Nam-shub-02, the second graphics processing unit (GPU)-capable compute server at the Department of Mathematics is now operational. The server is configured identically to every other compute resource at the department, offering the same software stack you are already familiar with. The machine consists of the following hardware:
- 2x Intel(R) Xeon(R) Gold 6226R CPU @ 2.90GHz (16 cores, 32 threads each)
- 768GB RAM
- 4x Nvidia A10 GPUs
Compatibility with older GPU-based software
The A10 GPUs were released in 2021 and older versions of software installed at the department may not support this hardware. Versions of PyTorch and TensorFlow from the fosscuda 2020b toolchain and onwards, such as PyTorch/1.9.0-fosscuda-2020b, are explicitly designed and configured to work with these cards. If you require the use of older versions of PyTorch, please use nam-shub-01 instead.