NVMe SSD temp sanity and heat spreader suggestions

[ArmedAviator]

Hi all,

 

I'm running a Samsung 970 EVO and a 970 EVO Plus in a Btrfs RAID0 on my main desktop.

 

I was running a Btrfs scrub (reads all data from the SSDs and ensures there's no corruption) and the SSD that does not have a heat spreader on it got quite hot.  Total duration for 1.6TB of data was 6 minutes 5 seconds.

 

My motherboard is an MSI MAG Tomahawk Z390 which has the improved "Frozr" heat spreader for only one of the two M.2 modules - which my 970 EVO Plus has affixed to it, but not my 970 EVO.  The case airflow is adequate in a Fractal Design S, and the ambient temp is about 68F in the room.

 

The below temperatures were achieved in only about 1 minute of the 6 minute operation.   Peak temperature I saw was 110C on the SSD in question and hovered in that area for the remainder of the scan.

 

Also, the SSD never seemed to thermal throttle.  Now I question if it's going to cook itself and fail on me.

Quote

rich@area51 ~ $ sudo smartctl -a /dev/nvme0n1 | grep "Temperature Sensor" && sudo smar
tctl -a /dev/nvme1n1 | grep "Temperature Sensor"
Temperature Sensor 1:               56 Celsius  <<< 970 EVO Plus with MSI Frozr
Temperature Sensor 2:               55 Celsius  <<< 970 EVO Plus with MSI Frozr
Temperature Sensor 1:               79 Celsius  <<< 970 EVO with no heat spreader
Temperature Sensor 2:               108 Celsius <<< 970 EVO with no heat spreader

 

Here is an example of idle temps:

Quote

rich@area51 ~ $ sudo smartctl -a /dev/nvme0n1 | grep "Temperature Sensor" && sudo smar
tctl -a /dev/nvme1n1 | grep "Temperature Sensor"
Password:  
Temperature Sensor 1:               38 Celsius   <<< 970 EVO Plus with MSI Frozr
Temperature Sensor 2:               37 Celsius   <<< 970 EVO Plus with MSI Frozr
Temperature Sensor 1:               40 Celsius   <<< 970 EVO with no heat spreader
Temperature Sensor 2:               48 Celsius   <<< 970 EVO with no heat spreader

 

Output of dstat -cdn during scrub:

Quote

--total-cpu-usage-- -dsk/total- -net/total-
usr sys idl wai stl| read  writ| recv  send                                            
 3  66  31   0   0|6588M   37M| 331B  405B
 2  66  31   0   0|6721M 1592k| 218B  218B
 3  66  30   0   0|6655M    0 | 594B  594B
 4  66  30   0   0|6645M    0 | 392B 1171B
 3  66  31   0   0|6666M    0 | 858B  858B
 2  66  31   0   0|6667M    0 | 408B  408B
 3  66  31   0   0|6598M 9560k| 324B  387B
 3  66  31   0   0|6642M    0 | 456B  456B
 3  66  30   0   0|6642M  304k| 336B  422B
 3  65  32   0   0|6503M    0 | 801B 1249B
 3  66  31   0   0|6595M    0 | 253B  253B
 3  65  32   0   0|6356M 5672k|1988k   52k
 3  65  33   0   0|6462M    0 | 330B  330B
 6  65  29   0   0|6483M    0 |  66B   66B
 7  65  28   0   0|6616M    0 |  66B   66B
 8  67  25   0   0|6383M    0 | 794k   33k
 3  66  30   0   0|6456M  960k|1731B  258B
 6  69  25   0   0|6309M    0 | 716B 1067B

6 GB/sec read performance for a Linux gaming and web surfing desktop (the CPU usage was high because both SSDs are encrypted with LUKS/dm_crypt).