Information on current GPU Architectures

Question

I have decided that my bachelors thesis will be about general purpose GPU-computing and which problems are more suitable for this than others. I am also trying to find out if there are any major differences between the current GPU architectures that may affect this.

I am currently looking for some scientific papers and/or information directly from the manufacturers about the current GPU Architectures, but I can't seem to find anything that looks detailed enough. Therefore, I am hoping that someone may be able to suggest some papers or at least point me in the right direction.

Thank you

Answer 1

Read litterature about OpenCL and Cuda. Perhaps OpenGPU project site might give more links.

_{GPU manufacturers usually don't want to give detailed information. ATI seems to tell more than Nvidia. Some of ATI GPU chips have published VLIW instruction sets.}

The trend is to give (e.g. in OpenCL runtime) several functions to query the hardware, and then enable you to tune your (OpenCL) application to what the hardware wants.

Answer 2

This paper from NVIDIA provides some details on their Tesla architecture:

• NVIDIA Tesla: A Unified Graphics and Computing Architecture

Of course, Tesla is one generation old at this point; I have not seen a peer-reviewed publication describing Fermi, NVIDIA's latest architecture. The best information from NVIDIA on Fermi may be in this white paper:

• NVIDIA’s Next Generation CUDA Compute Architecture: Fermi

Unfortunately, I'm not aware of similar papers from AMD/ATI. I have seen a few third party papers that do provide some information; for example:

• A Micro-benchmark Suite for AMD GPUs

One of the best third party sources of architecture information, in my opinion, is Real World Technologies. Here are some relevant articles from RWT:

• AMD's Cayman GPU Architecture
• AMD Fusion Architecture and Llano
• Inside Fermi: Nvidia's HPC Push
• NVIDIA's GT200: Inside a Parallel Processor