I'm trying to understand why my Core i5-14400F performs normally in single-core tests but loses power and performance during sustained multi-core loads. My system uses a Gigabyte B760M K V2 DDR4 motherboard, RTX 2060 Super, 16GB DDR4, a 550W PSU, and an ICE C612 V2 cooler running Windows.
The CPU scores about 569 in Cinebench 2026 single-core, but only around 3011–3046 in multi-core. Cinebench 2024 multi-core is around 720–732. During testing, the processor reaches roughly 4.69GHz on the P-cores and 3.49GHz on the E-cores, while temperatures remain near 75°C, so ordinary thermal throttling does not seem to be the cause.
The unusual behavior is that package power starts around 100–105W, then falls to approximately 60W for most of the sustained workload and sometimes drops even lower near the end. I changed the BIOS turbo limits from Auto to manual settings, including PL1 at 100–135W, PL1 time at 56 seconds, and PL2 at 148W, but the behavior did not change. I have not adjusted voltage or overclocked the processor.
What could be forcing the CPU down to about 60W? Could it be a current or EDP limit, another motherboard power setting, Gigabyte BIOS behavior, VRM protection or overheating, Windows power management, or the board's power-delivery design? What is the safest way to identify the active limit and allow the processor to sustain higher multi-core power?
2 Answers
A likely explanation is the motherboard’s VRM or current-delivery limit. This board has a fairly modest power section and may be designed to run a non-K processor at Intel’s standard settings, even if the BIOS lets you enter higher turbo limits. If the VRM becomes too hot or reaches a current limit, the firmware can reduce CPU package power despite the PL1 and PL2 values being higher.
Monitor the VRM temperature sensor in HWiNFO if the board exposes one, and watch the CPU throttling or limit-reason indicators during the benchmark. If no VRM sensor is available, cautiously checking heatsink temperature after a sustained run can provide a clue, but the software limit flags are more reliable. A current or VRM limit may appear even when the CPU itself is only around 75°C.
Run Intel XTU, or another monitoring tool that reports throttling reasons, while Cinebench is running. Look specifically for power-limit, current/EDP, thermal, or VRM-related flags; the package-wattage reading alone doesn’t identify which limit is active. Also consider whether the score actually affects your normal workloads—if real-world performance is fine, chasing a benchmark number may not provide much benefit.
How can I check whether the VRM is overheating or whether the board is triggering a current limit?