Differences

This shows you the differences between two versions of the page.

--- team_gpu [2023/11/24 15:16] – created Luc Bonnefond
+++ team_gpu [2024/02/09 09:50] (current) – Luc Bonnefond
@@ Line 3: / Line 3: @@
 The GPU server is installed ate IGBMC for the moment
-Specs :
+**Specs** :
-CPU : 2x Intel Xeon-G 6330
+  * CPU : 2x Intel Xeon-G 6330
-RAM : 16x 32GB DDR4-2933/3200 RDIMM (512GB total)
+  * RAM : 16x 32GB DDR4-2933/3200 RDIMM (512GB total)
-GPU : 4x Nvidia A10 (4x 24GB)
+  * GPU : 4x Nvidia A10 (4x 24GB)
-HDD : 6x 1.92 To SSD SAS (1 disk for OS, 5 for scratch)
+  * HDD : 6x 1.92 To SSD SAS (1 disk for OS, 5 for scratch)
-years support
+  * 7 years support
+It is part of the IGBMC HPC, so it's accessible through ssh (with MobaXterm for instance)
+<code>ssh -X login@hpc.igbmc.fr</code> and then: <code>ssh -X login@phantom-node39</code>
+space2, mendel and storage can be accessed
+<code>
+/shared/space2/lamour-ruff/
+/shared/mendel/projects/
+/shared/mendel/teams/lamour-ruff/
+/shared/misc/cbi-teams/lamour-ruff/
+</code>
+The CryoSPARC instance can be accessed online at :
 [[https://cavarelli-cryosparc.igbmc.fr]]
-It is part of the IGBMC HPC, so it's accessible through ssh
+The submission script has to be modified to increase the allocated RAM for some jobs (not enough memory for 2D classes Job with 200 classes in the Extensive validation EMPIAR 10025 Benchmark Mode - Edvanced jobs enabled)
-ssh login@hpc.igbmc.fr
-ssh loggin@phantom-node39
+<code>#SBATCH --mem={{ (ram_gb*(ram_adjust|int))|int }}M</code>
+with ram_adjust set to 1500 (1,5x times the memory allocated by default)