Radixia

NSCC Singapore ISC2026 HPC Around The World APAC Stream

SessionWednesday · 14:15–15:30 · Hall 4 - Ground Floor · ~1,605 words

Speakers: Gih Guang Hung (NSCC SINGAPORE)

Session summary

Gih Guang Hung of NSCC Singapore updates the ISC audience on Singapore's national supercomputing infrastructure and strategy in the HPC Around the World APAC stream. NSCC, funded by the National Research Foundation, anchors its vision on capacity and diversity, operational excellence, and talent development, and serves as the compute backbone for national initiatives including the National AI Strategy, AI for science, AI for health, and quantum, supporting localized multimodal LLM training and the Symphony medical foundation model program. The production fleet includes Aspire 2A (AMD CPUs and NVIDIA A100 GPUs, 10 petaflops peak), Aspire 2A Plus with 320 H100 GPUs, and on-premises systems for sensitive healthcare research. The newly launched Aspire 2B quadruples national capacity with 115 petaflops peak, over 180,000 CPU cores, 1,500 NVIDIA H200 GPUs, 66 petabytes of storage on Slingshot 400, and ranks 44th on the TOP500; a 98-qubit Quantinuum quantum system will be connected for hybrid quantum-classical computing. Hung describes shifting workloads from massive training toward inference services and fully agentic AI workflows combining training, inference, simulation, and analysis, prompting rethinking of scheduling, reservations, and token-based accounting. To support growing numbers of non-expert users, NSCC is developing an HPC Copilot vision including PerfPredict, an open-source resource-sizing tool that recommends job configurations from historical benchmarks and achieved over three times speedup in VASP production cases, alongside workload classification work with RIKEN.

Topics: national HPC infrastructure · AI supercomputing · hybrid quantum-classical computing · agentic AI workflows · resource sizing tools · HPC user enablement

AI-generated summary of an auto-generated transcript (~1,605 words in full). Details may be imprecise — verify against the session recording.

Auto-generated captions from ISC 2026 session recordings · transcription errors likely, verify quotes against the video · timestamps are offsets into each recording · independent tool, not affiliated with ISC · a Radixia Labs experiment