The NUC Wasn't the Problem. The Power Strip Was.
The NUC Wasn't the Problem. The Power Strip Was.
WHERE IT STARTED
I spent my Saturday dissecting an Intel NUC.
Not because I planned to. Because it kept shutting down — every 15 minutes, no warning, no gradual degradation, no error message. Just off. And when your firewall goes down, your entire network goes with it.
I needed to know if the hardware was salvageable before I committed it to anything. So I ran a stress test, took it completely apart, cleaned it, replaced the thermal paste, and traced the actual problem to its source.
This post documents the full process — what I found, what I did, and what I decided afterward.
The Machine.
WHAT I WAS WORKING WITH
Intel NUC D34010WYKH. This is a 2014 fourth-generation NUC — one of Intel's early compact PC kits designed for home theater and small office use. On paper it's modest by today's standards, but it's built to run continuously and takes up almost no space.
Full specs:
- Processor: Intel Core i3-4010U — 1.7 GHz, dual-core, hyperthreaded (4 logical cores)
- RAM: up to 16GB DDR3L 1333/1600 MHz (two SO-DIMM slots)
- Graphics: Intel HD Graphics 4400
- Storage: 2.5" HDD/SSD bay + mSATA slot
- Connectivity: 4x USB 3.0, Gigabit Ethernet, Mini HDMI, Mini DisplayPort, WiFi, Bluetooth
- Power: 19V, 65W DC adapter
- Dimensions: 116.6 x 112 x 51.5mm — fits in your hand
Mine was running OPNsense as the firewall for a 1,250 square meter property — house, offices, storage, store — managing 30+ devices at any given time. It ran 24/7. For years.
Then it started shutting down every 15 minutes. No warning. Just off. And with it went IP management, traffic rules, content filters, the kids' internet schedule — everything that the firewall controlled.
Step 1 — Stress Test Before Opening Anything.
IS THE HARDWARE STILL GOOD?
Before pulling a single screw, I needed a baseline. There's no point doing a full teardown on hardware that's already dead. The stress test tells you whether you're dealing with a hardware failure or something else.
The process is simple and requires nothing installed — just a bootable USB with Ubuntu Live. No changes to the machine, no installation, nothing permanent.
What you need:
- Ubuntu Desktop ISO (ubuntu.com/download/desktop)
- Rufus to write the USB (rufus.ie)
- USB drive — 8GB minimum
- Internet connection on the machine being tested
Creating the bootable USB in Rufus:
- Select your USB drive
- Select the Ubuntu ISO
- Partition scheme: GPT (important for UEFI machines like the NUC)
- Click Start — takes about 5-10 minutes
Booting the NUC from USB:
- Connect the USB, power on the NUC
- Press F10 repeatedly as it boots — this opens the boot menu
- Select the USB from the list
- On the Ubuntu screen, choose "Try Ubuntu" — not Install
Once Ubuntu loads, open Terminal and install the two tools you need:
sudo apt update && sudo apt install -y stress-ng lm-sensors
Reading baseline temperature before the test:
watch -n 1 sensors
This refreshes temperature readings every second. Let it run for a minute until the numbers stabilize — that's your idle baseline. Take a photo. That's your "before."
My NUC at idle:
- Package: 48°C
- Core 0: 46°C
- Core 1: 44°C
That's high for idle. Normal for a machine with 10+ year old thermal paste that ran 24/7 without maintenance. Not alarming — but confirming that the paste needed replacing regardless of what the test showed.
Running the stress test:
Open a second terminal — keep the first one running with sensors visible — and execute:
stress-ng --cpu 0 --timeout 30m --metrics-brief
--cpu 0 uses all available logical cores. --timeout 30m runs for 30 minutes. If you want a quicker initial test, 10 minutes gives you a solid read — that's what I ran first.
While it runs, watch the sensors window. You'll see temperatures climb within the first minute and then stabilize at a peak. That peak is what matters.
My results after 10 minutes at 100% CPU:
- Peak temperature: 64°C
- Temperature after completion: 50°C — came down on its own
- Passed: 4/4 cores — Failed: 0
64°C peak under full load, dropping back to 50°C on its own. Zero failures. The hardware passed. Whatever was causing the shutdowns, it wasn't a failing CPU or a broken board.
Step 2 — Taking It Apart.
THE TEARDOWN
Hardware confirmed good. Now it gets opened.
A NUC that's run 24/7 for over a decade without being opened needs maintenance regardless of test results. Dried thermal paste, accumulated dust, oxidized contacts — these are certainties, not possibilities.
What you need for the teardown:
- Precision screwdriver set — I use the JOREST 40-piece kit. Phillips, Torx, Triwing — everything you need for electronics in one compact set.
- Isopropyl alcohol 70%+ for cleaning old paste
- New thermal paste — a rice grain sized amount is enough
- JOREST 40-Piece Precision Screwdriver Set — Phillips, Torx, Triwing. Everything you need for electronics in one kit.
- Gunk NM1 Electric Contact Cleaner — Non-conductive, no residue. USB ports, power connectors, any oxidized contact.
- Compressed air or a soft brush for dust
- A clean, flat surface with good lighting
Tools Used in This Post.
WHAT I ACTUALLY USED
Opening the NUC D34010WYKH — step by step:
1. Power off completely and disconnect all cables. Not sleep — fully off, power adapter unplugged.
2. Flip the unit upside down. You'll see four rubber feet and four screws underneath them. The screws go through the bottom panel into the chassis.
3. Remove the four bottom screws. Keep them somewhere safe — they're small.
4. Remove the bottom panel. It lifts straight off once the screws are out. The 2.5" drive bay is part of this panel on the WYKH model — if you had a drive installed, it comes out with the panel.
5. The board is now accessible. You'll see the fan, the heatsink, the SO-DIMM RAM slots, and the mSATA slot. Everything is compact — work carefully and don't force anything.
6. Remove the fan connector. Small connector on the board — pull straight up, gently. Don't yank it by the wire.
7. Remove the heatsink screws. There are screws holding the heatsink/fan assembly to the board. Remove them in a cross pattern to release pressure evenly.
8. Lift the heatsink assembly. It may resist slightly if the old paste has hardened and is bonding the surfaces. Twist gently — don't pull straight up with force. Work it loose slowly.
9. This is what you'll find.
That gray, cracked, hardened material on the CPU is thermal paste that has completely dried out. When thermal paste dries, it loses its ability to transfer heat efficiently from the CPU to the heatsink. The CPU runs hotter than it should, thermal throttling kicks in, and in extreme cases the machine shuts down to protect itself.
Mine was almost solid. I needed to work carefully with isopropyl alcohol and a soft cloth to remove it without scratching the CPU surface.
Step 3 — Cleaning Everything.
THE MAINTENANCE
Removing the old thermal paste:
Apply isopropyl alcohol to a lint-free cloth or cotton swab. Work in small circular motions on both surfaces — the CPU and the heatsink contact plate. Take your time. You want both surfaces clean and shiny before applying new paste. Any residue left behind reduces thermal transfer.
Cleaning the fan:
The fan blades and heatsink fins accumulate dust over time, reducing airflow. Use compressed air or a soft brush to clean between the fins. On a machine this age, you'll find more than you expect.
Cleaning the USB ports and power connector:
This step matters more than people realize. Oxidized contacts on USB ports cause intermittent read errors — the kind that look like software problems but aren't. Spray contact cleaner into each USB port and the power connector. Let it dry completely before reassembly.
Applying new thermal paste:
Less is more. A single dot the size of a grain of rice in the center of the CPU. The pressure from the heatsink will spread it evenly when you reinstall. Don't spread it manually — let the heatsink do it.
Reassembly is the reverse of teardown:
- Heatsink back on — screws in cross pattern, snug but not overtightened
- Fan connector back in
- Bottom panel back on
- Four screws back in
Step 4 — The Actual Problem.
IT WASN'T THE HARDWARE
After reassembly, I looked at where the NUC had been installed. Old power strip. Bad contact at the socket. Visibly worn — the kind of thing you stop seeing because it's been there for years.
I replaced the entire power strip and moved the NUC to a clean, dedicated connection.
The machine that was shutting down every 15 minutes wasn't failing — it was starving. An old power strip with a bad contact was cutting the supply intermittently. The NUC was doing exactly what it should: shutting down cleanly rather than corrupting data mid-operation.
The USB instability was a combination of both problems — oxidized contacts on a machine that hadn't been opened in years, and the same unstable power causing read errors. After the contact cleaner and the new power connection, the USB ports work cleanly.
The fix cost zero dollars and less than an hour. Check the basics before you assume the hardware is dead.
What's Next for This Machine.
HONEST STATUS
The NUC is mechanically sound. The electrical problem is solved. The contacts are clean. The thermal paste is fresh.
What it still needs: storage. OPNsense requires a drive — running a firewall from USB is not acceptable. USB drives aren't designed for the constant read/write cycles a firewall generates. The drive would fail in months. I'm waiting on an SSD before putting OPNsense back on this machine.
And even then — I've made a decision about what role this NUC will play going forward.
A firewall is critical infrastructure. If it goes down, the entire property loses network control instantly. That's not a job for hardware I'm still rebuilding confidence in. The OptiPlex 780 will take OPNsense — more thermal headroom, purpose-built for continuous operation, and stress-tested before deployment.
The NUC earns a secondary role: services that can tolerate an occasional restart. Pi-hole, Wireguard, Immich, Jellyfin. Useful, always-on, but not mission-critical.
A machine doesn't have to be perfect to be useful. It just has to be honest about what it can and can't handle. This NUC has earned a place in the homelab — just not the one I originally planned for it.
What This Actually Taught Me.
THE REAL LESSON
Before opening the NUC, I assumed hardware failure. A 2014 machine running 24/7 for over a decade — it made sense to suspect the CPU, the board, the fan.
The stress test said otherwise. The hardware was fine.
The real problem was an old power strip with bad contact. The fix cost nothing. The maintenance — thermal paste, contact cleaner, cleaning the fan — was overdue regardless and would have been done eventually.
Two things to take from this:
- Test before you assume. A stress test takes 30 minutes and tells you whether you're diagnosing hardware or environment. Run it before anything else.
- Check the power. Unstable power is invisible until it's a problem. Old power strips, worn sockets, overloaded circuits — these cause symptoms that look like hardware failures. It's always worth eliminating the obvious before opening the machine.
This post contains Amazon affiliate links. If you purchase through them, I may earn a small commission at no extra cost to you. Every product listed here is something I personally own and use daily.
RELATED POSTS
- My Network Has No Firewall Right Now. Here's Why That's Temporary. — Post #009
- 4 Dell OptiPlex 780s Rescued From the Trash. Here's My Plan. — Post #013
Comments
Post a Comment