Some Xbox 360 and PS3 questions

Liandry

Regular
I would like to start a new thread about Xbox 360 and PS3. There is still some moments I didn't understood yet. I will try to ask questions by one and maybe someone can explan. Also would be interesting if some discusion starts here.
So question. Maybe someone haven't red this:
https://www.copetti.org/writings/consoles/xbox-360/
There in Organising the content section is info about Xbox Procedural Synthesis.
  • Firstly, the CPU’s front-side bus was adapted so the GPU can directly fetch data from the CPU’s L2 cache (where the CPU may procedurally generate the geometry) in blocks of 128 Bytes (the size of a cache line).
  • Secondly, a separate cached location was added into Xenon so the GPU can notify its current state to the CPU as fast as possible. Microsoft calls it the Tail pointer write-back and keeps both components in sync while the CPU updates the L2 cache and the GPU pulls from it. According to Microsoft, this routine provides a theoretical bandwidth of 18 GB/sec
As I know bandwith between CPU and GPU was 10.8 GB/s in each side, and here is info about 18 GB/s from CPU to GPU. How that is possible and is that info correct?
 
Damn, I didn't know this website. A lot of interesting data everywhere. I've don't know the answers to your questions, but thanx for the url.
 
I would like to start a new thread about Xbox 360 and PS3. There is still some moments I didn't understood yet. I will try to ask questions by one and maybe someone can explan. Also would be interesting if some discusion starts here.
So question. Maybe someone haven't red this:
https://www.copetti.org/writings/consoles/xbox-360/
There in Organising the content section is info about Xbox Procedural Synthesis.
  • Firstly, the CPU’s front-side bus was adapted so the GPU can directly fetch data from the CPU’s L2 cache (where the CPU may procedurally generate the geometry) in blocks of 128 Bytes (the size of a cache line).
  • Secondly, a separate cached location was added into Xenon so the GPU can notify its current state to the CPU as fast as possible. Microsoft calls it the Tail pointer write-back and keeps both components in sync while the CPU updates the L2 cache and the GPU pulls from it. According to Microsoft, this routine provides a theoretical bandwidth of 18 GB/sec
As I know bandwith between CPU and GPU was 10.8 GB/s in each side, and here is info about 18 GB/s from CPU to GPU. How that is possible and is that info correct?

They are perhaps changing the direction of one side of the bi-directional bus for a period of time.
 
This is just taking into consideration the compression feature:
A previously mentioned feature very important to XPS is the addition of D3D compressed formats that we implemented in both
the CPU and the GPU. To get an idea of this feature’s usefulness, consider this: Given a typical average of 2:1 compression and an XPS- targeted 9 Gbytes/s FSB bandwidth, the CPU cores can generate up to 18 Gbytes/s of effective geometry and other graphics data and ship it to the GPU 3D core. Main memory sees none of this data traffic (or footprint).
source
 
I still make some discoveries about 7th gen consoles. Anyone can tell is that true or false? Cell could do some staff on SPUs what can't be done on Xbox 360 and are there were games with wtaht stuff and if yes can anyone name some?
 
I still make some discoveries about 7th gen consoles. Anyone can tell is that true or false? Cell could do some staff on SPUs what can't be done on Xbox 360 and are there were games with wtaht stuff and if yes can anyone name some?
Curious how was the MLAA handled in GoW3? That game had excellent IQ
 
I still make some discoveries about 7th gen consoles. Anyone can tell is that true or false? Cell could do some staff on SPUs what can't be done on Xbox 360 and are there were games with wtaht stuff and if yes can anyone name some?
That's way too general. There's nothing SPUs could do that XB360 couldn't - the question is whether XB360 could do it fast enough without using up other resources needed elsewhere. You'd need to pin down something enabled on PS3 that was outside the reach of XB360 in games. For example, perhaps physics was better in one game on PS3 because the SPUs were well suited to the solvers of the time, but in that same game XB360 likely had advantages elsewhere.

At a low level, both consoles had inbuilt hardware features the other lacked. XB360 had a single-cycle dot-product instruction for example. XB360 could therefore calcuate a dot product in one cycle which was not possible on SPUs. But, did that get you anything in games written to the strengths of the systems? AFAIK this single-cycle instruction wasn't included in other AMD processors, so it likely didn't contribute meaningfully. So what do you count as possible, the low level execution or what was achieved on screen? Your question needs to be be more specific on that matter. ;)
 
Xenon was designed by IBM. AMD didn't have any involvement with the Xbox 360 development from what I recall. 360 came put in 2005, ATI was purchased in 2006.
 
At a low level, both consoles had inbuilt hardware features the other lacked.
Can you name some what was on PS3 please?
XB360 had a single-cycle dot-product instruction for example.
What is it?
XB360 could therefore calcuate a dot product in one cycle which was not possible on SPUs.
That was calculated on CPU or GPU?
So what do you count as possible, the low level execution or what was achieved on screen? Your question needs to be be more specific on that matter.
As I know in some games Cell was used for some of post processing effects and if I remember correctly those was beyond shader model 3.0 and couldn't be done on X360. But I don't know if thta's true.
 
Xenon was designed by IBM. AMD didn't have any involvement with the Xbox 360 development from what I recall. 360 came put in 2005, ATI was purchased in 2006.
Yeah, brain-fart. So used to talking about AMD and rarely see IBM these days.
 
Can you name some what was on PS3 please?
Things like a ring bus. Or the amount of cache. Or even 9 addressable cores versus three in Xenon.
What is it? That was calculated on CPU or GPU?
A CPU instruction to calculate the Dot product of two vectors in one cycle instead of using multiple cycles.
As I know in some games Cell was used for some of post processing effects and if I remember correctly those was beyond shader model 3.0 and couldn't be done on X360. But I don't know if that's true.
I don't know if that's true, but if so, they could have been done on the XB360 CPU. However, not fast enough to use in games. So what are you counting as 'impossible'?

This conversation is too broad to go anywhere.

Edit: I'll add for those who weren't here, B3D was the home of an intended Cell developers Forum. It was opened by Mike Acton, Engine Director of Insomniac games at the time, as a place to discuss best practices and algorithms. It seems to be no-longer accessible, sadly.
 
Last edited:
Things like a ring bus. Or the amount of cache. Or even 9 addressable cores versus three in Xenon.
Two moments here. First, Cell have 7 cores, because one SPU is disabled and one was used for some security calculations. Second, I heard from one developer, and if I remember correctly other agreed, for programmer there is 6 cores cpu in Xbox 360, because there is multi threading and 2 VMX blocks per physical core, so one for each thread. And he also said that second thread give up to 85% of performance as first thread. Maybe that was what his team achieved on X360 and second developer agreed because studios where he worked achieved similar result.
So what are you counting as 'impossible'?
Impossible is when performance to low to add feature in final game. Like MLAA.
 
This article seems to praise the results of MLAA.
But despite how well it refined the image, I m curious why we didn't see more of it in games
 
This article seems to praise the results of MLAA.
But despite how well it refined the image, I m curious why we didn't see more of it in games
After GOW 3 most if not all Sony first party games used MLAA. Motorstorm sequels, Killzone 3, Resistance 3, Twisted Metal, LBP3, maybe more.
 
As I know anly beggers can be stored in EDRAM on XBox 360. Bandwith between main core and EDRAM core is 32 GB/s, so that is 16 GB/s each direction. If we devide that to 30 (for 30fps game) then that is almost 550 MB per frame. But why ATI made GPU with so much bandwith? I mean if game uses 1 tile that is 10 MB per frame, so 10 MB will be moved each direction. Can anyone explain please?
 
As I know anly beggers can be stored in EDRAM on XBox 360. Bandwith between main core and EDRAM core is 32 GB/s, so that is 16 GB/s each direction. If we devide that to 30 (for 30fps game) then that is almost 550 MB per frame. But why ATI made GPU with so much bandwith? I mean if game uses 1 tile that is 10 MB per frame, so 10 MB will be moved each direction. Can anyone explain please?
beyond3d probably has the best article I've read on Xenos. Lots of water passed under the bridge since then, but the article remains as a treasure.
 
I know about that article for years. :) But there 8s no answer for my question. Interesting I hwven't thought about that untill today. For PS2 that is understandable. All that massive bandwith was used for multipass, because data swould be moved many many times, and for textures, that also was used many many times. But I don't understand for what that bandwith was used on XBox 360.
 
Back
Top