GPU cluster

From Wikipedia, the free encyclopedia

This article is orphaned as few or no other articles link to it.
Please help introduce links in articles on related topics. (September 2006)

This article does not cite any references or sources. (December 2006)
Please help improve this article by adding citations to reliable sources. Unverifiable material may be challenged and removed.

A GPU cluster is a computer cluster in which each node is equipped with a Graphics Processing Unit (GPU). By harnessing the computational power of modern GPUs via General-Purpose Computing on Graphics Processing Units (GPGPU), very fast calculations can be performed with a GPU cluster.

1 Hardware (GPU)
2 Hardware (Other)
3 Software
4 Algorithm Mapping

[edit] Hardware (GPU)

The hardware classification of GPU clusters fall into two categories: Heterogeneous and Homgeneous.

Heterogeneous

Hardware from both of the major IHV's can be used (ATi and nVidia). Even if different models of the same GPU are used (ie 7800GT mixed with 7800GTX) the gpu cluster is considered hetergeneous.

Homogeneous

Every single GPU is of the same hardware class, make, and model. (ie a homogeneous cluster comprised of 100 Sapphire 7800GTs, all with the same amount of VRAM)

Classifying a GPU cluster according to the above semantics largely directs software development on the cluster, as diffent GPUs have different capabilities that can be utilized.

[edit] Hardware (Other)

Interconnect

In addition to the computer nodes and their respective GPUs, a fast enough interconnect is needed in order to shuttle data amongst the nodes. The type of interconnect largely depends on the number of nodes present. Some examples of interconnects include Gigabit Ethernet and InfiniBand.

[edit] Software

The software components that are required to make many GPU-equipped machines act as one include:

1) Operating System
2) GPU driver for the each type of GPU present in each cluster node.
3) Clustering API (such as the Message Passing Interface, MPI).

[edit] Algorithm Mapping

Mapping an algorithm to run a GPU cluster is somewhat similar to mapping an algorithm to run on a traditional computer cluster. Example: rather than distributing pieces of an array from RAM, a texture is divided up amongst the nodes of the GPU cluster.