1. Current Location: Home >  NAS >  Is 52 degrees Celsius considered high for NVMe solid state in a NAS? I read the three sensors clearly before I came to a conclusion.

Is 52 degrees Celsius considered high for NVMe solid state in a NAS? I read the three sensors clearly before I came to a conclusion.

NVMe solid-state temperature monitoring in NAS: Three sensor readings, WCTEMP threshold and cooling three-layer account

Last month, I installed a 128G Laozhidao PC005 with a PCIe adapter card into the Feiniu small host at home to use it as a download cache disk. The transfer card article The installation process is written here. After installing it, I ran smartctl to check the quality of the disk, and the Temperature line in the output clearly said 52 degrees. I've been panicked when the mechanical disk turned 52 degrees. The article about 52 degrees and 58 degrees There are also special stalls. At that time, I had already reached into the shopping cart to add a cooling vest, but then I thought about it.: Are NVMe temperature accounts the same thing as mechanical disks? After checking it out, I discovered that I almost spent money unnecessarily and almost misunderstood the key points.

Find out first: The temperature of NVMe is more than one number

There is only 194 Temperature in the SMART of the mechanical disk._Celsius, the disk core temperature, can be explained clearly in one number. NVMe is not. knock smartctl -A /dev/nvme0 , the output is completely different from that of a mechanical disk. There is no number line like 194, and Temperature is followed only by "comprehensive temperature". If you want to see all the sensors, you have to use smartctl -x /dev/nvme0 , scrolling down there will be independent entries such as Sensor 1 and Sensor 2.

My PC005 has three readings: Composite is 52 degrees, Sensor 1 is 58 degrees, and Sensor 2 is 41 degrees. I was confused at first sight - the same chart, but the three numbers were more than ten degrees different, which one should I believe? After reading the information and comparing it with my colleague's Samsung PM981, I finally understood.: Composite is the weighted comprehensive value calculated by the main control. Sensor 1 to 3 are physical probes distributed on the main control, cache particles, and flash memory. . The main control is the hottest, so Sensor 1 is the highest. It is normal for the flash memory one to be more than ten degrees cooler. The 52 degrees of the mechanical disk and the 52 degrees of the NVMe Composite have completely different gold content.: The former plate body is an aluminum shell that heats evenly, while the latter is just a compromise between various components.

So the first conclusion is enough to save many people from misjudgment.: When you see NVMe reporting 52 degrees, don’t put the dial on the mechanical dial up yet. Mechanical disk 52 degrees is recommended for intervention. NVMe's 52 degrees may be 58 degrees for main control and 41 degrees for flash memory. They are all in their respective comfort zones.

Where is the real red line?: WCTEMP and CCTEMP

There is a scale for judging the temperature. The NVMe scale is not in the SMART output. It is also used in the Identify information. smartctl -x To bring it out, find two words: WCTEMP and CCTEMP . WCTEMP is the warning temperature. When the disk reaches this number, the firmware starts to suppress the performance to protect itself. ; CCTEMP is the shutdown temperature. When it reaches the temperature, it will directly shut down to protect the hardware. The WCTEMP of my PC005 is 70 degrees, and the CCTEMP is empty and not reported. The CCTEMP of common consumer-grade disks is between 70 and 80 degrees.

This set of scales can be seen more clearly when compared with the mechanical dial. If the mechanical disk exceeds 60 degrees, you will start to worry about bad sectors and try again. The official upper limit of the NAS disk is usually 60 degrees. ; The daily working temperature of NVMe is 40 to 50 degrees, which is normal, and the red line is around 70 degrees. The difference of 20 degrees is a physical account of the fact that flash memory has no mechanical parts and the heat resistance range is naturally wider. In my small host, the disk Composite is 52 degrees and the main control is 58 degrees, which is still a long way from WCTEMP. You really shouldn’t panic.

There is another counter-intuitive reading that needs to be taken out separately.: There is a line of "Available Spare" in the output with remaining spare space. If the temperature continues to be high, flash memory wear will accelerate, and the consumption of spare blocks will increase accordingly. The temperature is the "thermometer", and the spare space and "Percentage Used" are the condition itself - how to read the RAW value of SMART without stepping on the trap? RAW value article I have dismantled it using the example of a mechanical disk. The field names on the NVMe side have changed, but the idea is the same.: Look at trends, not single points.

What does overheating look like?: I reproduced it once when I used it as a seed tray.

I can’t feel it just by looking at the numbers, but let’s talk about the overheating I’ve actually seen. The QLC disk in the previous seeding machine continued to slow down external writes, and qBittorrent hung up all night. The next day I saw that the Composite had climbed to 68 degrees, and the disk was obviously slowing down - the speed of dropping out of the buffer after the cache was full was normal, but that night it dropped harder than usual. This is the thermal underclocking of NVMe: When the disk is near WCTEMP, the firmware automatically speeds up and cools down. Save your life first and then work. . There will also be protection against overheating of the mechanical disk, but the frequency reduction of NVMe comes earlier and more obvious, and the performance is "suddenly slowed down a lot."

Fortunately, how to check the external writing speed and how to distinguish between cache and disk? The speed test article There is a complete test method. It is easy to mix up frequency reduction and slow down speed. The way to distinguish them is to keep an eye on the temperature.: The temperature is stable when the external speed is slowed down, and the composite moves upward when the frequency is reduced. When I was making a seed machine, I limited the download speed. After half an hour, the temperature returned to 55 degrees, and the speed returned. It didn’t cost a penny, and the problem was solved.

If your NVMe composite in the NAS is below 65 degrees all year round, you don’t need to do anything. It hovers between 65 and 70 degrees all year round. Don’t rush to spend money yet. The next section is free and done in advance. When frequency reduction is frequently triggered after 70 degrees Celsius, it is the turn for hardware-level intervention.

Free cooling: Location and air duct account for 80%

The position of my disk in the small host was not good at first.: The M.2 slot is directly under the CPU radiator. After the adapter card is installed, the entire disk lies in a dead space without air duct. After moving the position - to another slot closer to the chassis fan, the Composite dropped directly from 52 to 47 degrees, five degrees, at zero cost. Do this account first, and then think about spending money.

Beyond the location is airflow. A common problem in small hosts and NAS cases is not the lack of fans; The air duct is blocked by wires and hard drive racks . The mechanical disk array in the four-bay chassis is itself a wind wall, and NVMe hiding on its leeward side is even worse than if it were close to the CPU. Manually moving cables and adjusting the order of trays is more effective than any heat dissipation sticker. The silently modified NAS at home, Replace shock absorber fan That time I took care of the wiring once, and the temperature difference was more than five degrees in places where the air outlet couldn't blow.

There is another free item that is easily overlooked: Remove heavy loads that don’t do work. Many people use NVMe in their NAS as a cache disk. If the cache strategy is not matched, it means continuous idle writing. How to configure the cache and how to evaluate whether it is worth it on your disk? The 430 yuan solid-state cache article The conclusions of three months of actual testing can be copied directly. After the load was straightened out, the temperature dropped by itself.

Should I buy a vest?: I bought it twice and I’ll give you the account

Just now arrived at the vest. An aluminum shell vest plus a thermal pad costs more than a dozen yuan each. Whether it is worth it depends on the situation of the disk. I bought it twice: For the first time, Sensor 1 dropped by six or seven degrees, which is the value of the disk that the downloader continued to write to. The second time I installed the 52-degree plate in the small console, the composite only dropped three degrees - that plate was originally in the health zone, and three degrees was a psychological comfort. So don’t reverse the order of judgment.: First check how far the temperature is from WCTEMP, and then check whether the position and air duct have to be moved. After everything is done, it is still hot, and the vest only adds the last few degrees. .

disk situation my actions expected effect
Composite is below 60 degrees all year round do nothing You are already healthy, spending money is comfort
65 degrees hovering around Move slots, manage air ducts, and check cache load Lower the temperature by 3 to 8 degrees at zero cost
Frequently approaching WCTEMP, obvious frequency reduction After completing the free items, add the vest The vest needs to be replenished by another 3 to 7 degrees.
No blind spots in the air duct + upper pressure mechanical disk Move your position first, don't expect a vest The waistcoat is also hot even in the blind spot.

There are two small things to pay attention to when buying a vest. First, the thickness of the thermal pad: There are two common types of pads for vest belts: 0.5 mm and 1 mm. For M.2 2280 single-sided discs, use 0.5 mm for a solid fit. For double-sided discs, use 1. If the pad is too thin and cannot be attached, it means it is not installed. If the pad is too thick, it will cause stress. Second, don’t buy a model with an RGB light panel. Lighting the lights in the NAS only consumes electricity, and the light panel also creates an extra layer of heat. When installing a disk on an adapter card, check the height of the adapter card's cover before installing the vest. Some half-height blade cards have screws that conflict with the thick vest. The transfer card article I mentioned this size pitfall.

wrap up: Which numbers should you focus on and how often should you look at them?

Collect the entire monitoring line into a habitual action. Every time I look at the market, smartctl -x Take a trip and only focus on three lines: The distance between the Composite and WCTEMP, whether Sensor 1 (the main control one) has climbed abnormally, and whether the Available Spare has dropped. Do a comprehensive physical examination once a year, and also take a look at the firmware - the windows and rules for firmware upgrades. The firmware article It has been written here that if you are not sick, don’t get promoted randomly. The temperature trend should be viewed together with the life indicator, how much TBW is left, and how to estimate the retirement time? TBW article The accounting method can be used immediately.

In terms of frequency, home NAS really doesn’t need to be monitored every day. My rhythm is "routine once every six months, expedited when problems arise"」: The disk suddenly slows down, the seeding task slows down, and the system log reports NVMe-related errors. If any of these three signals appears, you should run smartctl to check the temperature that day. I usually find the WCTEMP line and write it down, and leave the rest to the judgment that "temperature is a thermometer with other problems." The conclusion of the mechanical disk article still holds true on NVMe, but the scale needs to be read with a new one.

Read More


Copyright Notice Scan to read on mobile
All Rights Reserved: 《SHUNOT》 => 《Is 52 degrees Celsius considered high for NVMe solid state in a NAS? I read the three sensors clearly before I came to a conclusion. 》
Article URL: https://www.shunot.com/en/nas/1568.html
Unless otherwise stated, all articles are original by 《SHUNOT》. Reposting is welcome! Please indicate the original URL when reposting, thank you.

Contact Us

Online Consultation: Click here to send me a message

WeChat ID: master_135

Scan to follow