SLVAFH6 November 2023 LM25066 , LM5066 , LM5066I , TPS25984 , TPS25985 , TPS25990 , TPS536C9T
A typical server platform contains one or more processor units (CPU) as well as other sub-systems and is powered by one or multiple power supplies (PSU). In a broader sense, the PSUs are categorized as common redundant power supply (CRPS) and non-CRPS PSUs. CRPS PSUs have crisply defined peak power capabilities, whereas non-CRPS PSUs have no standard of behavior beyond their 100% rated power. The most critical case of PSU configurations is one where a single PSU is operational and that cannot handle all the projected peak power envelope demanded by the platform. Therefore, the total power drawn by the platform must be continuously monitored, and there must be a protection mechanism to throttle the power demanded by the platform quickly once a pre-programmed threshold is exceeded.
The system input power monitor, PSYS is designed to provide information necessary for the processor to dynamically allocate power between different system components, such as the CPU, memory, and the rest of the platform, within a designated power budget. This helps in optimal use of total available platform power by facilitating the dynamic employment of surplus system power to assist the platform turbo. In the absence of accurate system current monitoring, the platform designer sets the peak power limit too low, potentially settling for a too conservative power envelope. This results in sub-optimal utilization of available platform power.
PSYS_CRIT#™ is a must-have requirement in the case of non-CRPS generic PSUs to handle peak power events. Otherwise, these peak power events can, depending on their dwell time, shut down or damage platform power sources. Implementing the PSYS_CRIT#™ functionality enables the use of PSUs with generic specifications without making it mandatory to procure CRPS PSUs.