Nexus Compute

Data Science Workstation

High-core-count, high-memory compute for analytics, modeling, and large in-memory datasets.

Overview

The Data Science Workstation prioritizes CPU cores and memory capacity for analytics teams whose work is often more data- and CPU-bound than GPU-bound. Nexus Compute configures the platform around large in-memory datasets, with optional GPU acceleration where your models benefit. This is the configuration we recommend when the honest answer is that a GPU would not help much. A great deal of production analytics (joins, aggregations, feature engineering, gradient-boosted trees, simulation) is limited by memory capacity and core count, and buying a large GPU for that work is money that would have bought more useful RAM. The rule of thumb worth applying is that a dataset which fits entirely in memory can be worked interactively, and one that does not forces you into chunking, spilling to disk, or a distributed framework, all of which cost analyst time on every iteration. So the first sizing question is how large your working dataset is, uncompressed and in memory, which is usually several times the size of the file on disk. Where GPU acceleration does help, such as deep learning or some dataframe and graph libraries, we add it deliberately rather than by default.

Warranty & Support