XBOX 360: Inside & Out (NO CPU die photo!)

Titanio said:
Megadrive1988 said:
how many ALUs make up a conventional pixel pipe and a conventional vertex pipe?

I believe the pixel shaders in NVidia's and ATi's latest have 2 ALUs per shader, not sure about the vertex shaders, though. And they may not be very comparable to the ALUs in the R500.

AFAIK ATI themself counting 1 5D ALU per vertex pipe. They don't count scalar and vector seperate. You can expect 1 shader op/cycle/pipe but because it can co-issue a scalar and vector op often you can get 2ops

Pixelshader ALU from r420 is 1 full 4D + 1 mini 4D per pipe. As with the vertex shader ALU it can co-issue a scalar and vector op for both full and mini ALU.

The mini ALU however has only limited cababilitys but i'm not sure how much they can do compared to a full one
 
Whitepaper said:
Each vertex processing unit actually includes two Arithmetic Logic Units (ALUs): a 128-bit floating point vector ALU plus a 32-bit floating point scalar ALU.
Jawed
 
tEd said:
The mini ALU however has only limited cababilitys but i'm not sure how much they can do compared to a full one

It has the PS1.4 modifiers and they've always been a little coy on telling what, if any, other instructions they have.
 
I don't think it matters how you count them. Important is that every ALU from the 48 in r500 is capable of co-issue a scalar and vector op like those in r420. That's where the 48billion shader ops/s comes from in the first place

It means if you count the vector and scalar unit as 2 ALU's in r420 you need to do the same thing with the ALU's in r500 which means r500 has 48 vector ALU's and 48 Scalar ALU's
 
I don't really even wan't to pretend that I understand even half of this stuff, but how many shader operations would for example X850XT-PE produce?.
 
I think the shader and ligntning power of Xbox360 are lower than
a R520,the 10mb eDRAM uses alot of transistors,and the other problem
is the systembus with only 22,4 Gbyte/s,to low to move big 3d-worlds
with 60 frames/s in HDTV resolution,it is not a big jump to the 6,4Gbyte/s
of Xbox1.I think the Bussystem of Xbox360 is a big bottleneck.
 
Ted you're having a laugh.

If ATI says there are 92 ALUs in R420 and the leak says there are 48 ALUs in R500, and coincidentally the peak fill-rate of R500 is half that of R420 per clock, then don't you think it's about time to admit that R500's ALU capacity is roughly half that of R420?

Jawed
 
Jawed said:
I don't think 12 ALU ops (each ALU doing vector + scalar) per clock per fragment is viable.

Register dependencies between lines of code will mean most of your ALUs are sitting around doing nothing most of the time.

Jawed

A lot more (48???) than 8 fragments are processed at once, 8 are just the max output per cycle (just like the NV43 are processing 8 fragments at the time, but can only output four per cycle).
 
Jawed said:
Ted you're having a laugh.

If ATI says there are 92 ALUs in R420 and the leak says there are 48 ALUs in R500, and coincidentally the peak fill-rate of R500 is half that of R420 per clock, then don't you think it's about time to admit that R500's ALU capacity is roughly half that of R420?

Jawed

No actually i think r500 has roughly double the shader power of a r420 per cycle.

Pixelfillrate is not so important as we go on anymore. Shader ops/cyle is. You don't need 8GPixel/s for 1280*720 res.
 
"Pipes" are not really comparable between the two...let's not even bring that into this discussion for sake of clarity. ;)

If someone would clarify..is this correct?

1) R500 has about 2x the shader op power per cycle. vs. R420

2) R500 has about 1/2 the "non AA" fill rate power per cycle vs. R420.
 
Joe DeFuria said:
"Pipes" are not really comparable between the two...let's not even bring that into this discussion for sake of clarity. ;)

If someone would clarify..is this correct?

1) R500 has about 2x the shader op power per cycle. vs. R420

2) R500 has about 1/2 the "non AA" fill rate power per cycle vs. R420.

Imo that's about correct
 
Jawed is still anchored to an 'old' conception of ALUs and pixel pipelines coupling.
 
Agreed on all that Joe. That's why prefer to refer to them as Unified Shader Engines.

1 and 2 are both correct in my opinion.

Jawed
 
nAo said:
Jawed is still anchored to an 'old' conception of ALUs and pixel pipelines coupling.

:LOL: Anything but. A Unified Shader Engine has an Arbiter that distributes workload to an ALU Engine and a Texture Unit. The command threads (vertex or pixel) use ALU or texturing resources as they become available.

The arbiter juggles the command threads to ensure that both the ALU Engine and the Texture Unit are running as close to 100% as possible.

Jawed
 
JF_Aidan_Pryde said:
Okay, but they don't have to use it. But being denied the possibility to use keyboard and mouse makes no sense IMO.

Of course it does. It would make it completely unfair to those who didn't use a mouse and keyboard in say online FPS games like Perfect Dark Zero. In order to maintain gameplay balance they have to remove M/K support for gameplay so the majority of users with controllers aren't at a major disadvantage when playing online multiplayer.
 
Ruined said:
Of course it does. It would make it completely unfair to those who didn't use a mouse and keyboard in say online FPS games like Perfect Dark Zero. In order to maintain gameplay balance they have to remove M/K support for gameplay so the majority of users with controllers aren't at a major disadvantage when playing online multiplayer.

Sad, but true.

The frustrating part is some of us are really good with a KB and only ok with a gamepad. One of the problems I have with gamepads is their size... small hands do not play nicely with giant gamepads. Nor do sweaty hands...

But having a level playing field is important. But with USB 2.0 I can see certain game makers going with KB/MS option...
 
That's why you include it in the base package.
That way, everybody has it. And game developers can create games that are actually designed for M/K instead of just the oddball hacked implementation that exists now (which sucks in most cases).
I'd rather have a cheap mouse and keypad included in my retail X360 box than a stupid changeable faceplate. I'd even give up DVD functionality (without purchase of extra remote) in order to gain a M/K. They could put the stupid thing in a 40 pound oak barrel for a case, and I'd still be happy as long as I got a mouse and keypad to control my games.
There is no excuse for requiring gamers to use a poor solution to control their games (FPS and 3rd person action titles mostly) when the best possible control setup exists and can be included cheaply.
 
Back
Top