Mainboard power delivery configurations
Mainboard power delivery configurations
A SOC is a system-on-chip that integrates various components like the CPU, memory controller, and more. It serves as a versatile solution, often used in APUs where it handles both processing and memory management efficiently. Its importance extends beyond just APUs—it's valuable for any system needing integrated performance and resource control.
The output filter relies, and it might involve two inductors, though for the MOSFETs, it's somewhat similar.
For future reference, you might consult datasheets for different VRM controller chips and DC-DC switching controllers. For instance, here’s a two-phase controller featuring integrated drivers: NCP5383 - ONSMS14374-1.pdf
A mosfet doesn’t operate at a single fixed frequency. Its peak performance depends on factors like gate capacitance, resistance, how quickly the gate can charge or discharge, and the presence of doubling circuits. On motherboard VRMs, typical speeds reach up to 600 kHz, especially with double converters that allow two 300 kHz signals. Without these doubling stages, frequencies drop to 300–500 kHz. Higher speeds aren’t always better—they can lead to smaller components (like inductors and capacitors), but may reduce efficiency and increase noise, requiring more filtering. Stronger drivers are needed to efficiently push energy into the gate and quickly charge/discharge it. The model you referenced has a relatively high on-resistance (7–11 mΩ), which is typical for some packages, though it might not be ideal for mainstream VRM applications. Figures 7–10 suggest this design could help estimate optimal frequencies based on voltage and driver selection.
When selecting a mainboard, evaluate VRMs by checking both current capacity and voltage regulation features. Look at specifications like switching frequency, maximum current, and smart power stage capabilities. For instance, the CSD87350Q5D offers up to 1500kHz switching and 40A, while the ISL9927B supports 60A with SPS technology. The SM4503 can handle up to 80A. Compare based on MOSFETs and board layout for optimal performance.
Comparing VRMs isn't straightforward since many factors are involved. The peak frequency doesn't tell much because most VRMs operate efficiently between 300-600 kHz. Their power efficiency changes with frequency, often decreasing as speed increases—this means sacrificing some efficiency for the ability to use smaller components like inductors and capacitors. The peak current is also not a complete picture; VRMs distribute load across multiple phases, so a single device's total draw is spread out. For instance, a CPU drawing 120W at 1.2V with 100A means each MOSFET or power stage handles about 12.5A, well under typical limits. The actual heat generated depends on how the current is shared among phases. A high peak current alone isn't enough to judge performance; it's the overall thermal management and efficiency that matter. The Rds(on) value is crucial—it indicates resistance when current flows, with higher values leading to more heat and lower efficiency. Datasheets often show maximum ratings, but real-world use varies. For example, a certain chip might support 80A but needs special cooling or design tweaks to function reliably. Manufacturers balance size, cost, and thermal performance, so the right choice depends on specific requirements rather than just numbers.
The efficiency and output voltage matter, but if the heat output stays in a desirable range like 93% efficiency with solid cooling, what comes next is enhancing voltage quality or VRM performance. Doublers with load balancing and load monitoring between MOSFETs can help—when one MOSFET handles more current, they adjust to spread the load evenly. While traditional phases might seem simpler, advanced doubled configurations (like 4×2) often outperform average real phases (8). Similarly, high-quality doubled MOSFETs that monitor temperature and load are better than standard ones.