• 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"

JonnyLH

Banned
He talks about audio raycasting on the GPU IIRC, something that SHAPE doesn't do.
Raycasting is the rendering of a 3d image from a 2d texture surface?

Like he says, the post-processing and big audio tasks will be performed by the GPU which will provide a "hybrid" approach.
 

ElTorro

I wanted to dominate the living room. Then I took an ESRAM in the knee.
http://www.vgleaks.com/world-exclusive-orbis-unveiled-2/

More information inside that article. The PS4's 4 'hardware balanced' CU's are extra ALU's. These are primarily for mathematical and logical operations. This is very smart actually, these will be quite different from the other CU's as they're balanced for this operation.

The article literally just says "balanced at 14+4", which might indicate that this schedule of GPU time is a reasonable one for many games (see quote below). Everything else you said is just made up.

Digital Foundry: Going back to GPU compute for a moment, I wouldn't call it a rumour - it was more than that. There was a recommendation - a suggestion? - for 14 cores [GPU compute units] allocated to visuals and four to GPU compute...

Mark Cerny: That comes from a leak and is not any form of formal evangelisation. The point is the hardware is intentionally not 100 per cent round. It has a little bit more ALU in it than it would if you were thinking strictly about graphics. As a result of that you have an opportunity, you could say an incentivisation, to use that ALU for GPGPU.

(CUs are basically just arrays of ALUs).

http://www.eurogamer.net/articles/digitalfoundry-face-to-face-with-mark-cerny
 

KidBeta

Junior Member
Raycasting is the rendering of a 3d image from a 2d texture surface?

Like he says, the post-processing and big audio tasks will be performed by the GPU which will provide a "hybrid" approach.

ray casting is the name for any algorithm that fires rays and tries to model where they go and how they interact with the environment. It doesn't have to have anything to do with graphics.

this GPGPU hybrid approach I imagine is the same as with the XBONE. Theres a lot of stuff SHAPE cannot do.
 
Every modern AMD GPU includes FF hardware to encode and decode h264 video on the fly without using any of the CU's.

Also the GDDR5 latency rumour has been debunked multiple times, the latency is the close enough if the not the same as DDR3.

In the GPU its a non-issue. In the CPU its a completely different story. It does have massive effects where latency dependent tasks like an audio buffer, come to suffer.

I'm not trying to degrade the PS4 in anyway, I'm just stating that the extra beef in the GPU was a smart move and will be used for tasks the CPU will come to struggle with.

Actually the latency of the DDR3 chips MS picked could be equal or even worse.

Identified X1 memory chips are Micron "2133 Mhz CAS 14" which is not exactly high end modules. Translation in real world latency :

(14 / 1066) x 1000 = 13.1 ns (it's in their .pdf also)

Some GDDR5 chips have better latency than this. Even with a CL 20 it could be close, cause the base clock is higher (1375).
Btw, the X360 never had struggle on the cpu side with GDDR3.

I think either way it'snot really a concern, knowing what kind of stuff is going to run on the cpu, and on the contrary what's going to run on the gpu (almost everything)
 

JonnyLH

Banned
The article literally just says "balanced at 14+4", which might indicate that this schedule of GPU time is a reasonable one for many games (see quote below). Everything else you said is just made up.



http://www.eurogamer.net/articles/digitalfoundry-face-to-face-with-mark-cerny
He says its not formal and confirms its true by saying:
It has a little bit more ALU in it than it would if you were thinking strictly about graphics
Like the article says, these can not be used for this purpose but they're there. It was a sneaky way of getting round the question which the non-technical person would understand.
 

ElTorro

I wanted to dominate the living room. Then I took an ESRAM in the knee.

KidBeta

Junior Member
Actually the latency of the DDR3 chips MS picked could be equal or even worse.

Identified X1 memory chips are Micron "2133 Mhz CAS 14" which is not exactly high end modules. Translation in real world latency :

(14 / 1066) x 1000 = 13.1 ns (it's in their .pdf also)

Some GDDR5 chips have better latency than this. Even with a CL 20 it could be close, cause the base clock is higher (1375).
Btw, the X360 never had struggle on the cpu side with GDDR3.

I think either way it'snot really a concern, knowing what kind of stuff is going to run on the cpu, and on the contrary what's going to run on the gpu (almost everything)

Indeed, not to mention the entire point of the cache on a CPU is so that you don't have to goto the memory often, making the latency even less worrying.
 

Jack_AG

Banned
Its a common conception that some CU's will be reserved for OS tasks. For example, what do you think decodes your screen and encodes it into a 15 minute video for you to share?

I hate to bring this topic subject up again, but CPU's are hindered by GDDR memory. Tasks like audio processing, which would normally get handled by the CPU will get offloaded into the GPU. Same goes with physics. GDDR latency isn't as apparent in the GPU.

In conclusion, for actual CU's which are 'dedicated' to games so to speak, it may very well decrease.
My response was to the "all you need" comment. What is allowed/locked is an entirely different discussion.
 

ElTorro

I wanted to dominate the living room. Then I took an ESRAM in the knee.
Like the article says, these can not be used for this purpose but they're there. It was a sneaky way of getting round the question which the non-technical person would understand.

What do you think is the difference between "More ALU" and "just 4 more compute units"? There is no hint at any difference or special sauce in 4 CUs. There are just 18 compute units, and every developer can use them as he pleases. Even the architectural overview of the PS4's GPUs shows no difference but talks about a "unified array of compute units".

gpu_queues.jpg
 
Raycasting is the rendering of a 3d image from a 2d texture surface?

Like he says, the post-processing and big audio tasks will be performed by the GPU which will provide a "hybrid" approach.

Another thing the PlayStation 4 team did to increase the flexibility of the console is to put many of its basic functions on dedicated units on the board -- that way, you don't have to allocate resources to handling these things.

"The reason we use dedicated units is it means the overhead as far as games are concerned is very low," said Cerny. "It also establishes a baseline that we can use in our user experience."

"For example, by having the hardware dedicated unit for audio, that means we can support audio chat without the games needing to dedicate any significant resources to them. The same thing for compression and decompression of video." The audio unit also handles decompression of "a very large number" of MP3 streams for in-game audio, Cerny added.
http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php?page=3
no?

The GPU can do different types of audio processing but the PS4 has an audio DSP. But then the PS4 has 8 ACEs and 64 compute queues compared to Xbox One 2 ACE 4 compute queues so it's hardly going to be a meaningful performance hit.
 

JonnyLH

Banned
What do you think is the difference between "More ALU" and "just 4 more compute units"? There is no hint at any difference or special sauce in 4 CUs. There are just 18 compute units, and every developer can use them as he pleases. Even the architectural overview of the PS4's GPUs shows no difference but talks about a "unified array of compute units".

IF VGLeaks are correct, they're hardware balanced to support the ALU operation. That's different. Yet, it depends on their leaks. Which they have been correct so far. Even Mark Cerny sort-of confirms it in his interview.
 

KidBeta

Junior Member
IF VGLeaks are correct, they're hardware balanced to support the ALU operation. That's different. Yet, it depends on their leaks. Which they have been correct so far. Even Mark Cerny sort-of confirms it in his interview.

Whats hardware balanced mean, what hardware has changed, where does it sit within the GPU, and what does it change.
 

ElTorro

I wanted to dominate the living room. Then I took an ESRAM in the knee.
IF VGLeaks are correct, they're hardware balanced to support the ALU operation. That's different. Yet, it depends on their leaks. Which they have been correct so far. Even Mark Cerny sort-of confirms it in his interview.

What does "balanced to support the ALU operation" even mean? I don't think that you really know what you are talking about when you refer to ALU.

By the way, I love how we are bouncing from issue to issue without any conclusion after every argument gets debunked. First, we are talking about a OS reserve of GPU time on the PS4. Without addressing the counter arguments, you switch the debate to the PS4 "sorely" relying on the GPU. Without addressing the arguments here, you switch to a debate that 4 CUs are supposedly different, and you dig out the latency argument that was already debunked a few posts ago.

That does not bring any debate forward.
 
Nope, offloading in-game chat is fine but then he refers to just the encoding and decoding of over 200 streams. That doesn't mean processing or any of the more labour orientated tasks which comes with audio, especially next-gen.

PS4 has 8 ACE's and 64 compute queues, Xbox One has 2 ACE's and 4 compute queues. Pointless having all that processing power and not using it but that's hardly "sorely relying on GPU". Not only that but using compute means they can take advantage of advanced effects and techniques as they become available, doesn't an Audio DSP have fixed function?
 

JonnyLH

Banned
Whats hardware balanced mean, what hardware has changed, where does it sit within the GPU, and what does it change.

What does "balanced to support the ALU operation" even mean? I don't think that you really know what you are talking about when you refer to ALU.

By the way, I love how we are bouncing from issue to issue without any conclusion after every argument gets debunked. First, we are talking about a OS reserve of GPU time on the PS4. Without addressing the counter arguments, you switch the debate to the PS4 "sorely" relying on the GPU. Without addressing the arguments here, you switch to a debate that 4 CUs are supposedly different, and you dig out the latency argument that was already debunked a few posts ago.

That does not bring any debate forward.
Its hard to say without seeing a detailed spec sheet. This could be more cache, less texture units. It literally could mean anything, I've personally not really looked into the subject before. What VGLeaks implies is that there is something different architectural about those 4 cores which means they're for the ALU purpose.

EDIT: The whole point of this discussion is whether these 18 cores will be dedicated to the physical rendering of the game, everything graphical. Personally, I sorely don't believe they will be. Hence why Sony opted for the larger GPU. It wasn't just "Rawr more powah".
 

Oppo

Member
They are, did you know most games on the 360 used to dedicate a full PowerPC core to processing audio, because the chip only provided so much?

Most, no. A few used a whole core. Of that old PPC chip.

Ah you're just kicking sand up.
 

KidBeta

Junior Member
Its hard to say without seeing a detailed spec sheet. This could be more cache, less texture units. It literally could mean anything, I've personally not really looked into the subject before. What VGLeaks implies is that there is something different architectural about those 4 cores which means they're for the ALU purpose.

EDIT: The whole point of this discussion is whether these 18 cores will be dedicated to the physical rendering of the game, everything graphical. Personally, I sorely don't believe they will be. Hence why Sony opted for the larger GPU. It wasn't just "Rawr more powah".

But the rest of the vgleaks numbers and data lines up with all 18 CU's being able to do graphic tasks (i.e. it has 18 texture units). The cache argument doesn't make any sense either as even GPGPU tasks need cache and it would be retarded to remove something like that, costing time and money, to remove flexibility.
 
EDIT: The whole point of this discussion is whether these 18 cores will be dedicated to the physical rendering of the game, everything graphical. Personally, I sorely don't believe they will be. Hence why Sony opted for the larger GPU. It wasn't just "Rawr more powah".

Cerny already said that ALL of the CUs have beefed up ALU and that they have been modified such that it can process graphics and compute in parallel whereby a modern PC GPU can only execute compute or graphics at a time.
 

JonnyLH

Banned
You're the only one seeing that implication, bro.
By looking at the article on VGLeaks, its very ambiguous. GPU's are very good maths calculators, and dedicating 4 CU's to the ALU purpose is a very smart idea since it will work much faster than the ALU's in the CPU. If you're denying this, then the PS4 will be much a slower machine. I'd rather have faster ALU's then 4 extra CU's.
Most, no. A few used a whole core. Of that old PPC chip.

Ah you're just kicking sand up.
Nope, I'm making a relevant statement to say, audio takes a lot of power. Especially this generation. A chip which's primary purpose is to encode streams and work in-game chat doesn't have the "umph" to handle next-generation audio.
 

KidBeta

Junior Member
By looking at the article on VGLeaks, its very ambiguous. GPU's are very good maths calculators, and dedicating 4 CU's to the ALU purpose is a very smart idea since it will work much faster than the ALU's in the CPU. If you're denying this, then the PS4 will be much a slower machine. I'd rather have faster ALU's then 4 extra CU's.

Nope, I'm making a relevant statement to say, audio takes a lot of power. Especially this generation. A chip which's primary purpose is to encode streams and work in-game chat doesn't have the "umph" to handle next-generation audio.

Oh okay, your even trolling or missing a fundamental understanding of what a GPU does.

CU's are ALU's...
 
By looking at the article on VGLeaks, its very ambiguous. GPU's are very good maths calculators, and dedicating 4 CU's to the ALU purpose is a very smart idea since it will work much faster than the ALU's in the CPU. If you're denying this, then the PS4 will be much a slower machine. I'd rather have faster ALU's then 4 extra CU's.

Nope, I'm making a relevant statement to say, audio takes a lot of power. Especially this generation. A chip which's primary purpose is to encode streams and work in-game chat doesn't have the "umph" to handle next-generation audio.

Are you ignoring my posts on purpose?

The Xbox One only has 2 ACE's with 4 compute queues and a fixed function Audio DSP, the PS4 has 8 ACE's with 64 compute queues which can execute compute and graphics threads in parallel as well as a basic Audio DSP.

It has more than enough compute "umph" compared to XBone.
 

JonnyLH

Banned
Cerny already said that ALL of the CUs have beefed up ALU and that they have been modified such that it can process graphics and compute in parallel whereby a modern PC GPU can only execute compute or graphics at a time.
Which would be fine, but if you're using a CU for ALU, you can't make it do anything else. If the machine is hardware balanced at 14, this means you have to specify to use those 4. Just you then have to make sure everything is fine regarding the CPU.
 

KidBeta

Junior Member
Which would be fine, but if you're using a CU for ALU, you can't make it do anything else. If the machine is hardware balanced at 14, this means you have to specify to use those 4. Just you then have to make sure everything is fine regarding the CPU.

With GPGPU you don't have to have anything fixed like that, you can use 18 for rendering, 2 for rendering, or 0 for rendering.

You just miss understand how it works.
 
Which would be fine, but if you're using a CU for ALU, you can't make it do anything else. If the machine is hardware balanced at 14, this means you have to specify to use those 4. Just you then have to make sure everything is fine regarding the CPU.

Ugh yes you can, they have modified the GPU so that it can execute BOTH graphics and compute threads in PARALLEL unlike a modern PC GPU that can only execute one at a time and has to wait for the other to complete. That's not just 4 of the CU's, that's ALL of the CU's.
 

JonnyLH

Banned
Are you ignoring my posts on purpose?

The Xbox One only has 2 ACE's with 4 compute queues and a fixed function Audio DSP, the PS4 has 8 ACE's with 64 compute queues which can execute compute and graphics threads in parallel as well as a basic Audio DSP.

It has more than enough compute "umph" compared to XBone.
Basic Audio DSP? You're looking at a lot more than than it the next-generation. You've just proved what I'm saying though, the GPU will most likely be doing audio tasks in games. This is perfectly fine, but you're not going to see 18 CU's dedicated to graphical rendering.

I don't get this, I'm being pin-pointed like I'm attacking a religion. The way Sony have architectured their machine like this is a very smart and productive thing. Are people that bothered about the CU's?
 
I don't understand the emphasis on this difference in power at all, it seems to me that we're making a mountain out of a molehill. The Ps3 proved that you can turn things around even if your console is weaker and very hard to code for, so why is this considered such a huge deal by some? The PS3 launched a full year later than the Xbox and was much more expensive yet it still did great, right? Why do you consider a much smaller price difference and an almost simultaneous launch such a disadvantage for the One? It doesn't make sense.
 

Skeff

Member
Which would be fine, but if you're using a CU for ALU, you can't make it do anything else. If the machine is hardware balanced at 14, this means you have to specify to use those 4. Just you then have to make sure everything is fine regarding the CPU.

Everything in this post is wrong, you don't need to dedicated a CU to do GPGPU.
 

KidBeta

Junior Member
Basic Audio DSP? You're looking at a lot more than than it the next-generation. You've just proved what I'm saying though, the GPU will most likely be doing audio tasks in games. This is perfectly fine, but you're not going to see 18 CU's dedicated to graphical rendering.

I don't get this, I'm being pin-pointed like I'm attacking a religion. The way Sony have architectured their machine like this is a very smart and productive thing. Are people that bothered about the CU's?

you do realise there are times in the pipeline when the CU's sit idle whilst the FF hardware is maxed out / working, you could do GPGPU during these times with a minor/tiny hit to rendering.
 

bonus_sco

Banned
I don't understand the emphasis on this difference in power at all, it seems to me that we're making a mountain out of a molehill. The Ps3 proved that you can turn things around even if your console is weaker and very hard to code for, so why is this considered such a huge deal by some? The PS3 launched a full year later than the Xbox and was much more expensive yet it still did great, right? Why do you consider a much smaller price difference and an almost simultaneous launch such a disadvantage for the One? It doesn't make sense.

The PS3 SPUs act a lot like GPGPU does. It was a very powerful machine, just hard to use. GPGPU is easier.
 
Exactly, extra load on the GPU. I don't see your point?

I don't see your point either, you don't have a clue what you are talking about, 14+4? This is old, and I might add, debunked information. It is extra load on the GPU but it has been massively beefed up making the point moot?

You are talking like its going to massively bog down the system noticeably.
 

Hollow

Member
Which would be fine, but if you're using a CU for ALU, you can't make it do anything else. If the machine is hardware balanced at 14, this means you have to specify to use those 4. Just you then have to make sure everything is fine regarding the CPU.

What does this even mean? There is no hardware division for the CUs. There is just 18 CUs and they can be used as the developer sees fit.

There's a chart for the GPU posted on the last page that shows that.
 

JonnyLH

Banned
you do realise there are times in the pipeline when the CU's sit idle whilst the FF hardware is maxed out / working, you could do GPGPU during these times with a minor/tiny hit to rendering.
Not so much fun to do this in practise. Its better to 'dedicate' hardware so you know exactly whats going on and when.
 

gruenel

Member
I don't get this, I'm being pin-pointed like I'm attacking a religion. The way Sony have architectured their machine like this is a very smart and productive thing. Are people that bothered about the CU's?

People are bothered because you're spreading FUD. You're making lots of claims without valid sources (PS4 DSP only does decoding, no dedicated video chip, "most" games on 360 use a whole core for audio, etc...).
 
Offtopic but damn I feel vindicated, I have been crowing this stuff at another forum since Feb which Jonny is a member of, I eventually got banned because all the Xbox One circle jerkers reported me for being a "troll".

It's nice to know that what I was saying was right by people who are more intelligent than me.

Also the PS Eye outputs uncompressed video so surely the PS4 has a Video DSP as well?
 
Top Bottom