⚡ Samsung and SK hynix use Nvidia AI to make chip simulations up to 1,600x faster
At Nvidia AI Day Seoul, Samsung said an AI workflow built on Nvidia's PhysicsNeMo cut a chip-level thermal and mechanical simulation from about 18 hours on a 32-core CPU to 41 seconds, roughly 1,600 times faster. SK hynix said its full-chip heat and deformation analysis now runs more than 30 times faster than on CPUs, and an AI model predicts etch-process shapes dozens of times faster than conventional simulation.
⭐ Faster virtual testing means fewer trial production runs, a real edge as memory makers race to build custom HBM for each AI chip customer.
#AI
At Nvidia AI Day Seoul, Samsung said an AI workflow built on Nvidia's PhysicsNeMo cut a chip-level thermal and mechanical simulation from about 18 hours on a 32-core CPU to 41 seconds, roughly 1,600 times faster. SK hynix said its full-chip heat and deformation analysis now runs more than 30 times faster than on CPUs, and an AI model predicts etch-process shapes dozens of times faster than conventional simulation.
⭐ Faster virtual testing means fewer trial production runs, a real edge as memory makers race to build custom HBM for each AI chip customer.
#AI
