How to check a QNAP NAS for CPU stability and overheating

CPU Lab in StorageDiag Workbench loads the NAS processor in consecutive rounds without pauses. A run stops by itself once the score stops changing, after about five minutes at the latest. It records the peak CPU temperature and the drop in core clock speeds. BlackBox records temperatures and fan speeds. Together, these readings help you assess the cooling and identify what needs further investigation.

Preparesign in as an administrator and choose a quiet period for the NAS
Run CPU Labwatch the CPU load
Open BlackBoxmonitor temperature sensors and fan speeds
Read the resultcheck the verdict, peak temperature and clock speed
Repeatrun consecutive tests and use Lab mode if other processes interfere

01Before you start

CPU Lab, Pulse and BlackBox are on the Utilities desktop and monitor the NAS on which StorageDiag is installed. Sign in as an administrator. Pulse and BlackBox are available only to administrators.

Choose a quiet period. CPU Lab runs its worker threads at real-time priority. Shares, apps and virtual machines may respond slowly or time out during a run.
Stopping a run. During the first Run after opening the window, the button reads Running and cannot stop the test. Later runs in the same window can be stopped with Stop. Closing the window ends a run, including the first one.
Publishing. By default, a result with the verdict clean or throttled is published on the public CPU board, along with the NAS model, processor and memory. To keep your results off the board, open Compare and untick publish automatically before clicking Run. This choice is saved in the browser, so untick it in each browser you use. A Lab run (section 06) always publishes results with these verdicts.
A greyed-out Run button. When the StorageDiag daemon is set to Safe mode or Locked down, Run and Lab are greyed out and the window displays a notice. An administrator can change this level in Access management, on the Daemon tab.
History needs no setup. BlackBox records the NAS state every 20 seconds from the moment the package starts, and temperatures and fan speeds once a minute.

02Run the test

Open CPU Lab: Start menu → Benchmarks → cpuhound, or click the desktop icon of the same name. The Processor card shows the CPU model, thread count, maximum clock speed and frequency governor.
Click Run in the title bar. Each round performs a set of compute tests. The table lists the eight tests used to calculate the index. Each test runs on one thread first, then on all logical CPUs available to the test. Rounds follow without pauses. The run ends when the last four rounds agree within 1.5 %, when five consecutive rounds no longer reduce the spread, or after about five minutes.
Watch the CPH index after each round and the status line showing the round number and spread (round 8 · ±8.6 % → ±1.5 %). The table highlights the test in progress, and the Data computing tiles show the individual logical CPUs.

Figure 1. CPU Lab during a run (values from a recorded run). Round 8 on a NAS with a 12-thread Core i5-1235U. The index is still changing: the status line shows the spread of recent rounds against the 1.5 % target.

The pink contended badge and red Wait cells mark a test that spent more than 5 % of its time waiting for CPU time. With the default real-time scheduler settings, the kernel reserves about 5 % of each second for other tasks, so these marks are common. A run receives the contended verdict only when the wait exceeds about 10 % over the last four rounds. The run in figure 1 ended with a throttled verdict (figure 3).

When the processor reduces its clock speed as temperatures rise, the index falls over the first few rounds and then levels off. In the run in figure 1, the first round scored 9 762. From the sixth round onwards, scores ranged from 7 160 to 7 808. The settled score shows the performance sustained under those operating and cooling conditions.

On a NAS with four logical CPUs or fewer, CPU Lab leaves one available for the system. Its tile stays dark, and the test itself does not use 100 % of the total CPU capacity.

03Watch temperatures and fan speeds

CPU Lab displays its temperature reading only when the run ends. To follow the sensors and fans during the test, open BlackBox alongside it:

Open BlackBox: Start menu → Monitoring → BlackBox.
In the title bar, choose Charts, the 1h range and the Absolute scale. This scale starts the temperature and fan-speed axes at zero.
Click Temp in the title bar to display the temperature sensors together (up to twelve).
To include fans and CPU load, open the series picker using the tab at the left edge. In the Thermal group, tick the Temperature row and click Fan speed. Each fan has its own button on a NAS with several fans. In the CPU group, click sys in the CPU busy row to show overall CPU usage. Each unit of measure (°C, rpm, %) gets a separate panel with a shared time axis, as in figure 2.

Figure 2. BlackBox during three CPU Lab runs (example data from a NAS with a four-core, eight-thread CPU). The package and core sensors climb to about 76 °C and level off. The enclosure temperature reported by system1 rises by a few degrees, and the fan speeds up as the temperature rises.

Sensors. package and core0, core1, … are Intel CPU sensors. An AMD processor reports tctl. When the kernel exposes no processor sensor, the CPU temperature reported by QNAP appears as cpu. system1 and similar labels identify enclosure sensors reported by QNAP. A network card may add its own sensor. Fans have numeric labels; a CPU fan is labelled cpu1.
One point per minute. BlackBox reads temperatures and fan speeds once a minute, so a short run produces only a few points. Section 05 explains how to extend the load and collect more readings.
Heat map. Heat map in the title bar displays each sensor as a row of coloured cells: amber from 65 °C and red from 80 °C.
Events. BlackBox marks Critical temperature on the timeline when a core, the package or cpu reaches 90 °C, an enclosure sensor reaches 60 °C, or any other sensor, including AMD's tctl, reaches 85 °C. It marks Fan stopped when a fan reported as present reads 0 rpm.
Live clock speeds. Pulse (Start menu → Monitoring → Pulse, CPU tab) refreshes each logical CPU's readings every two seconds: Busy, Current clock speed (⚡ above the base clock), Min, Max and Governor. A substantial clock-speed drop under sustained load may indicate a performance limit. Assess it alongside the temperature and governor setting.
Kernel log. Start menu → Live logs → Kernel Log opens Kernel Log · LIVE, which displays new kernel messages as they arrive. A line such as Package temperature above threshold, cpu clock throttled is marked with the CPU Thermal Throttle pattern. A [Hardware Error] line reports a hardware error from the processor or firmware, usually a machine check. Decode it in MCE Decoder (Start menu → Analysis).

04Read the result

When the run ends, the CPH index card shows the verdict and number of rounds. A Thermal line appears on the Processor card, showing the peak CPU temperature and the lowest sustained core clock speed as a percentage of that core's peak. CPU Lab and BlackBox each use their own temperature thresholds, the same for every processor. The processor's operating limit depends on its model and the NAS design. A reading of 90 or 95 °C alone does not establish overheating. Compare the temperature with the CPU and NAS specifications, clock-speed changes and kernel log messages.

Figure 3. The result of the run in figure 1. throttled · 14 rounds, index 7 589 (±5.2 %), and Thermal: 98 °C peak · clock 90 %. The processor exceeded CPU Lab's 95 °C threshold and sustained 90 % of its peak clock speed. The result was published, so the card shows its position on the board.

The verdicts of CPU Lab
Verdict	What the run measured	On the board
`clean`	Every core sustained at least 90 % of its own peak clock speed, the CPU stayed below 95 °C, and the wait for CPU time stayed below the contended threshold.	yes
`throttled`	A core's clock speed fell more than 10 % below its peak, or the hottest CPU sensor reached 95 °C. The score represents CPU performance under those conditions.	yes
`contended`	Other processes occupied the CPU, leaving the test waiting for CPU time. The score is a lower-bound estimate of performance.	no
`untrusted`	Other processes competed for CPU time while the clock speed fell or the temperature threshold was reached.	no
`indicative`	The score did not settle: even the four closest consecutive rounds differed by more than ±5 %. Only a run that would otherwise be clean receives this verdict.	no

Heat or a clock-speed limit. A throttled verdict can result from temperature or a drop in clock speed. If the peak is above 90 °C, as in figure 3, check the cooling and compare the peak with that processor's specifications. A clock-speed drop with a peak in the 60s or 70s suggests checking power limits and the governor.
Check the temperature even on a clean run. clean allows a peak below 95 °C. A run peaking at 90 °C can therefore be clean according to CPU Lab while also reaching BlackBox's Critical temperature threshold.
CPU diagram. An amber outline indicates throttling; a pink outline indicates waiting for CPU time. The Wait column turns red above 5 %. With the default real-time scheduler settings, part of that wait may be the kernel's own reserve (section 02).
Index. The CPH index gives a single-thread score and an all-core score. On this scale, a single thread of a current mid-range NAS processor scores about 1 000. The ring beside it shows the precision reported with the score.

05Keep the CPU under load for longer

A run lasts from about one to five minutes. With effective cooling, the score may settle within four to six rounds. To observe temperatures for longer, start each run as soon as the previous one finishes:

When the verdict appears on the CPH index card, click Run again. Three runs on a NAS that heats up may take about a quarter of an hour, as in figure 2. If the score settles quickly, repeat the test until BlackBox shows ten minutes of load.
Watch BlackBox. With effective cooling, CPU temperatures level off and the fan maintains a higher speed. If temperatures are still rising at the end of the last run, the system has not yet reached thermal equilibrium. Check whether clock speeds are falling or the model's temperature limits are being exceeded.
Open Compare, then the This unit tab. It lists the last 40 runs for this NAS, with the date, type (Run or Lab), verdict and index. Select a run to compare its all-core test results with the latest other run made with the same build of CPU Lab's test program. A run from a different build is marked other build. Similar scores across runs indicate stable performance under the test conditions. Falling scores accompanied by rising temperatures call for a cooling check.

06Test with services stopped (Lab)

A contended or untrusted verdict means other processes limited the CPU time available to the test. Lab runs the benchmark with services stopped:

Close every QTS or QuTS hero tab in the browser. An open panel can restart services.
Click Lab. The panel explains the sequence, recorded data, estimated duration and preflight checks. It also lists the services that will stop: file sharing, App Center packages, scheduled jobs, media, snapshots, replication and antivirus. StorageDiag, the admin panel and helpdesk remain running.
Lab will not start while a virtual machine is running, an iSCSI session is connected, or an array or pool is degraded, rebuilding, resyncing or scrubbing. A failed check turns red and explains the reason.
Select the confirmation and click Start Lab run.

Figure 4. The Lab panel before a run (example services and checks). Estimated time: one to three minutes to stop services (at most 30 minutes), up to five minutes for the test, and one to three minutes to restart services.

Lab sets the governor to performance, stops services and waits 30 seconds for the CPU to settle. It then runs the same test. Afterwards, it restores the previous governor and restarts services. A Lab run continues on the NAS after you close the browser. Cancel ends the test and restores services. When it finishes, check that your shares and apps are working again.

07Assess the symptoms

What the run and BlackBox show, and what to do
What you see	How to interpret it	What to do
clean, a peak well below 90 °C, stable temperatures and a fan maintaining a higher speed	The cooling handles the load in this test.	Keep the result for later comparisons.
throttled with a peak of 95 °C or more and fans at high speed	The CPU reached the temperature threshold used by CPU Lab. The clock figure shows the sustained clock speed.	Compare the temperature with the model's limits and check the cooling (section 08). Repeat after addressing any problems.
throttled with a peak in the 60s or 70s	A power limit or the governor may be lowering the CPU's clock speed.	Check Governor in Pulse. Lab uses `performance`.
A fan at 0 rpm, or Fan stopped	A possible problem with the fan, its connection or the speed reading.	Stop the test. Check whether the fan is turning and properly connected. Replace it if a fault is confirmed.
A fan does not speed up under load	The settings and fan operation need checking.	Increase the fan speed (section 08). If the fan does not respond, check it before replacing it.
Critical temperature recurs after cleaning at a normal room temperature	A possible cooling problem, such as a worn fan or poor heatsink contact.	Check the model's temperature limits. If the symptoms indicate a cooling fault, contact a repair service.
The NAS restarts during a run or remains unresponsive afterwards	Further diagnosis is needed. Areas to check include the power supply, motherboard, CPU, memory and temperature.	Read the kernel log and BlackBox, test the RAM, then report the problem with these results.
`[Hardware Error]` lines appear in the kernel log during the run	The processor or firmware reported hardware errors.	Decode the entries in MCE Decoder, which identifies the subsystem (memory, CPU or bus), and send the result to a repair service.
contended	Other processes occupied CPU time needed by the test.	Repeat in Lab mode (section 06).

CPU Lab does not stop when the temperature rises. It also measures performance with reduced clock speeds. The processor has its own thermal protection, but do not continue testing after detecting a cooling problem. Stop the run and do not start another if BlackBox shows Fan stopped or 0 rpm. Check the fan first. If the NAS hangs, investigate the cause before repeating the test. CPU Lab measures performance but does not verify the results of its calculations. A stable index therefore does not establish that the processor is fully functional. Calculation errors can come with application crashes, restarts and machine checks, so read the kernel log as well. Memory faults can cause similar symptoms. Test the RAM before attributing the problem to the processor.

08Cooling a NAS that runs hot

Dust. Remove dust from the heatsink, fans and air intakes.
Fans. Check their operation and response to load (BlackBox, Fan speed). If they remain slow while the CPU is hot, review the Smart Fan settings in QTS or QuTS hero and increase their speed: Control Panel → System → Hardware → Smart Fan.
Airflow. Leave space in front of the intakes and behind the exhaust. Keep the NAS out of closed cabinets and away from other hot equipment. Lower the room temperature if you can.
Heatsink. Have a repair service or someone experienced in NAS servicing reseat the heatsink or replace its thermal paste.
Retest. After each change, repeat section 05. Compare the peak temperature, fan speed and index with earlier runs.