JP Wright

Regional Sales Manager

University of California, Santa Barbara BrainChip

Winchester, California, United States 500+ connections

1. JP Wright commented on this

   

   Kevin D. Johnson 3rd+

   Field CTO – HPC, AI, LLM & Quantum Computing | Principal HPC Cloud Technical Specialist at IBM | Symphony • GPFS • LSF

   3h

   Follow

   As promised, below you'll find a brief video of Symphony and Akida working together to demonstrate neuromorphic classification of real market data running just minutes ago right before market close.Why This MattersThe real story isn't one chip running one model. It's the architecture made possible by joining Symphony and Akida together.Symphony isn't legacy technology. It's battle-tested infrastructure running some of the world's largest trading grids. Leading investment firms operate Symphony clusters with hundreds of thousands of cores, processing millions of tasks per second. This is proven technology at institutional scale.What's changed is the hardware landscape. The compute capability that filled server rooms a couple of decades ago now fits on a $150 edge board. An Intel N100 with an Akida M.2 card delivers neuromorphic inference in microseconds while barely drawing 20-40 watts—versus hundreds or even thousands of watts for server-class hardware.Symphony now recognizes and capably manages Akida's neuromorphic resources alongside traditional compute. This isn't bolted-on integration—Akida nodes become known participants in the cluster, schedulable and governable through Symphony's resource framework. You could run multiple classifiers across a grid, or coordinate a distributed neural network that responds in force to market events. Symphony handles the orchestration.This opens two paths for large-scale trading grids:Path 1: Cost Reduction. A 500,000-core institutional grid typically dedicates a large part of its compute to classification—market regime detection, signal generation, risk classification, anomaly detection. These are high-volume, low-complexity workloads perfect for neuromorphic hardware. Replace classification cores with 15,000 Intel N100 edge nodes running Akida M.2 cards. Hardware cost: roughly $6 million. Power consumption drops by an order of magnitude. Five-year TCO savings measured in nine figures.Path 2: Capability Expansion. Keep all existing cores. Add 15,000 Akida edge nodes as a classification tier. Cores previously dedicated to classification are now freed for optimization, execution, and backtesting. A $6 million investment liberates over $100 million worth of compute capacity. You've doubled your analytical throughput.The Symphony/Akida AdvantageSymphony and Akida make this work together through neuromorphic hardware and resource abstraction. Applications request classification capacity; Symphony routes work to available neuromorphic nodes. No application changes required. The same scheduling, monitoring, and governance that handles CPU and GPU workloads now extends to Akida's neuromorphic silicon.This is heterogeneous computing done right: a dynamic compute platform that treats neuromorphic chips as capable, known resources—routing work to the right hardware automatically.Architecture validated. Symphony and Akida work together. Classification was the proof point, but not the only one.

   …more
   245 Comments2 Reposts
   
   

   CommentShare
   
   JP Wright

   Regional Sales Manager

   The IC used for this analysis, the AKD1000 operates at ~1W, our AKD1500 can run at 0.1W for SPI interface and 0.3W for PCIe interface, offering even more power saving and much less heat dissipation!