The Puhti supercomputer was launched on September 2, 2019. It is an Atos cluster system with a variety of different node types. It is targeted at a wide range of workloads, and it can also be used to run larger simulations until Mahti becomes available.
Puhti has a total of 682 CPU nodes, with a theoretical peak performance of 1,8 petaflops. Each node is equipped with two latest generation Intel Xeon processors, code name Cascade Lake, with 20 cores each running at 2,1 GHz. The interconnect is based on Mellanox HDR InfiniBand. The nodes are connected with a 100 Gbps HDR100 link, and the topology is a fat tree with a blocking factor of approximately 2:1.
The Puhti AI artificial intelligence partition has a total of 80 GPU nodes with a total peak performance of 2,7 petaflops. Each node has two latest generation Intel Xeon processors, code name Cascade Lake, with 20 cores each running at 2,1 GHz. They also have four Nvidia Volta V100 GPUs with 32 GB of memory each. The nodes are equipped with 384 GB of main memory and 3,6 TB of fast local storage. This partition is engineered to allow GPU-intensive workloads to scale well across multiple nodes. The interconnect is based on a dual-rail HDR100 interconnect network connectivity providing 200 Gbps of aggregate bandwidth in a non-blocking fat-tree topology.
A detailed list of the nodes:
|Name||Number of nodes||Compute||Cores||Memory||Local disk|
|M||532||Xeon Gold 6230||2 x 20 cores @ 2,1 GHz||192 GiB|
|L||92||Xeon Gold 6230||2 x 20 cores @ 2,1 GHz||384 GiB|
|IO||40||Xeon Gold 6230||2 x 20 cores @ 2,1 GHz||384 GiB||3600 GiB|
|XL||12||Xeon Gold 6230||2 x 20 cores @ 2,1 GHz||768 GiB|
|BM||12||Xeon Gold 6230||2 x 20 cores @ 2,1 GHz||1,5 TiB|
|GPU||80||Xeon Gold 6230
|2 x 20 cores @ 2,1 GHz
4 GPUs connected with NVLink
4 x 32 GB
In addition to the compute nodes above, Puhti has two login nodes with 40 cores and 2900 GiB local disk each.
The login nodes can be used for light pre- and postprocessing, compiling applications and moving data. All other tasks are to be done in the compute nodes using the batch job system. Programs not adhering to these rules will be terminated without warning. Note that compute nodes can be used also interactively
The Quantum Learning Machine
Quantum computers differ from their classical counterparts when it comes to the basic computational operators. Before QPUs can be utilized, they require tailor-made programs and algorithms. With Kvasi, the user can explore and develop algorithms for quantum computers. Read here detailed instructions on how to access
Kvasi provides an ecosystem for developing and simulating quantum algorithms in both ideal, and realistic, noisy conditions. With Kvasi, you can optimize your algorithm for a specific hardware (QPU), with specific connections and basic gate operations.
The algorithms can be developed either at a level close to the hardware, using the Atos Quantum Assembler (AQASM) language, or using a higher level, Python based language and ready-made libraries. The QLM comes with several ready-made examples. You can also download and run locally myQLM - a light-weight version of the QLM ecosystem.