[SOLVED] Nvme drive temperature sensor seems wrong.

[Marcelektro]

Hello,
My nvme drive has gotten very slow in the past week.
After checking the temperatures, sensor1 says it's 47 *C, however sensor2 says it's 74 *C.
As far as I know, the drive starts throttling the performance as the temperature is above 70 *C.
I touched the drive, and it's pretty cool, I'd say it has like 40*C.
I think, that the sensor2 is broken or something.
How can I fix that? The write speeds don't get above 150 kb/s...

 

[Colif]

what make/model NVME is it? could be a way to test it and check it out.

My 970 evo plus shows 47c for sensor 1 and 50c for sensor 2,

 

[Marcelektro]

what make/model NVME is it

Model: Corsair Force MP500

47c for sensor 1 and 50c for sensor 2,

Yea, similar on my another drive.

The drive has been running since 2018 and has total of 69 TB data read and 241 TB write.

 

[Marcelektro]

Write speeds are extremely low. It's still stuck on 74 *C.
Running Linux Ubuntu Mate 18.04 lts. Can this be software issue?
Again, sensor 1 works good, but sensor 2 is stuck.

 

[fzabkar]

I believe that sensor 2 is the core temperature of the flash controller. Sensor 1 is a "composite" temperature value.

 

[Marcelektro]

Oh. Well, still the temperature difference shouldn’t be that high, and the value shouldn’t be the same for a whole day. Can the connection be the issue?

 

[fzabkar]

Is there a heat sink?

 

[Marcelektro]

No, but since I had this issue, I added a fan blowing at the drive so the air passes through it.
Also the same temperature on sensor 2 is while it’s writing data, and when it’s completely idle.

 

[Colif]

I would see if this can recognise drive and if it has any tests for it - https://www.techspot.com/downloads/6075-corsair-ssd-toolbox.html

 

[Marcelektro]

The website doesn't seem to have Linux version, so I can't really use that.

 

[Colif]

sorry, I sort of assumed it was windows. there are plenty of benchmark tools on linux but none I see that are diagnostic

I would look into a 3rd party heatsink for the nvme, I didn't realise it didn't have one. That should help dissipate the heat from the memory . Otherwise they will get toasty

you could ask on here otherwise if heatsink doesn't help - https://forum.corsair.com/v3/forumdisplay.php?f=188

 

[Marcelektro]

So do you think that's really overheating? It's not sensor issue? Why would it be stuck on the same value then?

I've been using the drive since 3 years and it has never been that warm.

 

[Colif]

It could be a busted sensor but if it only happened recently, its at least worth looking at. 3 years normal, suddenly hot.
Its possible the sensor is used to control the speeds and you said it has been sluggish.
Have you run any speed benchmarks to see what performance is like?

 

[Marcelektro]

It could be a busted sensor but if it only happened recently, its at least worth looking at. 3 years normal, suddenly hot.

How would I check if it's busted? Also why 74*C. Nothing heats that much in my setup.

Also yes, the sensor most likely controls the speeds.
I just uploaded a 10 gb file as I did daily, but one day (when it got super slow due to the "heating") the speeds dropped from 500 mb/s to like 150 - 200 kb/s.
I saw a diagram on one website that showed if temperature exceeds 70, it drops performance, and if it reaches 75% the performance will be nearly 0.
In my case it's 74*C, so the performance is 25% or below. (I guess below, it's horribly slow).

The diagram I mentioned

 

[Colif]

I don't know why the 74c is where it is, its sort of why I linked to the Corsair forums as they might know more and tell you. if it is the sensor itself, you have no way of really knowing - it is only way to tell actual temp. But other tests might reveal something

Speed tests - https://forums.scotsnewsletter.com/...peed-test-readwrite-ssd-performance-in-linux/

Health tests - https://askubuntu.com/questions/325283/how-do-i-check-the-health-of-a-ssd