Unusual Temperatures and Power Consumption
Unusual Temperatures and Power Consumption
I'm reaching out because I've never encountered this situation before. I'm noticing unusual wattage and temperature patterns on my CPU. My setup includes a 5950x on an Asus B550-I STRIX with a Nactua NH-D15s, running in a Lian Li TU-150. Last night I had to power down my machine due to stability problems while playing a game. In Ryzen Master, the temperature reached 90°C. Today I ran benchmarks using Furmark's CPU burner. With a 32-core load, I observed around 125 watts and temperatures under 70°C according to OpenHardwareMonitor. It's noticeably warmer than when idle. When switching to an 8-core load, the readings were about 145W with temperatures between 88-89°C. Below that core count, power consumption drops and heat output decreases. It appears the processor draws more power when using fewer cores. I intend to repaint now, as my last attempt didn't resolve the issue. The results suggest increased power usage at lower core loads. In Ryzen Master, certain cores are prioritized and receive more workload. I'm sharing this because it's unusual—especially under lower core usage. I'm planning automated tests to explore how different wattages and thermal behavior affect performance across thread counts. If I've overlooked something or if this is common, I'd appreciate any insights or advice. My work often involves heavy multitasking with many cores, so this issue stands out. This scenario seems rare, but it tends to occur during gameplay when fewer cores are active.
Not necessarily. For one, in an all core load your cores are running at a lower frequency and thus lower voltage, which makes them more efficient. During an limited core load, the individual cores are allowed to consume far more power to eek out just a bit more performance. On the other hand, you might also run into other bottlenecks. Depending on your load, your CPU might be limited by your RAM. This would mean, your CPU cores would just wait for the next instruction to load, which consumes less power overall. If only 8 CPUs need to work on the task, the overall computation speed is lower, so you might not be limited by the RAM anymore. There can also be some interesting effects with the specific design of the CPU. If you have a lot of data sharing between the CCDs, you might end up having a better time just running the same application on 8 cores, that are on the same CCD. This is all speculation though.
This is a purely synthetic load intended as a burn-in test. I don't think RAM speed has much to do with it. Maybe shared cache could play a part. I tried to run separate processes completely (instead of separate threads, launch the burn in test multiple times to ensure they are independent) and had similar results. If I write a benchmark for this I'll write something that can run within cache and something that uses more memory. I'd like to see if others see similar results on closely related CPUs.
Well, then you can simply verify if it's the initial one. For my 5900X, I hit the EDC/TDC (current) limit when doing a full core load and the PPT (power) limit when using a 6C/12T setup. Is this the same for you? (Reminder: someone needs to quote your response for notification.)
First, remove Ryzen Master and reset the CMOS. Don’t reinstall it. Also take out Fumark and don’t reinstall it—both are considered harmful. If you need anything else, let me know. Hope this assists and good luck!
It seems you're questioning how an all-core load without external constraints affects CPU performance. It does seem unusual. You might want to share a screenshot of the power monitoring section in Ryzen Master for reference.