What is a Graphics Processor Unit (GPU) Cluster and why was it developed?
SUrge was developed to assist researchers with computational research tasks that can take substantive advantage of the speed-up GPU’s can provide. We developed the cloud methodology to make allocating the GPU’s to a number of different workloads a simple process.
From Wikipedia: 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.
What specific computing needs does it serve?
SUrge is designed to provide GPU access to researchers to significantly increase computation capabilities, allow for development utilizing CUDA/OpenCL, and gain familiarity with GPU capabilities. Additionally, GPU’s are used in a wide variety of academic areas; they provide a significant speed increase over traditional CPU’s for certain types of mathematical operations.
What hardware and system configurations are available?
- 270 commodity GPU’s are available through SUrge (NVIDIA GeForce GTX 750 Ti’s)
- Both Linux and Windows operating systems are supported
- Nodes can range from 1 to 16 GPUs and from 1 to 32 cores. The appropriate CPU to GPU ratio is dependent on your specific application
- Each GPU operates with 2 GB of memory
- Both CUDA and OpenCL are supported as GPU programming languages
Who can use this system?
SUrge is open and free of charge to all SU-affiliated researchers.
When will it be available?
SUrge is available now! We are looking for initial researchers who would like to take advantage of the new resource.
Who can provide assistance in using this system?
For help contact our research computing group – email firstname.lastname@example.org.
Does this system replace or complement OrangeGrid?
Both! SUrge will be available as standalone virtual machines and also as part of OrangeGrid.