This is straight from a technical article written by two high level engineers in the Xbox group.
HD, alpha blending, z-buffering, antialiasing, and HDR pixels take a heavy toll on memory bandwidth. Although more effects are being achieved in the shaders, postprocessing effects still require a large pixel-depth complexity. Also as texture filtering improves, texel fetches can consume large amounts of memory bandwidth, even with complex shaders. One approach to solving this problem is to use a wide external memory interface. This limits the ability to use higher-density memory technology as it becomes available, as well as requiring compression. Unfortunately, any compression technique must be lossless, which means unpredictable—generally not good for game optimization. In addition, the required bandwidth would most likely require using a second memory controller in the CPU itself, rather than having a unified memory architecture, further reducing system flexibility.
EDRAM was the logical alternative. It has the advantage of completely removing the render target and the z-buffer bandwidth from the main-memory bandwidth equation. In addition, alpha blending and z-buffering are read-modify-write processes, which further reduce the efficiency of memory bandwidth consumption. Keeping these processes on-chip means that the remaining high-bandwidth consumers—namely, geometry and texture—are now primarily read processes. Changing the majority of main-memory bandwidth to read requests increases main-memory efficiency by reducing wasted memory bus cycles caused by turning around the bidirectional memory buses.
Results 1 to 12 of 12
Thread: Why the 360 GPU uses EDRAM
Why the 360 GPU uses EDRAM
I still don't really understand it. Perhaps a techie on here can explain it in average terms?
- Join Date
- Oct 2006
- Rep Power
And to the techie, please tell the downside to using edram. Thanks!
- Join Date
- Dec 2005
- Rep Power
- Join Date
- Jun 2006
- Rep Power
And? We have known this(well at least that has been my argument to defend the Xbox 360s low memory bandwidth)
Basically from what I have read, lots of memory bandwidth is needed. One solution to use less memory bandwidth is to use a wide external memory interface. It limits the ability to use higher-density memory technology and require lossless compression which is "unpredictable—generally not good for game optimization". It also require using a second memory controller in the CPU instead of the unified memory architecture which Xbox 360 use so it reduce Xbox 360's system flexibility.
In order to handle the problem from Xbox 360's unified memory architecture and excess amount of memory bandwidth in use, EDRAM is used for Xbox 360 by removing whether the memory bandwidth is used for from the main-memory bandwidth and make it more efficient. It also mean that the remaining high-bandwidth consumers, geometry and texture, are now primarily read processes. By changing many of the main-memory bandwidth to read requests, it make the main-memory make it efficient by reducing wasted memory bus cycles caused by turning around the bidirectional memory buses (from CPU?).
Is my interpretation correct? What I have noticed that is that geometry and texture processes are now primarily read processes and I recalled something about Xbox 360's GPU lacking the geometry shader that process vertex, pixel, and geometry shaders more efficiently. Was it because of the processes being primarily read or is there other reasons for this?"If the evolutionary roads like the PS3 are closed off, the industry will no longer grow." - Hideo Kojima
"Oh God... please don't let me die today! Tomorrow would be so much better!"
"Management is like playing a strategy game." - GalCiv2
Weekly Console Hardware Sales in Japan last updated on March 26, 2009
I'm not seeing the downside.
note: i didnt say there isnt one$sql->Query("SELECT * FROM users WHERE clue != 0");
0 rows returned
Member # 8 of the balanced gamers group
10-24-2006 #9When it comes to porting Final Fantasy XIII to Xbox 360, Square's of the opinion that it's not the size of the disc that matters; it's how you use it.
10-24-2006 #10The /)octo/2Guest
I honestly don't think there is a downside.
Ok I'll try and explain for you all.
When you render a scene (thats the picture you see on the screen), you basically draw every single thing, or simliar thing , then draw the next thing, then the next etc etc,
Now imagine when you draw, say a hillside, you write all this to the screen, then you draw some trees on top of this. You've now written twice to some parts of the screen, then you draw your hud on top of this, now you've written over say the original hillside maybe 3 times.
Everytime you "draw" on the screen it's actually wirting back to you video memory.
So the normal process is...
MEMORY-> GPU-> BACK TO MEMORY
As you can see this happens for everything you draw, so you writing as much as your reading to and from main memory. This happens probably about 200-500 times for anyone "frame".
Now what the edram allows to happen is this
So you not ever writing back to the main memory untill the frame is "complete", in otherwords when you finished the completed scene, then you write the image back to the main memory once, not hundreds of times.
Now before you write this image back you can do post processing effects, what this really means is, you scene is completed but you decided to say blur some of it or all of it as is the case of antiailising, you can do all this in superfast edram, then write it back to main memory to be displayed.
In short you're not going read model, write screen , read model, write screen.
You're going read read read read read, finished, write.
So you required bandwidth is infact pretty much halved.
I.e. xbox360 bandwidth is infact roughly twice that of the PS3's vram specs.
Assuming PS3 hasnt done something similar.
Users Browsing this Thread
There are currently 1 users browsing this thread. (0 members and 1 guests)
By dexvex in forum PlayStation 3 and Legacy ConsolesReplies: 34Last Post: 07-29-2006, 08:34