[SOLVED] SSD controller chip overheating

[saman81]

Hello and thanks in advance.
I have an omen 17t laptop. It has a Samsung 512 GB PCIe® NVMe™ M.2 SSD which is only used as the boot drive and not for gaming or any heavy load. I've recently noticed some high temps in HWINFO. Temp 1 (supposedly for memory chip) reaches 60s in gaming but never reaches 70. And it's 50 celcius average when idle. But Temp 2 (controller chip) is between 70 to 80 when idle. And reaches and stays on low 90s when gaming for hours. I've been googling and found some contradictory results. Some say it's harmful and some say the controller chip heat itself up to work more efficiently and effectively. And they say it is the memory chip temp that should stay under 70 c. I'm really lost. Can anybody help?

PS. I'm not sure if these high temps are something new to my laptop. Cause i never cared to check the ssd temps before. Sorry for the bad English. Thanks😁

 

[lvt]

Memory chips can withstand 100°C without problem, but for how long only God knows.

Better keep them under 85°C to prevent loss of data.

 

[siaan312]

Those are high, but not crazy or anything.
The controller is basically a cpu and likes to stay as cool as possible, so, 90 is high, but not dangerous, if you want, try thermal epoxying on a heatsink.

The NAND itself likes to stay a bit toasty, but it's optimal temp is around 45-50c, not 70, though 70 won't harm it either.

 

[fzabkar]

I believe that "temp 1" is a "composite temperature" which is a temperature that is calculated by applying some kind of weighted average to the readings from the physical temperature sensors. Therefore I can't see how it could be interpreted as a NAND flash temperature, especially since these chips don't have their own temperature sensors (ICBW, though).

 

[lvt]

I believe that "temp 1" is a "composite temperature" which is a temperature that is calculated by applying some kind of weighted average to the readings from the physical temperature sensors. Therefore I can't see how it could be interpreted as a NAND flash temperature, especially since these chips don't have their own temperature sensors (ICBW, though).

Maybe they use old school sensor like on this AMD motherboard ( a smaller one).

 

[fzabkar]

How would an SSD sense or report the temperatures of all the NAND ICs? Some SSDs have a discrete temperature sensor IC, but that senses the air temperature.

 

[Maxxify]

Samsung NVMe drives have one composite sensor and another one for the NAND/flash. The flash may have a thermal limit of 85C, however it's usually cooler than the controller. Samsung's warning/critical temps for composite are 82C/85C but you can see throttling in the 70-80C range. However, on some drives the second sensor seems to biased incorrectly. In any case, the hardware will still throttle as necessary.

 

[fzabkar]

So where are the physical sensors located?

 

[Maxxify]

"The technique results in notifying the host that the NAND is operating beyond a normal temperature range rather than reporting the temperature. In an example, the notification may be provided automatically (e.g., without being requested) at the end of valid array operations. Here, upon execution of the valid array operation, an embedded temperature sensor is queried to determine the temperature that is then compared against the threshold. " Although normal reporting of the temperature is of course possible. In the quoted patent this method is used to simplify "NAND temperature acquisition for NAND die, reduces overhead (e.g., there is less signaling)" - obviously this is not reporting temperatures individually though.

 

[Phaaze88]

Why not see what Samsung Magician has to say about the drive temperature?

 

[fzabkar]

Thanks. Interesting read. I wasn't aware that current NAND ICs had on-die temperature sensors.

https://patentimages.storage.googleapis.com/70/22/37/fa2d8001bd9c9d/US10331377.pdf

The patent refers to earlier implementations using MLBi (Media Local Bus):

https://ww1.microchip.com/downloads/en/DeviceDoc/TB0400AN4V2 - MediaLB Specification 4v2.pdf

I never noticed these extra pins in any NAND flash datasheet. It all makes sense now.