Without an efficient way to extract additional computing power from existing infrastructure, organizations are often forced to purchase additional hardware or postpone projects. This can lead to longer wait times for results and possible losses compared to competitors. This problem is further exacerbated by the rise of AI workloads that require high GPU compute loads.
ClearML has come up with the perfect solution to this problem, according to it: fractional GPU capabilities for open source users, making it possible to “split” a single GPU so that it can run multiple AI tasks simultaneously.
This move is reminiscent of the early days of computing, when mainframes could be shared between individuals and organizations, giving them the ability to utilize computing power without having to purchase additional hardware.
Fractional capabilities for Nvidia GPUs
ClearML says the new feature enables DevOps professionals and AI infrastructure leaders to divide their Nvidia GTX, RTX and data center-grade, MIG-capable GPUs into smaller units to support multiple AI and HPC workloads, allowing users can switch between small R&D tasks and larger, more demanding training jobs.
The approach supports multi-tenancy and provides secure and confidential computing while limiting hard memory. ClearML says stakeholders can run isolated parallel workloads on a single shared computing resource, increasing efficiency and reducing costs.
“With our new free offering supporting fractional capabilities for the widest range of Nvidia GPUs than any other company, ClearML is democratizing access to computing as part of our commitment to helping our community build better AI at any scale, faster,” says Moses Guttmann, CEO and co-founder of ClearML. “We hope that organizations that have a mix of infrastructure can use ClearML and get more out of the computing power and resources they already have.”
The new open source fractional GPU functionality is available for free at ClearML’s GitHub page.