Differences

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

--- en:user:server [2017/02/20 08:52] – [Compute servers] add compu4, fix spelling, fix RAM, add OS, add fingerprints link Jan Henrik Sylvester
+++ en:user:server [2026/09/07 10:49] (current) – [Compute servers] restart with no CPU load Jan Henrik Sylvester
@@ Line 3: / Line 3: @@
 ===== Login servers =====
   * login1.math.uni-hamburg.de and login2.math.uni-hamburg.de
-  * are available from outside math.uni-hamburg.de via ssh (see [[user:server:fingerprints]]),
+  * are available from outside math.uni-hamburg.de via ssh (see [[en:user:server:fingerprints]]),
   * have limited resources, what makes them suitable for
     * logging into compute servers,
@@ Line 12: / Line 12: @@
 ===== Compute servers =====
-  * compu1.math.uni-hamburg.de with 24 CPU cores and 40 GB RAM running Ubuntu 14.04 and
+  * compu3.math.uni-hamburg.de with 20 CPU cores and 384 GB RAM running Ubuntu 24.04,
-  * compu2.math.uni-hamburg.de with 24 CPU cores and 48 GB RAM running Ubuntu 14.04 and
+  * <del>compu4.math.uni-hamburg.de with 24 CPU cores and 512 GB RAM running Ubuntu 24.04</del> (network problems),
-  * compu3.math.uni-hamburg.de with 32 CPU cores and 480 GB RAM running Ubuntu 16.04 and
+  * compu5.math.uni-hamburg.de with 64 CPU cores and 1504 GB RAM running Ubuntu 24.04,
-  * compu4.math.uni-hamburg.de  with 24 CPU cores and 512 GB RAM running Ubuntu 14.04
+  * compu1.math.uni-hamburg.de with 192 CPU cores and 1536 GM RAM running Ubuntu 26.04 (new machine in 2026),
-  * are only available from inside the IP-area math.uni-hamburg.de via SSH (see [[user:server:fingerprints]]), for example from login servers or desktop computers,
+  * gpu1.math.uni-hamburg.de with 2 NVIDIA A30 (good for double precision, 24GB GPU memory each), 24 CPU cores and 128GB RAM running Ubuntu 24.04, and
-  * will usually only be restarted after prior announcement on the department blog,
+  * gpu2.math.uni-hamburg.de with 1 NVIDIA RTX PRO 6000 (better for single precision, 96GB GPU memory), 96 CPU cores and 768GB RAM running Ubuntu 26.04 (new machine in 2026, will be shared with PHYSnet later).
-  * are suitable for computation jobs.
+  * are only available from inside the IP-area math.uni-hamburg.de via SSH (see [[en:user:server:fingerprints]]), for example from login servers or desktop computers,
+  * will usually only be restarted after prior announcement on the department blog or if there is no CPU/GPU load,
+  * are suitable for computation jobs,
+  * have shared resources, please watch your usage with `top`, `nvtop`, and similar tools and coordinate with other users. Especially GPU memory must always be freed after calculations.
 ===== Computations independent of the ssh connection =====
@@ Line 29: / Line 33: @@
 ''kpagsh''
-You needs to obtain new credentials:\\
+You needs to obtain new credentials [[en:user:server:kpagsh|(Explanation)]]:\\
 ''%%kinit -l 2w --renewable -r 3month%%''