-a-chip):

  • 2x Jaguar modules (8 cores total)
  • integrated memory controller-- 2133MHz DDR3 (Xbox ONE); 5.5GHz GDDR5 (PS4)
  • integrated southbridge (SATA, USB, LPC, SPI, etc.)
  • Integrated Radeon-based GPU with possible UVD and HDMI out
Unique features to Xbox ONE SoC:

  • 32 MB eSRAM (102 GB/sec bandwidth total; approx. 50 GB/s per direction)
  • 853 MHz GPU
  • 1.75 GHz CPU
  • 2133MHz DDR3 256-bit memory controller (max 68 GB/sec bandwidth)
  • 12 CU (768 cores) at approx. 1.31 TFLOPS (approximately between a Radeon HD 7750 and 7770, single-precision)
  • DirectX 11.2 supported GPU
  • possible hUMA implementation
Unique features to PS4 SoC:

  • 5.5 GHz GDDR5 256-bit memory controller (max 176 GB/s bandwidth estimated)
  • 18 CU (1152 cores) at approx. 1.83 TFLOPS (approximately between a Radeon HD 7850 and 7870, single-precision)
  • OpenGL 4.4 support
  • hUMA support
We can surmise that the Playstation 4's GPU and CPU components are/were clocked higher than the Xbox ONE's based on two very obvious clues:

  1. Microsoft increased CPU speed to 1.75 GHz
  2. Microsoft increased GPU speed to 853 MHz
It leads me to believe that the PS4 is clocked higher, possibly closer to 2.0 Ghz CPU speed (maximum for Jaguar modules) and that the GPU is probably between 860 MHz (7850) to 1 GHz (7870 GHz Edition). A higher clocked GPU and CPU in the Xbox ONE means it's possibly compensating for lesser hardware performance compared to the PS4.

There is also no possible way that the Xbox ONE's SoC can do 6 operations per clock cycle compared to 4 operations per clock cycle of the PS4's SoC if the Jaguar modules are STATIC in design and structure. One Jaguar module won't do more operations per clock cycle than another unless AMD changed the Jaguar compute unit itself. Its instruction pipeline is still 13 cycles. Customization of the SoC is beyond the modules themselves-- memory controller, southbridge, integrated GPU, etc. And, that is up to Microsoft's and Sony's engineers to do.

AMD is providing the modules, or Jaguar compute units, to OEMs. It's up to the OEMs to design the SoC outside the modules (albeit with help from AMD).

Another thing is: eSRAM is not the end-all, be-all solution here. It's like trying to connect a highway at 80 MPH speed limit to a highway with a 60 MPH speed limit. There will be bottlenecks somewhere along the memory pipeline going from DDR3 to eSRAM to GPU/CPU. It's going to come down to software, again, on how the developers and the Xbox ONE's dev kit will utilize the higher speed, lower latency eSRAM and how to get 3D polygonal, texture, pixel/vertex data to-and-from the GPU and main memory within a 32 MB space. It is also a cache, so data has to be small enough or compressed to be small enough to utilize it and exploit its advantages. D3D tiled resources might help here but I don't see how large textures can be compressed to such a small size without loss of quality.

It's going to take some magical software engineering and programming to compensate for the lower performance of the GPU and the amalgamation of eSRAM, DDR3 RAM, and GPU/CPU to work together in the Xbox ONE.

Also, which handles memory better? Games? Software? Hardware?

Windows 8 kernel (apps, XBLA games, DVR function, cable TV overlay app, etc.) with 3 GB pre-allocated plus Xbox OS (games) with 5 GB pre-allocated on top of a custom Hyper-V implementation (Windows 8/8.1 kernel) or FreeBSD-based OS on top of a security hypervisor (similar to PS3's hypervisor) split between 4.5 GB (games), 3 GB (OS, apps, etc.), and 512 MB "flexible memory".

Regardless how Sony and Microsoft implemented and customized their SoCs, games will look very similar between the two consoles if developers keep the games in a one-to-one parity with the lowest common denominator out of the two-- Xbox ONE. But, the PS4 will be able to and should be able to showcase them better. Also, the PS4 is a more unified, standard PC setup compared to the Xbox ONE that tries to be everything more than a game console. We'll have to wait and see how this pans out in the next few years as the developers get more familiar with the hardware.
Click to expand...

This post is completely full of assumptions. Your ASSUMING that Microsoft knows what the CPU Speed of the PS4 is going to be, yet it has not been released as public knowledge, which i find hilarious because every single sony supporter called absolute foul when microsoft claimed a 10% CPU speed advantage because Microsoft couldnt possibly know the PS4 CPU specs. Completely contradictory.

As for the post on the Ops, that post was made in May, which is before Microsoft went into ANY details on the APU architecture, well before the press release, and well before the HotChips conference which does not invalidate what Microsoft said. I have yet to see someone disprove it with evidence coming out AFTER Microsoft made the announcement. Yes the timing of released information and information that contradicts it does matter.

As for NeoGaf, if this were any other topic, such as fox news on any political argument, you would cry foul without knowing exactly who your sources are. Neogaf does not differentiate insiders, meaning a person with insider status may still strictly be a Sony employeed person, or first party dev. It does not mean they have direct experience with both platforms.

As for the legality issue, your making the assumption that these insiders are releasing information that is held secret just because they hold insider status, and there is no guarantee of that, just because someone is listed as an insider DOES NOT mean they are giving you secret information.

Your also misquoting the fact that the PS4 is hUMA which it is not, because one defining point of hUMA according to AMD is full cache coherency between CPU and GPU and current information is saying PS4s caches are only coherent in a single direction which does not make it truely hUMA compliant.
 
Revdarian said:
-.-'

And here i was gonna give you some benefit of doubt... dividing by 0 yet again.
Click to expand...

why because im simply stating facts? show me where on neogaf it tells us an insiders credentials, show me where on neogaf someone is releasing clearly secret information. show me where in the ps4 information it supports it being true hUMA
 
If anything said on neogaf werent true, then just as there are ps4 only insiders, there are mixed ones and MS only ones, and they would call foul.

That is the basics of the basics, but your blatant bias doesn't allow you to see that.

Denial, denial, denial denial.
 
Revdarian said:
If anything said on neogaf werent true, then just as there are ps4 only insiders, there are mixed ones and MS only ones, and they would call foul.

That is the basics of the basics, but your blatant bias doesn't allow you to see that.

Denial, denial, denial denial.
Click to expand...

not necessarily true and you cant prove that without neogaf actually telling us who these insiders are, thats also the basic of basics, your making an assumption, and thats the simple truth, and even so, again show me where actual secret information is being released as opposed to clarification of previous statements
 
Vlad, start giving numbers to prove yourself. Come on you're a hardware/systems engineer right? I'm an engineer myself (albeit not in the electronic business) and would love to see some numbers coming from you.
 
I cant give numbers when benchmarks on anything by component level has been done, and thats what i have been stating this whole time.

I cannot claim the XBox One has any direct advantages without having direct access to their hardware. And the same is completely true for the sony side of things.

Just because i can look at a memory speed does not mean i know everything i need to know about it to draw a real legitimate conclusion. I can look at the number of CU's sure, but i cant see where they are actually embedded into the system, in fact theres talks that perhaps only 14 of the 18 in the ps4 are for gaming. can i prove it? no. can i prove that all 18 also are? no

Can i prove how much the eSRAM will add to the ones performance? no.

every single conclusion that anyone is drawing is strictly on speculation, and if you cant realize that then there is absolutely no convincing you of otherwise
 
It is totally obvious that neoGAF is full of lies, i mean, why other reason would Albert Penello try and get an account there?

Anyone who reads this thread will know in an objective manner the truth, and the obvious bias.
 
Revdarian said:
It is totally obvious that neoGAF is full of lies, i mean, why other reason would Albert Penello try and get an account there?

Anyone who reads this thread will know in an objective manner the truth, and the obvious bias.
Click to expand...

And yet every single reply back to me is done in an incredibly biased fashion. Your taking theoretical numbers on the PS4 and NOT comparing them to theoretical numbers on the XBox One, you cant be objective and biased at the same time.

I have stated absolutely no bias, i have stated the fact that YOU CANNOT make comparisons on hardware or performance without really knowing anything about it. You cannot site a source without knowing who the source really is. Just because someone is an insider does not mean they are giving you secret information, and it does not mean that that insider does not have a bias of their own.
 
VladDracule said:
And yet every single reply back to me is done in an incredibly biased fashion. Your taking theoretical numbers on the PS4 and NOT comparing them to theoretical numbers on the XBox One, you cant be objective and biased at the same time.

I have stated absolutely no bias, i have stated the fact that YOU CANNOT make comparisons on hardware or performance without really knowing anything about it. You cannot site a source without knowing who the source really is. Just because someone is an insider does not mean they are giving you secret information, and it does not mean that that insider does not have a bias of their own.
Click to expand...

Actually you can make predictions on performance based on hardware information. For example people know DDR5 and the bandwidth advantage over DDR3 allows for significant advantages when rendering high detail 3D scenes. Scenes are rendered with textures residing in that RAM and that data needs to pass along the memory controller. Also the same people understand that AMD knows how to feed data into a GPU, especially when paired with something like a Radeon GPU.
 
VladDracule said:
This post is completely full of assumptions. Your ASSUMING that Microsoft knows what the CPU Speed of the PS4 is going to be, yet it has not been released as public knowledge, which i find hilarious because every single sony supporter called absolute foul when microsoft claimed a 10% CPU speed advantage because Microsoft couldnt possibly know the PS4 CPU specs. Completely contradictory.
Click to expand...

He never said anything about any of this in his post.
You're starting to pull things out of thin air and are splitting hairs at this point.

Most of what you have been posting at this point has just been your opinion or just fud in general, sorry to say.
 
Red Falcon said:
He never said anything about any of this in his post.
You're starting to pull things out of thin air and are splitting hairs at this point.

Most of what you have been posting at this point has just been your opinion or just fud in general, sorry to say.
Click to expand...

completely wrong. he said the PS4 supports hUMA, he also tried to enstantiate an argument that the original clock speed on the CPU and GPU were higher than the XBox Ones,read it again
 
"Unique features to PS4 SoC:
hUMA support"

"We can surmise that the Playstation 4's GPU and CPU components are/were clocked higher than the Xbox ONE's based on two very obvious clues:"

right there
 
"32 MB eSRAM (102 GB/sec bandwidth total; approx. 50 GB/s per direction)" - from octoberasian posts

"Peak bandwidth is indicated to be 204 GB/s, with minimum bandwidth of 109 GB/s." - From microsofts hotchips conference.

which makes that also a completely incorrect citation
 
ocellaris said:
Actually you can make predictions on performance based on hardware information. For example people know DDR5 and the bandwidth advantage over DDR3 allows for significant advantages when rendering high detail 3D scenes.
Click to expand...
OK so give us a prediction on what effect the xbone's 32MB eSRAM will have on effective system bandwidth and/or total system performance. Bear in mind that the X360 used a similar memory model for years and had great performance vs the PS3 which had, according to some (ie. Sony), a huge performance advantage. Which was certainly true on paper in some ways but didn't really pan out in the real world did it?

ocellaris said:
Also the same people understand that AMD knows how to feed data into a GPU, especially when paired with something like a Radeon GPU.
Click to expand...
If you trust that AMD knows how to feed its own GPU's data than why is it so hard to believe that the 2 consoles will actually perform about the same?
 
VladDracule said:
"32 MB eSRAM (102 GB/sec bandwidth total; approx. 50 GB/s per direction)" - from octoberasian posts

"Peak bandwidth is indicated to be 204 GB/s, with minimum bandwidth of 109 GB/s." - From microsofts hotchips conference.

which makes that also a completely incorrect citation
Click to expand...

Vgleaks has a wealth of info, likely supplied from game developers with direct access to Xbox One specs, that looks to be very accurate at this point. According to their data, there’s roughly 50GB/s of bandwidth in each direction to the SoC’s embedded SRAM (102GB/s total bandwidth). The combination of the two plus the CPU-GPU connection at 30GB/s is how Microsoft arrives at its 200GB/s bandwidth figure, although in reality that’s not how any of this works.
Click to expand...
http://www.anandtech.com/show/6972/xbox-one-hardware-compared-to-playstation-4/3

You don't magically get 200+ GB/s bandwidth from a combination of 68 GB/s DDR3 RAM and 102 GB/s from the eSRAM and then another 30 GB/s from CPU-GPU. You have two different types of memory here and different data buses. You have CPU <-> RAM via northbridge. You have CPU <-> GPU via northbridge. And, you have GPU <-> eSRAM and GPU <-> DDR3 RAM via the GPU's own memory bus. Adding all of that, of course, gets you 200+ GB/s of bandwidth. However, you're talking about different components on a system board passing and moving data along.

Like the example I gave above, it's like an interstate highway. You have a car going at the max speed limit of 100 MPH (eSRAM) entering the highway interchange to get onto a highway with a max speed limit of 60 MPH (DDR3 RAM). Developers have to get around this to ensure there are no perceived bottlenecks at any location of the data bus. And, hopefully, D3D tiled resources (DX 11.2) and other software tricks and compression algorithms will get around that limitation.

There will be bottlenecks and it's going to depend entirely on developers on getting around that. You can't magically shove 300 MB of texture data into 32 MB of eSRAM unless you have some very good compression algorithms, and you can't turn 68 GB/s of DDR3 RAM into 200 GB/s of memory bandwidth when it's not there.

But, you can't deny the fact that there is a slight processing advantage of the GPU in the PS4 compared to the Xbox ONE. Numbers are numbers. We know this from hardware reviews even those on [H]ardOCP and other review sites. You can't magically expect a game like Metro Last Light to perform the same on a Radeon 7850 and then on a Radeon 7770 at the same settings and resolution. If you were to take that game and set it to 1920x1080, AAA (Very High settings), and tessellation set to Very High, I assure you that the Radeon 7770 will not do it at a high enough framerate at 1080p. The 7870 GHz Edition already averages 42 FPS under those settings in [H]ardOCP's testing. Other reviews peg the 7770 at around 10 FPS less on those same settings, and higher settings dropping the framerate to just below 30 FPS.

Concessions and sacrifices in quality settings have to be made, as well as resolution at which the game is rendered. In the PS3 and the 360, this was done at smaller resolutions then upscaled. 540p? 768p? ... then expanded or contorted to 720p or 1080p. I would not be surprised developers will do this again on the Xbox ONE. We'll most likely see the first batch of games on the system at 720p, 60 FPS average and upscaled to 1080p. And, more than likely, PS4 ports will be the same for the first year or so. Thus, taking away any clear advantage the PS4 has in terms of GPU hardware and performance. We'll probably see very few 1080p titles but not at 60FPS, more than likely at 30FPS in the first year, and I would not be surprised some will be rendered at a lower resolution such as 960x540 then upscaled to 1920x1080.

Again, it's going to take some sort of software engineering and programming feat to get around the various components of the Xbox ONE. The games that do so will be the most noticeable to players.

And, any of us that have read reviews of AMD's APU since Llano know that memory bandwidth is the biggest issue, if not so. This is why the highest end Haswell mobile processor has eDRAM on-die at up to 128 MB of it. DDR3 RAM cannot provide the necessary bandwidth for high resolution gaming with many of the settings turned on. Again, hardware reviews between a video card with DDR RAM and GDDR RAM show the differences. It is also why for a more straightforward approach, Sony most likely went with the higher bandwidth GDDR5 over DDR3 regardless of the latency. More than likely hoping bandwidth usurps latency, which even I would know will not likely happen. Microsoft gets around this with various ICs called move engines, if you believe VG Leaks at all and the Digital Foundry articles posted after. They're less of a co-processor and more of a hard-coded PHY memory interfaces to get around the bandwidth limitations of the console to receive/decode data going from one memory type to another. Why? Maybe, again assumption, to prevent perceived bottlenecks within the system.

But, when you think about it-- the hardware setup in the PS4 is simpler compared to the Xbox ONE. It's more straightforward and looks almost as similar as various budget AMD APU-based computers-- APU, RAM, hard drive, etc. It's more of a console-ized PC dedicated to gaming and media services running a Linux-distro.

As for hUMA, that's entirely assumption as you said so yourself. But, given that AMD licensed the modular Jaguar modules to OEMs in a similar fashion to ARM licensing their designs, there is still a small, very small, possibility that Microsoft and Sony added hUMA support outside the modules themselves. Regardless if either system supports it or not is something only those with a dev kit and the companies themselves-- AMD, Sony and Microsoft-- will know for sure. hUMA support in the consoles would, in my opinion, be a great benefit for AMD and for us on desktop PCs. It'd be a great way to showcase and demonstrate what HSA and hUMA can do. Unfortunately, there's no way to confirm if either console supports it or not. If they do support it, developers will have to program games to specifically use it.
 
Semantics said:
But yeah it still doesn't take a ton of work but when people bitch about the probable 10-15% real difference we'll see in games why not bitch about the 2-5% of cpu cycles taken up by audio.
Click to expand...
On some X360 games audio apparently chewed up an entire core by itself according to developers who post on B3D. That is lots more than 2-5% of CPU cycles.
 
mesyn191 said:
OK so give us a prediction on what effect the xbone's 32MB eSRAM will have on effective system bandwidth and/or total system performance. Bear in mind that the X360 used a similar memory model for years and had great performance vs the PS3 which had, according to some (ie. Sony), a huge performance advantage. Which was certainly true on paper in some ways but didn't really pan out in the real world did it?


If you trust that AMD knows how to feed its own GPU's data than why is it so hard to believe that the 2 consoles will actually perform about the same?
Click to expand...

The 360 advantage was in the GPU, not the CPU. Considering titles like TLoU, God of War Ascension, Uncharted 3, and Beyond: Two Souls are arguably more advanced than 360 titles I'm not entirely sure that is a sustainable argument to be had and likely makes the Xbox One look worse from a historic perspective.
 
To continue,

Considering the 360 had the same type of setup along with a more powerful GPU and can't unequivocally beat the PS3 in even exclusive titles and lately third party titles how can one argue that the One could sustain any advantage when it's GPU is without question inferior?
 
360 was easier to code for, with a better gpu.
PS3 was a bitch to code for, with a worse gpu but much more powerful cpu relatively speaking, you had to use the cpu to balance the gpu in certain process and achieve parity or in the case of Naughty Dog, superiority.

Thing is, at the moment:
Xb1 is harder to code, with a weaker gpu.
PS4 is easier to code with a stronger gpu.

And the cpu difference is not big enough to help balance the gpu difference, hence the rumor is that parity will be most of the time in the form of optimized 900p vs 1080p.
 
next-Jin said:
The 360 advantage was in the GPU, not the CPU.
Click to expand...
This doesn't answer my question(s) or even make much sense really in the context of ocellaris's statement that performance is predictable.

next-Jin said:
To continue,

Considering the 360 had the same type of setup along with a more powerful GPU and can't unequivocally beat the PS3 in even exclusive titles and lately third party titles how can one argue that the One could sustain any advantage when it's GPU is without question inferior?
Click to expand...
I never argued xb1 would have an advantage over PS4. I've said they'll perform about the same.
 
Revdarian said:
Thing is, at the moment:
Xb1 is harder to code, with a weaker gpu.
Click to expand...
Thing is developers already have a lot of experience with working with a memory model similar to, but over all more difficult to use, the xb1's.

Revdarian said:
PS4 is easier to code
Click to expand...
Depends on what you're doing.
Everyone seems to be ignoring that the XB1 has a GPU with effectively 32MB of cache, compared to the PS4 in the range of 512k. So yes, as long as what they are processing in a frame is purely streaming, congruent data, then the PS4 will easily surpass the XB1. Make the data a bunch of different textures in different memory locations, or GPGPU physics calculations, or complex shaders that aren't just streaming data, then the result may surprise you. In those cases, the PS4 may take up to 10x longer to retrieve a piece of data than the XB1, stalling a non trivial amount of the GPU.
Click to expand...
 
mesyn191 said:
Thing is developers already have a lot of experience with working with a memory model similar to, but over all more difficult to use, the xb1's.


Depends on what you're doing.
Click to expand...

Not to mention ray tracing will be faster on the XBox one and that notion has been supported by a Dev.

People are doing nothing but looking at high level numbers, and thinking they can draw conclusions like "the ps4 is clearly 50% faster". In terms of actual memory bandwidth? yeah maybe, will it be realizable in practice? probably not.

Individual components, especially the speed of memory does not determine an overall systems performance.
 
mesyn191 said:
This doesn't answer my question(s) or even make much sense really in the context of ocellaris's statement that performance is predictable.


I never argued xb1 would have an advantage over PS4. I've said they'll perform about the same.
Click to expand...

It makes sense given my statement, it's predictable given overall design and implementation. It's effectively the same thing they did before taken up a notch. In which case it didn't help the platform at all.

It did not lead to better looking games and it had a superior GPU. What does that say now that they used a similar approach with a much weaker GPU?

I'm not saying one is going to be 1080p/60 while the One is 720p/30, but I don't think anyone would suggest parity between the two.
 
You must log in or register to reply here.
Back
Top