"The cross-platform comparison adds that a conventional workstation GPU generally provides lower latency, whereas Loihi 2 offers a substantial dynamic-energy advantage across the representative regimes considered here."
The main result is that convolutional LCA on Loihi 2 is feasible but highly regime-dependent. The Loihi-only analysis shows that the quality–efficiency tradeoff depends systematically on λ, stride, filter size, and target regime, rather than collapsing to a single operating point.
"The practical implication is therefore not that Loihi 2 uniformly outperforms conventional hardware, but that it occupies a different part of the design space. (...)"
By implementing ITD-aware thermal grouping of CPUs and inlet temperature adjustments, data center operators can optimize facility-level cooling and overall sustainability.
For the first time, researchers characterized inverse temperature dependence on production Intel Xeon CPUs and demonstrated that efficiency-optimal temperatures are CPU part-specific, and frequently higher than typical data center operating conditions.
arxiv.org/pdf/2606.11163
The case study investigated showed that this approach can reduce total data center energy by 4 – 13% without sacrificing performance or reliability.
Underfox
Underfox
Underfox
Underfox
Underfox
Underfox
In this paper, researchers have proposed a Loihi 2 implementation of convolutional sparse coding via the Locally Competitive Algorithm (LCA) and evaluate it against a conventional GPU baseline on the same inference problems.
arxiv.org/pdf/2606.08584
