cooling M.2 NVMe Drive overheating!

[Husky]

Hello!

 

I have just got a Samsung 970 EVO 250 GB M.2 NVMe SSD, and I installed it into my slot on my board, and then proceeded to test it out.

 

I noticed that the read speeds were consistent at around 3500MB/s, but the write speeds held 1500MB/s for 10 secs and then dropped to 300MB/s and stayed there for the rest of the write. I then touched my hand against the drive thinking that it was thermal throttling and boy was I right because it literally burnt me so badly I still have a mark on my finger!

 

What can I do about this? There is pretty decent airflow around it already. Can I attach a heatsink with thermal pads to it? Do M.2 heatsinks exist? I just built a Ryzen rig for a friend and that board (ASUS Prime X470-Pro) had an M.2 heatsink built in which helped a lot (no throttling) but my board (Z170-DELUXE) does not.

 

Any suggestions would be appreciated!

[Mira Yurizaki]

You can find plenty of M.2 heat sinks around the interwebs.

[CharminUltraStrong]

4 minutes ago, Husky said:

-snip-

There are a plethora of options for m.2 SSD heatsinks some have active coolers, some have RGB and some are passive. If you don't want to spend 30.00$ USD on cheap sheet of metal with a pad in between then grab some Raspberry Pi heat sinks. They come in all sorts of sizes and can be had for mere pennies. Sure they're not made for M.2 Drives but they could offer better performance.

[XR6]

There's loads of M.2 heatsinks for sale on eBay. 

 

[nick name]

EK makes a popular and attractive m.2 heatsink.  Sells on Amazon for a fair price, but there are plenty more options if you want to continue to browse.  

[TheGlenlivet]

Just outta curiosity...

What temps are you seeing on the drive?  Can you point a fan at it and see if speeds increase as temps decrease?  I've just never seen NVME speeds affected by temps on any of my systems, so I was curious if this would prove temps are the issue.

[bowrilla]

Well, how hot/warm is the drive? Thing is: NAND flash prefers running hot. Cooling them will lead to more degradation up to cutting their lifetime in half. JEDEC specified the working temperature with 40°C. You really do not want to cool your NAND flash chips! Now the controller chip on the other hand does prefer running cooler. If (!) you actually do run into thermal throttling it's solely because of the controller and not the NAND. Manufacturers of cooling solutions get this terribly wrong. It is very hard though to actually push your NVMe hard enough under realistic loads (meaning: not synthetic benchmarks) in order for your controller chip to overheat. If OP actually experienced thermal throttling then I assume it was during synthetic benchmarks. It is also possible that the Cache was just filling up. That's totally normal at some point.

 

See also: 

 

[Mira Yurizaki]

21 minutes ago, bowrilla said:

Well, how hot/warm is the drive? Thing is: NAND flash prefers running hot. Cooling them will lead to more degradation up to cutting their lifetime in half. JEDEC specified the working temperature with 40°C. You really do not want to cool your NAND flash chips!.

Do you have a source for this? Because the only thing I'm seeing about temperature being better points to research on self-healing of cells. But the temperatures those articles note is quite large. Everything else I've seen about temperature is related to data retention.

[bowrilla]

10 hours ago, Mira Yurizaki said:

Do you have a source for this? Because the only thing I'm seeing about temperature being better points to research on self-healing of cells. But the temperatures those articles note is quite large. Everything else I've seen about temperature is related to data retention.

Even though I did provide a source I accept that it's a secondary source. This article provides more data: https://www.eeweb.com/profile/eli-tiomkin/articles/industrial-temperature-and-nand-flash-in-ssd-products

 

Higher temperatures will reduce data retention but also reduce stress during writes. 

 

I still regard GN and their sources as reliable.

[Mira Yurizaki]

8 hours ago, bowrilla said:

Even though I did provide a source I accept that it's a secondary source. This article provides more data: https://www.eeweb.com/profile/eli-tiomkin/articles/industrial-temperature-and-nand-flash-in-ssd-products

 

The conclusion I have when reading that article is not that it's good to leave flash memory at elevated temperatures all the time, but only at the time of erase/programming:

Quote

At high temperature, programming and erasing a NAND cell is relatively less stressful to its structure, but data retention of a NAND cell suffers.

Considering that in most cases people aren't constantly writing to flash memory, I thinking keeping it cooler for the sake of better data retention should be preferred. Even when used as a system drive, and thus it may be expected that more writes occur due to pagefile swapping, statistics from Microsoft suggest that write events are still fairly infrequent:

Quote

• Pagefile.sys reads outnumber pagefile.sys writes by about 40 to 1,
• Pagefile.sys read sizes are typically quite small, with 67% less than or equal to 4 KB, and 88% less than 16 KB.
• Pagefile.sys writes are relatively large, with 62% greater than or equal to 128 KB and 45% being exactly 1 MB in size.

From https://blogs.msdn.microsoft.com/e7/2009/05/05/support-and-qa-for-solid-state-drives/

 

Besides that, there's no data suggesting what the overall life expectancy of flash memory. I've heard the reason why you should keep electronics cooler is that for every 10C the part is hotter, its life expectancy is cut in half. Except... what is that life expectancy to begin with?

 

EDIT: On that note, Wikipedia has a blurb that semiconductors actually run better when hot.

[Paul Sabian]

On 1/22/2019 at 1:29 PM, Husky said:

Hello!

 

I have just got a Samsung 970 EVO 250 GB M.2 NVMe SSD, and I installed it into my slot on my board, and then proceeded to test it out.

 

 

[Paul Sabian]

On 1/22/2019 at 1:29 PM, Husky said:

Hello!

 

I have just got a Samsung 970 EVO 250 GB M.2 NVMe SSD, and I installed it into my slot on my board, and then proceeded to test it out.

 

I noticed that the read speeds were consistent at around 3500MB/s, but the write speeds held 1500MB/s for 10 secs and then dropped to 300MB/s and stayed there for the rest of the write. I then touched my hand against the drive thinking that it was thermal throttling and boy was I right because it literally burnt me so badly I still have a mark on my finger!

 

What can I do about this? There is pretty decent airflow around it already. Can I attach a heatsink with thermal pads to it? Do M.2 heatsinks exist? I just built a Ryzen rig for a friend and that board (ASUS Prime X470-Pro) had an M.2 heatsink built in which helped a lot (no throttling) but my board (Z170-DELUXE) does not.

 

Any suggestions would be appreciated!

One of my Drives died yesterday as well due to heat problem, even Corsair D900 Case with descent cooling was unable to save it.

This is the only disk which was installed into motherboard and second one i have in cooling cartridge from Angel Bird Wings PX1 Company and it stays around 42 c and the one which was unprotected was able to reach 72c in idle mode.

i did send request for warranty already but still waiting for them to respond.

Again i would recommend something like Bird Wings Adapter in order to cool it down .

It works Great!!!!!............

 

 

[Undermoose]

Get a modern heat sink. I had one of those, but it's EOL and dramatically slows down a 960 PRO/EVO. It'll work though.