UMA hardware resourcesNavigation : Computational servers | Cluster | GPGPU | Members and users
The computational servers
- 1 computer : Proc. 24 logical cores with Hyperthreading / 12 physical cores - 256 Go RAM - 1 To HD (2015)
- 1 computer : Proc. 8 logical cores with Hyperthreading / 16 physical cores - 768 Go RAM - 1 To HD (2015)
- 1 computer : Proc. 64 logical cores with Hyperthreading / 32 physical cores - 1 To RAM - 300 Go HD (2014)
- 2 computers : Proc. 32 logical cores with Hyperthreading / 16 physical cores - 128 Go RAM - 935 Go HD (2013 - 2014)
- 2 computers : Proc. Intel Sandy Bridge XEON E3-1280 at 3,49 GHz / 4 coeurs - 32 Go RAM - 500 Go HD (2011)
- 2 computers : Bi-pro. Intel Nehalem XEON 5670 at 2,93 GHz / 6 coeurs - 24 Go de RAM - 500 Go HD (2010)
In order to meet the needs of high performance computing and scalable algorithms, in 1999 the UMA adopted a cluster consisting of several interconnected PCs via a bi-processor standard network layer. This cluster consisting of 32 heterogeneous machines is renewed every four year thereby maintaining a degree of investment. Today, the UMA has 64 processors with 512 cores interconnected by a Gigabit network layer.
This type of configuration can offer a significant computational power in a context of parallel code execution and scaling. The cluster now utilizes a Linux operating system which is relatively simple platform to implement thereby reducing the cost of administration.
The graphic unit (GPU)UMA also has two GPU servers:
- Compute Node Hybride CPU/GPU CUDA 4220 (8 Intel cores XEON E5-2637 / 1 x Tesla K80),
- Nvidia Tesla S1070-500 16G (4 x Nvidia Tesla C1060).
All UMA hardware resources are fully administered and managed by Christophe Mathulik.