SURFsara adds 96 GTX 1080Ti cards to Lisa Rekencluster




SURFsara has extended the Lisa Rekencluster with a gpu cluster, to give scientists the possibility of the parallel computing power of the graphic chips. The cluster has a peak strength of 1 petaflops.

Lisa ‘s gpu cluster consists of 23 nodes each with four Nvidia GTX 1080 Ti video cards and an Intel Xeon Bronze 3104 processor. Each node can have 256GB of ram and 1.5TB of ssd storage as scratchdisk. Furthermore, duo nodes with both four Nvidia Titan V cards have been added for testing purposes. According to SURFsara , the gpu system is mainly used for machine learning and modeling and pattern recognition within research fields such as molecular dynamics, radio astronomy and cryo-transmission electron microscopy.

UvA PhD student Bas Veeling was already allowed to use the cluster to make models more accurate in order to recognize tumor tissue based on scans. “On a good cpu computer, training my model might take a month or two, and with Lisa’s gpu cluster I’m ready in an hour, ” he says . Pinaki Kumar of Eindhoven University of Technology thinks that the system can be used to simulate earthquakes. “I develop and test my model on a desktop computer with a hefty GPU, and when I run it at SURF, I can easily mount as many GPUs as necessary: ​​the speed of the simulation grows linearly.”

Universities can use the cluster if they have a subscription to SURF’s Research Capacity Computing Service, but the organization also offers the computing power on a pay-per-use basis. Lisa runs Linux and consists of a large number of nodes that scientists can use for their calculations. The system is managed by SURFsara and the Vrije Universiteit Amsterdam and the UvA are also involved.

SURFsara Lisa


In: A Technology & Gadgets Asked By: [19085 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »