CPU throttles to low frequency (1800 MHz) under load despite not actually overheating (50 °C)

Question

Solution

I solved the problem, thanks to the answer below. It was indeed the VRM heating up over 80 °C and it throttled CPU. CPU was not heating up, but motherboard made it look like it was so that it could throttle it. The solution was to cool down VRM. It could be solved through a couple ways:

• Apply heatsinks to MOSFET VRM if your board does not have them already.
  Heatsinks: https://www.newegg.com/Product/Product.aspx?Item=N82E16835708012
  Video demonstrating: https://www.youtube.com/watch?v=T20YIimV8vY
• Provide better airflow through your computer case, which can be achieved by:
  • Cleaning your case and your fans
  • Better cable management
  • Smarter case customization
  • More fans if really needed (usually unnecessary, better placement is more important)
  • Some case chassis have places to add fans on the side, if not, you can get creative
• Buy a new part:
  • New case
  • Better fan(s)
  • BETTER MOTHERBOARD!
• Additionally, you can make it so that the CPU fan blows over VRM. Cannot guarantee it'll work, because CPU fan will be exhausting hot air from CPU anyways, so it might even worsen. I didn't try, so just to let you know you might try it.

I had a old CPU fan, so I just installed it on the side of the case chassis, and it is blowing directly towards VRM. It is not exhaust but intake. Dropped my VRM temperatures by 10-20 °C. Downside to this is that now there is more noise coming while playing. It is not a big problem though because it is auto-adjusting its speed if there is no pin left out, if there is one, you still can do it via a program called SpeedFan. I plugged the CPU fan to System fan pins on the motherboard because there were no other pins. One is left out which is PWM pin. It doesn't auto adjust but I set it to 0% from SpeedFan while it is not necessary to use.

Problem definition:

CPU dropping its frequency down to 1800 MHz.

What do I know about it?

• Seems to occur when the CPU reaches 50 °C. I am not certain if the temperature is the cause.
• When it happens, the temperature jumps to 206 °C and the frequency drops to 1800 MHz.
• Only happens while playing games.(so far)
• It is not related to my CPU being overclocked because it happens if it is not, as well.
• The only solution I've discovered so far is restarting the computer.

Video

Please watch the following 15 seconds video that I captured from CPUID HWMonitor, it might prove to be useful to understand what's happening.

https://www.youtube.com/watch?v=2BEjC2OZIjE&feature=youtu.be

Specs

CPU: AMD FX-6100
Motherboard: MSI 760GM-P23
GPU: NVIDIA GeForce GTX 1060 3GB

Additional Notes

• It occurs the most frequently while I'm playing Tom Clancy's Ghost Recon® Wildlands.
• I've used CPUID HWMonitor to monitor the temperatures. I am not very certain if the program is accurate.

Important: I know there is something called throttling, CPU or motherboard does it to protect itself from harm, but 50 °C seems to be a safe temperature. I know other people using the same program as I have see temperatures higher than 80 °C.

Question

What are the possible causes for it to happen like this? How can I fix it? Also, any knowledge regarding the matter is appreciated, even if it does not solve the problem directly.

Answer 1

It looks like your motherboard's voltage regulator module (VRM) is overheating. The VRM is responsible for converting 12V power from the power supply into the low voltages used by the processor. Because it needs to convert large amounts of power, it can get very hot under load. On this motherboard, the VRM temperature is listed as TMPIN1 in HWMonitor.

While most quality motherboards have at least 6 VRM phases, this board has only a 3+1-phase VRM, which means it has relatively low power capacity. To make matters worse, there is no heatsink on the VRM as would typically be present on better motherboards. This means that the VRM on your board can overheat under sustained load. To prevent damage, the motherboard will report a fake CPU temperature to force it to throttle down, reducing the load on the VRM. The GIF in your post shows the VRM temperature hitting 90 °C before immediately dropping by several degrees when the processor is throttled with the fake temperature.

You may be able to address this problem by using a CPU cooler with a fan that blows towards the board instead of a tower cooler that blows air horizontally through a heatsink. Although these coolers tend to have lower performance, they provide airflow for the VRM, which can prevent it from overheating. Alternatively, adding case fans that blow towards or across the VRM can help.

The VRM on your motherboard is highlighted in green in this image:

More information can be found in this MSI forum thread. You might want to read this Gamers Nexus article for technical information on the design and operation of VRMs.

---

VRM overheating is not limited to cheap motherboards. In fact, at the far high end, with the advent of consumer high-core-count processors like the Intel Core i9 series (Skylake-X, based on server-grade Skylake-SP silicon), some processors can draw so much power that even the most advanced VRMs can overheat when the processor is overclocked. In the linked article, the VRM tested has the equivalent of 12 phases, with 10 for CPU power (5+1 with phase doubling), yet it still overheats under extreme overclocks, highlighting the need to keep the VRM cool under heavy loads.