Has Nintendo discovered a new way to render graphics?

Lac3y

Daemon Poster
Messages
1,110
I need a person who has a good understanding of graphics to interpret this for me:

http://appft1.uspto.gov/netacgi/nph...f=G&l=50&co1=AND&d=PG01&s1=nintendo&OS=ninten

The patent belongs to Nintendo and it is logical that it will be implemented in their new console, codenamed Revolution.

I saw it on another forum and the people there seem to think that Nintendo has found a more efficient way to render textures, making displaying graphics faster, cheaper and cooler all the while making the hardware much smaller.

If this is true will the Rev be as, or more, capable of displaying graphics similar to PS3, at a fraction of the cost?

One can only hope!
 
Well the Xenos does the same thing, kinda...

THG said:
In fact, ATI has gained a lead by designing an architecture that uses a unified-shader approach. "Run that past me again," you say? Today, a PC graphics chip comprises two main processing functions: vertex shaders and pixel shaders; one has the task of processing 3D objects while the other processes pixels. The Xenos has now put these two processing functions into a single processing unit, which specialists in 3D seem to feel is the way of the future. This choice is supposed to provide higher performance and more efficient processing.
 
This is from my post about the Xenos in the General Electricals section, but without the pictures.

alvino said:
On to the GPUs. The ATI Xenos chip is one of the most unique GPUs ever made. With the PS3's RSX and the PC's normal graphics cards, the pixel pipelines and vertex shaders are all individual, so there will be a certain number of pixel pipelines and vertex shaders (like the 7800 GTX's 24 pixel pipelines and 8 vertex shaders). The disadvantage to this is that sometimes some of the pipelines or shaders won't be completely used, which means that they just don't do anything. This just makes it unefficient and a waste.

ATI realized that separating pixel pipelines and vertex shaders is a relatively inefficient process. Obviously you need both shader types, but why fix the number of how many you need? Thus the Unified Shader Architecture was concieved. Instead, ATI created generic shader engines that can be dynamically assigned to either pixel or vertex functions as needed (48 of them in the Xenos' case). You can devote all your engines to vertex processing when theres a lot of triangles, or if theres only a few large triangles but lots of pixels, you could devote all the shaders resources to pixle processing. ATI claims this unified shader model yeilds the best performance (120 billion operations per second) with maximum efficiency. This flexibility allows the GPU to efficiently utilize it's pipelines instead of leaving some of them hanging like the RSX (which uses the traditionally independent shaders).

According to ATI, the Xenos is actually TWO chips. GPU is the "parent" die, but there is also a "daughter" die, which has embedded memory, and inside that memory is intellegence that does a lot of the graphics processing from within the memory. Instead of relying on the main system memory to do anti-aliasing, Z, alpha processing and stencil processing, this will be done within this embedded memory. Add that with the intelligence inside the memory, you never have to leave it. This pretty much gives you infinite bandwidth.

To be precise, it gives you 2TB per second of memory bandwidth. The big payoff from this high overhead is in anti-aliasing, which is a major factor in the Xenos' HDTV output. If you compare with the R520 (or better known as the X1800), there is a lot of heavy lifting with Multi-Sampling and Super-Sampling, which entials a lot of extra drawing and blending. With the Xenos, all the anti-aliasing happens within the graphics memory, so there's no impact on surrounding system resources and you're not bogging down the main system memory with samples that never get used. This totally helps everything become more efficient and resourceful.

In my opinion, the Xenos is one of the best and unique GPUs to date. ATI has really done a great job designing this and making this chip possible. Yes, the PS3's RSX may have 50mhz more (Xenos runs at 500mhz while the RSX runs at 550mhz) of clock speed, it's going to be inherently inefficient and will run on pure brute force instead of grace. Also, the RSX doesn't have the Xenos' unique intelligent memory, which means that the PS3's 256mb XRAM (another 256mb is for the CPU, while another 256mb is for the GPU) will have the unfortunate honor of doing all those processes with extremely slow memory bandwidth compared to the Xenos.

I'm sorry for the long read, but I thought I'd just explain the Xbox 360's GPU, because not many people actually understand how it really works.
 
Back
Top Bottom