GaN and SiC energy semiconductors from AOS assist NVIDIA’s 800-VDC energy structure for next-gen AI infrastructure, enabling knowledge facilities to deploy megawatt-scale racks for quickly rising workloads. Transferring from typical 54-V distribution to 800 VDC reduces conversion steps, boosting effectivity, chopping copper use, and enhancing reliability.
The corporate’s wide-bandgap semiconductors are well-suited for the facility conversion levels in AI manufacturing unit 800‑VDC architectures. Key system roles embrace:
- Excessive-Voltage Conversion: SiC gadgets (Gen3 AOM020V120X3, topside-cooled AOGT020V120X2Q) deal with excessive voltages with low losses, supporting energy sidecars or single-step conversion from 13.8 kV AC to 800 VDC. This simplifies the facility chain and improves effectivity.
- Excessive-Density DC/DC Conversion: 650-V GaN FETs (AOGT035V65GA1) and 100-V GaN FETs (AOFG018V10GA1) convert 800 VDC to GPU voltages at excessive frequency. Smaller, lighter converters free rack area for compute assets and improve cooling.
- Packaging Flexibility: 80-V and 100-V stacked-die MOSFETs (AOPL68801) and 100-V GaN FETs share a standard footprint, letting designers steadiness price and effectivity in secondary LLC levels and 54-V to 12- V bus converters. Stacked-die packages increase secondary-side energy density.
AOS energy applied sciences assist understand some great benefits of 800‑VDC architectures, with as much as 5% increased effectivity and 45% much less copper. In addition they cut back upkeep and cooling prices.
The put up AOS gadgets energy 800-VDC AI racks appeared first on EDN.
