<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>GPU Cluster |</title><link>https://albarqouni.github.io/compute/</link><atom:link href="https://albarqouni.github.io/compute/index.xml" rel="self" type="application/rss+xml"/><description>GPU Cluster</description><generator>HugoBlox Kit (https://hugoblox.com)</generator><language>en-us</language><image><url>https://albarqouni.github.io/media/logo_hu_43a0798e7da4e673.png</url><title>GPU Cluster</title><link>https://albarqouni.github.io/compute/</link></image><item><title>Marvin</title><link>https://albarqouni.github.io/compute/marvin/</link><pubDate>Mon, 11 Mar 2024 00:00:00 +0000</pubDate><guid>https://albarqouni.github.io/compute/marvin/</guid><description>&lt;p&gt;&lt;strong&gt;Marvin&lt;/strong&gt; is the large tier-3 HPC cluster of the University of Bonn, run by the University Computing Centre (HRZ) and in operation since March 2024. It is listed on the
list of the world&amp;rsquo;s most powerful supercomputers and is open to (HPC-experienced) scientists of the University, including the University Hospital Bonn (UKB) and their external project partners.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;GPU partitions&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;SGPU&lt;/strong&gt; — 32 nodes, each with 4 × NVIDIA A100 80GB (NVLink), 2 × AMD EPYC &amp;ldquo;Milan&amp;rdquo; 64-core, 1 TB RAM: for highly scalable GPU jobs.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;MLGPU&lt;/strong&gt; — 24 nodes, each with 8 × NVIDIA A40 48GB, 2 × AMD EPYC &amp;ldquo;Milan&amp;rdquo; 64-core, 512 GB RAM: for machine learning.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;CPU and storage&lt;/strong&gt; — 192 standard, 24 large-memory and 5 very-large-memory nodes (Intel Xeon &amp;ldquo;Sapphire Rapids&amp;rdquo;), 5.6 PB Lustre storage and an InfiniBand NDR 200 Gb/s network.&lt;/p&gt;
&lt;p&gt;Access requires a Uni-ID and an application via the
, where full technical details are listed.&lt;/p&gt;</description></item><item><title>Bender</title><link>https://albarqouni.github.io/compute/bender/</link><pubDate>Wed, 01 Jun 2022 00:00:00 +0000</pubDate><guid>https://albarqouni.github.io/compute/bender/</guid><description>&lt;p&gt;&lt;strong&gt;Bender&lt;/strong&gt; is a small GPU cluster installed by the University Computing Centre (HRZ) in 2022. It is mainly intended for teaching, and research projects are also given access when resources allow.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;A100 nodes&lt;/strong&gt; — 2 nodes, each with 8 × NVIDIA A100 80GB and 2 × AMD EPYC 7713 64-core.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A40 nodes&lt;/strong&gt; — 4 nodes, each with 4 × NVIDIA A40 48GB.&lt;/li&gt;
&lt;li&gt;Rocky Linux, Slurm batch system, CUDA; 200 Gb/s InfiniBand.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Access requires a Uni-ID; students apply with a supervisor. Apply and see the full specification on the
.&lt;/p&gt;</description></item><item><title>Lab GPU Server</title><link>https://albarqouni.github.io/compute/lab-gpu-server/</link><pubDate>Thu, 05 Nov 2026 00:00:00 +0000</pubDate><guid>https://albarqouni.github.io/compute/lab-gpu-server/</guid><description>&lt;p&gt;The lab&amp;rsquo;s own GPU server at the University Hospital Bonn, run together with UKB-IT. Its configuration is similar to the A100 nodes of the University&amp;rsquo;s
and
clusters.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;GPUs&lt;/strong&gt; — 8 × NVIDIA A100 80GB PCIe (640 GB GPU memory in total), in one node.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;CPUs&lt;/strong&gt; — 2 × AMD EPYC 7643, 48 cores each.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Network&lt;/strong&gt; — InfiniBand 200 Gb/s.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Software&lt;/strong&gt; — Linux, Slurm batch system, CUDA.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Access is restricted to members of the Albarqouni Lab with a UKB email account (@ukbonn.de) and an affiliation to our lab. To request an account, use &lt;strong&gt;Apply for access&lt;/strong&gt; above and write from your UKB email address.&lt;/p&gt;</description></item></channel></rss>