Hypnotoad

Hypnotoad is a small computational cluster administered by PSD Computing.  We provide access to PSD Faculty, Staff and Students upon request.  

We are targeting users who have modest computational requirements and no funding for dedicated high-performance computing equipment.   Users with serious compuational needs should contact the RCC to arrange for access to the Midway Cluster.

Hypnotoad may be small in comparison to Midway but there are some advantages to running jobs on the Toad.  We can hopefully be more responsive to any special needs that you may have on the system.  Hypnotoad is far less used than Midway so you can get un-interrupted time on nodes.  As of the time of this update we have no Wall time set so you can run jobs for as long as you require.

Hardware

  1. Hypnonode 0-0:  One AMD Trinity A-10 3.8GHz APU*.  4-cores, 16GB RAM, GTX620 GPUs**, 20Gb Infiniband
  2. Hypnonode 0-1:  One AMD FX8350 8-core @4GHz, 16GB, GTX 620**, 20Gb Infiniband
  3. Hypnonodes 0-2 through 0-7:  One AMD A8 2.9GHz quad-core APUs*.  4-cores, 16GB RAM, Nvidia GT620 GPU, 20Gb Infiniband.    
  4. Hypnonodes 0-9 through 0-12:  Dell PowerEdge C6105 w/ Dual AMD Opteron 4332HE 3GHz. 12 cores, 16GB RAM
  5. Hypnonodes 1-0 through 1-4:  AMD A8 2.9GHz quad-core APUs.  16GB RAM. Nvidia GTX550 GPUs, Hypnonodes 0-2 through 0-7:   AMD A8 2.9GHz quad-core APUs.  16GB RAM.  20Gb Infiniband 
  6. Hypnonodes 2-0 through 2-5:  Dual AMD Opteron 2.6GHz dual-core CPUs [4 cores total].  8GB RAM

* The AMD APUs have on-die OpenCL-compliant GPUs.  Nodes 0-2 to 0-7 & 1-0 to 1-4 are Llano-series APUs with strong GPGPU integer performance and poor FP performance.  Node 0-0 has a slightly newer 'Trinity'-series APU with stronger GPGPU FP performance.

** The GTX 620 GPUs are too slow to be useful GPGPU compute engines, but they will allow you to test your CUDA code.

BECOME A TOAD, REQUESTING AN ACCOUNT

We built hypnotoad primarially for the unfunded Students and Post-Docs who require CPU time but who have no other options available to them.  This was before the RCC changed their policy to not deduct student CPU time from their sponsor's grants.

For accounting, we would like you provide the name of your advisor or the PI who you're otherwise primarially working with.   We don't bill for time on Hypnotoad.  This is just information to keep you all straight.

You can request an account here.

USING HYPNOTOAD

Hypnotoad jobs are managed by the SLURM queuing software.  This is the same software package that the RCC uses on the Midway cluster.  Aside from being a compute resource, Hypnotoad is also a good place to familiarize yourself using queueing software before you move on to larger systems.

You can find information about creating and submitting compute jobs on hypnotoad at our Using Hypnotoad page.   The RCC has also posted some examples of SLURM job scripts on their website.