Search for a command to run...
No comments yet. Be the first to comment.
Every team has one complaint that never becomes a proper task. Someone says it, everyone nods, nothing happens. One of our complaints was "the VPN is slow sometimes, or the VPN feels too slow sometime
Why your reactive state fails structured clone, and the fix I shipped.
I built a recipe keeper called Recipe Jar. You paste a recipe link, it gives you a clean card with just the ingredients and steps, and you can save as many recipes as you want. No account, no ads, wor
Let me tell you about this week(Obv. mine). I pressed the power button on our office’s Ubuntu server. Fans started spinning. CPU light turned on. Everything sounded normal. But the monitor? Completely black. Nothing. Not even the BIOS screen showed u...
My Ubuntu server kept freezing. No warning. No error. Just completely stuck. Screen frozen, keyboard dead, only option was to hold the power button and force restart.
This happened randomly. Sometimes after a few hours, sometimes after a few days. I couldn't find a pattern. I checked RAM, disk, GPU, drivers, Docker, everything. All looked fine.
Turns out the problem was my CPU eating too much power. And the fix was literally one command.
My server runs on an Intel i9-14900K with a Gigabyte motherboard and an NVIDIA RTX 4090. It handles PostgreSQL, Docker containers, Python scripts, and some web scraping workers. Nothing crazy. But every now and then, when multiple things run at the same time, the whole system just freezes.
No logs. No crash dump. No kernel panic. Just dead.
Every Intel CPU has two power limits built in. Think of them as rules that tell the CPU how much electricity it's allowed to use.
PL1 (Power Limit 1) is the long-term limit. This is how much power the CPU can use continuously. All day, every day. Intel says for the i9-14900K, this should be 125 watts.
PL2 (Power Limit 2) is the short-term boost limit. When you suddenly need more power (compiling code, running heavy queries), the CPU is allowed to jump up to PL2 for a few seconds temporarily. Intel says this should be 253 watts. After a short burst, it must return to PL1.
So the normal behavior should look like this:
Normal load: 30W ... 50W ... 40W ... 60W
Sudden spike: 30W ... 50W ... 253W! ... 253W! ... back to 125W ... 60W ... 40W
^^^^^^^^^^^^^^^^^^^^
PL2 boost (few seconds only)
The CPU goes fast when needed, then calms down. That's healthy.
I checked my power limits with these commands:
PL1=$(cat /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw)
echo "PL1: $((PL1 / 1000000))W"
PL2=$(cat /sys/class/powercap/intel-rapl:0/constraint_1_power_limit_uw)
echo "PL2: $((PL2 / 1000000))W"
Result:
PL1: 253W
PL2: 253W
Both set to 253W. That means there is no "calm down" phase. The CPU is allowed to pull 253 watts forever if the workload demands it. No speed limit. Full power all the time.
This is a known issue with Gigabyte motherboards (and ASUS, MSI too). They ship with power limits set to maximum by default. Why? Because higher power means higher benchmark scores. Looks great in reviews. Not so great for stability.
Imagine a water pipe rated for 125 liters per minute. Someone removed the limit and now 253 liters can flow through it. Most of the time you only use 10-20 liters, so nothing happens. But when many taps open at once, 253 liters rush through. The pipe can't handle the pressure and bursts.
Same thing with the CPU. At idle, my server uses about 2-3 watts. Normal workload is maybe 30-60 watts. No problem. But when PostgreSQL runs a heavy query while Docker containers are scraping websites and Python scripts are processing data, all at once, the CPU pulls 200+ watts sustained. The chip gets extremely hot, the motherboard power delivery gets stressed, and the whole system locks up.
Intel actually acknowledged this problem for 13th and 14th gen CPUs. They released microcode updates and told everyone to set PL1 back to 125W. Many people were having crashes and didn't know why.
Step 1: Check your current power limits
PL1=$(cat /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw)
echo "PL1: $((PL1 / 1000000))W"
PL2=$(cat /sys/class/powercap/intel-rapl:0/constraint_1_power_limit_uw)
echo "PL2: $((PL2 / 1000000))W"
If PL1 shows anything above 125W, that's your problem.
Step 2: Fix it right now (instant, no reboot)
echo 125000000 | sudo tee /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw
That's it. One command. PL1 is now 125W. Your CPU will still boost to 253W for short bursts (PL2), but it won't stay there. It will always come back down to 125W.
Step 3: Verify it worked
PL1=$(cat /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw)
echo "PL1: $((PL1 / 1000000))W"
Should now show 125W.
Step 4: Make it permanent
The command above resets on reboot. You have two options to make it stick.
Option A: Fix in BIOS (recommended)
Reboot, enter BIOS (usually Del or F2 on Gigabyte boards), and look for one of these:
"Package Power Limit 1" or "PL1"
"Long Duration Power Limit"
Usually under "Tweaker" or "Advanced CPU Settings"
Set it to 125. Save and exit.
Option B: Set it automatically on every boot
Create a systemd service:
sudo bash -c 'cat > /etc/systemd/system/cpu-power-limit.service << EOF
[Unit]
Description=Set CPU PL1 Power Limit to 125W
After=multi-user.target
[Service]
Type=oneshot
ExecStart=/bin/bash -c "echo 125000000 > /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw"
[Install]
WantedBy=multi-user.target
EOF'
sudo systemctl enable cpu-power-limit.service
Now every time the server boots, PL1 gets set to 125W automatically.
Step 5: Keep Intel microcode updated
Intel released microcode patches specifically for this issue. Make sure it's installed and won't get auto-removed:
sudo apt install intel-microcode
sudo apt-mark manual intel-microcode
No. My server uses 2-3 watts at idle and maybe 30-60 watts during normal work. The 125W limit is more than double what I normally use. The only thing that changes is the CPU can't sustain 253W for minutes at a time anymore. For a server running databases and containers, you will never notice the difference.
Server kept freezing randomly. No errors, no logs, just dead. Turned out the motherboard had CPU power limits set way too high (253W sustained instead of Intel's recommended 125W). Under heavy load, the CPU would pull too much power for too long and the system would lock up.
Fixed it with one command:
echo 125000000 | sudo tee /sys/class/powercap/intel-rapl:0/constraint_0_power_limit_uw
Then made it permanent in BIOS.
If you have a 13th or 14th gen Intel CPU and your system freezes randomly, check your PL1. There's a good chance your motherboard set it way too high.
Thanks!
Happy Debugging!!😀
Moral of the story: sometimes the fix isn't in the software. Sometimes your motherboard is just being too generous with the power, and your CPU doesn't know when to stop.