Puget Systems print logo
Read this article at https://www.pugetsystems.com/guides/795
Article Thumbnail

Samsung 950 Pro M.2 Additional Cooling Testing

Written on June 6, 2016 by Matt Bach


In a recent article, we investigated how long it takes for a Samsung 950 Pro M.2 drive to throttle in a number of different M.2 slot locations and some of the results surprised us. In fact, in one instance we measured a 75% drop in performance after only 7 seconds!

If you are purchasing a M.2 drive because you need high transfer speeds, seeing our results probably threw up a big red flag for you. Having a high speed drive is great, but not so much if the speed only lasts a couple of seconds. Luckily, in most situations the Samsung 950 Pro drive actually lasts a good amount of time under full load before it begins to throttle. For those that need a M.2 drive to run at full speed for longer periods of time, however, we decided to run some quick tests to compare a number of different M.2 drive cooling methods.

If you prefer to skip all our testing and simply view our conclusions, feel free to jump ahead to the conclusion section.

Test Setup

While there are a number of different ways you could attempt to cool a M.2 drive, we are going to focus on the following three methods:

A basic heatsink added onto the drive. For this, we simply took a piece of 1.5mm aluminum flat stock, cut it to length, and attached it to the M.2 drive with a thermal pad and heatshrink. This is certainly not as good as using something like copper, but we had the aluminum on hand and wanted to give a simple solution like this a shot. One advantage to this method is that the M.2 drive is still thin enough to fit underneath the video card or on the underside of a motherboard.

A M.2 to PCI-E adapter with a large heatsink. This is actually what we have used in the past and while we know that it definitely helps, this will be a good chance to do in-depth testing on this cooling method.

A standard side panel fan. For this we actually tested three different levels of cooling. First, we tested a Fractal Design Silent Series R2 120mm fan at both 5V and 12V. After that, we tested a Cooljag Everflow 92mm fan at 12v. The 92mm Everflow fan is much higher flow than the Fractal fan and the smaller size actually makes the airflow a bit more targeted so more of the total CFM of airflow from the fan goes over the M.2 drive.

Our test platform is one of the ones we used in the Samsung 950 Pro Throttling Analysis article. Since we are simply comparing the different cooling methods, however, we do not need to test all the different M.2 slot locations we tested in that article.

Testing Hardware  

Asus Z170-WS

Test Hard Drive: Samsung 950 Pro 512GB M.2 x4 SSD
installed in bottom M.2 slot
CPU: Intel Core i7 6700K 4.0GHz Quad Core 8MB
CPU Cooler: Gelid Tranquillo Rev 3
RAM: 4x Crucial DDR4-2133 4GB
Chassis: Fractal Design Define R5 Titanium
(front and rear fan @ 5v)
GPU: NVIDIA GeForce GTX Titan X 12GB
OS Hard Drive: Samsung 850 Pro 512GB SATA 6Gb/s SSD
OS: Windows 10 Pro 64-bit
Software: IOMeter 1.1.0 (Performance benchmark)

Unlike our other article, we will be testing the Samsung 950 Pro in the bottom M.2 slot only as this is the only slot where adding a side fan might help (since the top slot is covered by the GPU). To see how much of an impact the ambient temperature inside the chassis has on each cooling method, we will be loading the GeForce GTX Titan X video card with Furmark to simulate the system being under a heavy load.

To make sure our results are as accurate as possible we used a combination of IOMeter .bat files and a custom AutoIt script to automatically benchmark the drive with different work loads (sequential read/write and random read/write). The script automatically wiped the drive prior to each round of testing, created the test file on the drive, then ran though each benchmark while pausing for 5 minutes between tests to ensure that the drive has fully cooled down. The total run time for each cooling method ended up being roughly three hours.

One thing you will notice in this article is that we are not showing any temperature readings from the M.2 drive. We actually logged the temperature from the drive using AIDA 64 during all our testing, but it turns out that the Samsung 950 Pro actually has two thermal sensors - one near the storage modules and one to control throttling. Unfortunately, the sensor that is read by AIDA 64 (and every other hard drive monitoring utility we tried) is actually from the thermal sensor near the storage chips, not the control sensor. The good part about this is that you can see if the drive is getting hot enough to potentially corrupt your data, but it does mean it is not a great indicator of whether your drive is close to throttling or not. We go into this in more detail in the Test Setup section of our Samsung 950 Pro Throttling Analysis article if you are interested in learning more.

Ideal (no system load) results

In an ideal situation, every component in the system is idle and producing a minimal amount of heat. In this situation, the M.2 drive should be able to run at full speed the longest before throttling due to drive temperature.

Sequential Read

Random Read

Sequential Write

Random Write

[+] View individual charts with temperature logs

Compared to the stock Samsung 950 Pro drive, all of the cooling methods gave us much, much better results. We still saw a large drop in random write performance after about 30 seconds, but since that happened in every situation regardless of how we cooled the drive we believe that this is simply always going to happen no matter what steps you take. In fact, this seems to be an indication that the drive is simply running out of overprovisioned area rather than being affected by the drive's temperature.

Overall, we were pleasantly surprised with how effective the different cooling methods were. In fact, with the system at idle the basic aluminum bar heatsink was the only cooling method that resulted in the drive throttling at all. Even in this case, the drive took about 2.5 times longer before it started to throttle and after throttling was about 10% faster than the stock drive. That is a huge improvement for what was essentially taking a piece a metal scavenged from our recycling bin and taping it to the drive.

What really surprised us was how even a small amount of airflow from a side panel fan was able to prevent the drive from throttling. The Fractal Design 120mm fan is a very quiet, fairly low flow fan when it is running at 5V and since it is six or seven inches away from the drive we did not expect it to prevent the drive from throttling nearly completely like it did.

Worst case (heavy system load) results

While the results we saw in the previous section are very interesting, they are also taken from an ideal situation where the system is not doing much work. To see how the cooling methods change when we introduce more heat into the system, we loaded the video card to 100% by using Furmark. This puts a much higher load on the video card than any normal workload, but since we are not loading the CPU at all (to ensure that the extra heat is the only thing that affects our results) it should be a pretty good simulation of a system running a heavy load (such as rendering, gaming, etc.)

Sequential Read

Random Read

Sequential Write

Random Write

[+] View individual charts with temperature logs

Unlike the ideal situation in the last section, with the GPU heavily loaded there were multiple cooling solutions that were not able to keep the drive from ever throttling. In fact, only the PCI-E adapter and the two 12V side fans were completely successful. However, even the worst performing solution (the aluminum bar heatsink) was still able to prevent the drive from throttling for three times longer than stock M.2 drive without any cooling modifications. To put this into perspective, since the Samsung 950 Pro has a read performance of 2.5GB/s, even with just the aluminum bar heatsink you would be still able to read 305GB of data (assuming you could 100% saturate the drive) before you saw any throttling.

The 120mm 5V quiet side fan was a step up from the simple aluminum bar heatsink and resulted in the drive taking 3-5 times longer to start throttling. In addition, after the drive throttled it was 50% faster than the stock M.2 drive. For such a small amount of airflow this is a pretty massive improvement, although it was not quite as good as the quiet and high flow 12V side fans which prevented the drive from throttling at all.


Compared to just a stock Samsung 950 Pro M.2 drive without any additional cooling, every single method we tested did extremely well. Even in the worst case with the simple aluminum bar heatsink, the drive took 2.5 times longer before it started to throttle compared to the stock M.2 drive with no additional cooling. If we were to rate the different methods from most to least effective, they would be:

  1. Tie between PCI-E Adapter w/ Heatsink, 120mm 12V Quiet Side Fan, 92mm 12V High Flow Side Fan. All three of these methods completely prevented the Samsung 950 Pro drive from throttling during our testing. If you fully load the drive for longer than we did (which would mean you need to read more than 875GB worth of data from a 512GB drive), the high flow side fan should perform better than the quiet side fan, but in a practical sense all three of these methods should effectively be able to prevent a Samsung 950 Pro drive from ever throttling
  2. 120mm 5V Quiet Side Fan. While this cooling method was not able to completely prevent the drive from throttling when the system was under load, it allowed the drive to take 3-5 times longer to throttle and after throttling was 50% faster than a stock Samsung 950 Pro. For such a small amount of airflow, this is a much bigger difference than we expected and means you could read 455GB of data (nearly the entire drive) or write 172GB of data continuously before you saw any drop in performance.
  3. Aluminum Bar Heatsink. Technically, this was the worst cooling method we tested but it was still a massive improvement over the stock drive without any cooling. The main downside to a simple heatsink like this is that the hotter the system gets, the less effective a heatsink can be. We only tested with a single GPU, but if you had two or more video cards under full load, it is very possible that a heatsink may be no better than a bare drive or in some situations may cause the drive to throttle even sooner.

The different methods we tested really boil down to two types of cooling: passive cooling with a heatsink and active cooling with a fan. Both can make a big difference, but one thing that was clear in our testing is that even a small amount of airflow over the drive can be extremely beneficial. While we did not specifically test it, even better would be to combine the two methods by having a heatsink on the drive along with a fan providing some airflow over the heatsink.

Keep in mind that in the real world, it is very uncommon to fully utilize a drive this fast to the same extent we did in our testing. Very few programs will actually be able to read from a Samsung 950 Pro at full speed for more than a very short period of time, but if you do have a situation where you need a M.2 drive to perform at full speed for longer periods of time this should give you an idea of what you may need to do to achieve this.

Tags: M.2, Samsung 950 Pro, Throttling
Yuhong Bao

I wonder what most laptops do so far.

Posted on 2016-06-08 22:01:51

Overheat and throttle :)

But really, I would wager that a large portion of computer users would never push their system hard enough to hit the thermal limits. Lets say that there was a laptop using the 950 Pro SSD, and that it was on par with the worst situations shown above. In the roughly 30 seconds you have of full performance before the drive starts to throttle, you could read 75GB of data or write 45GB. Those are both very large amounts, and average computer users probably don't do anything that involves such extended rear or write sessions. Enthusiasts and professionals can, though, in more demanding workloads... and they are the folks we had in mind when putting together this testing.

Posted on 2016-06-08 22:17:31

Thanks for the analysis.

I wonder how well a RAM heatsink on the controller chip would work compared to the aluminum stock. Also, the NUC and mini-STX platforms stack the WLAN under the SSD. That's gotta impact the time-to-throttle.

Posted on 2016-06-08 22:42:27
Lars Eriksson

Did you remove the label/sticker on the SSD before installing the heatsinks?
That is a pretty important detail :)

Im planning to install this heatsink (after some cutting and new screw holes) http://www.ebay.com/itm/381...
And i want to know how important it is to remove the sticker. Ive also heard that the sticker material is some kind of metal, is that correct?

I would rather not void the warranty if i can avoid it.

Posted on 2016-06-13 09:12:11

I experienced the same problem. I was experiencing thermal throttling of about 75% when bench marking a Samsung Pro 950 M.2 NVMe drive in my laptop. I wanted to add Copper RAM heat sinks on top of the Flash and Controller modules, but the manufacturer put their label right over these modules and I didn't want to void the five year warranty. I was, however, advised by the Senior Storage Reviewer for a major online publication to remove the label before installing heat sinks -- for best results.

So I removed the drive and used a hair dryer to heat up the adhesive under the label and (very) slowly pulled it off. I then stuck it to some acrylic material so it wouldn't tend to "roll up" and stick to itself. I will be able to put it back on should I ever need to send the drive in for warranty.

I ordered some small pure Copper RAM heat sinks to put on the Flash and Controller modules. I also ordered some thermal adhesive to attach the heat sinks and I am going to have to grind off part of the height of the heat sinks to get them to fit under the bottom cover of my laptop. This will, no doubt, reduce the efficiency somewhat but adding the heat sinks will probably be better than doing nothing. I will post the results when I am finished.

(Note: Just as an experiment, I taped three old copper pennies (modern pennies are 97.5% Zinc) to the flash and controller chips and it lowered the max temp of the drive under ATTO bench marking by about 6 degrees C. I was suprised. So I am hoping the RAM heat sinks I ordered will do even better.)


Posted on 2016-08-06 22:50:56
Dark Evry

Here this will solve all your problems, its 4mm pure copper and comes with everything you need to install, if its too high for a laptop, they have identical 2mm version


Posted on 2017-06-21 02:51:50

Thanks, I will give the 2mm version a try and see if it improves on the Copper RAM heat sinks I ground down to make them fit my laptop (although they did take care of the thermal throttling issue).

The heat sinks you linked to are shipped from China, I have ordered parts from China in the past. And even though they may say it takes two weeks, it generally takes 4 to 6 weeks for delivery. So anyone ordering them should be aware of this.

Posted on 2017-06-21 03:56:37
Dark Evry

Sometimes it takes 2 months, but thats Ok, you just order and wait.
You can file PayPal dispute for up to 6 months.
eBay disputes are faster and will remind you on product page, if you didn't get the order by certain date ebay will send you an email asking you to open a dispute, you get your money back automatically. And most of the some you also get the product eventually after 2-3 months
So its a win-win

Posted on 2017-06-21 04:52:29

Note: The RAM heatsinks mentioned in the above comment are no longer available. See my later post below for an update.

Posted on 2019-09-09 23:27:29
Erik Larson

The Samsung Pro line user a copper plated label, so actual helps with thermal transfer. It's not just paper.

Posted on 2017-05-08 05:59:39

Only the Samsung Pro 960 series. The earlier 950 Pro versions use a paper label -- and they can gain a significant (heavy use) performance improvement by heat sinking them.

Posted on 2017-06-21 04:02:51
Cody Gabel

My pcie nvme drive ( Sabrent rocket 1tb) had a copper sticker heatspreader. I removed this before applying a Jonsbo Battleship heatsink...gotta love that name.

Posted on 2020-05-19 00:15:44

Thanks to Puget Systems for writing this article, I've learned quite a lot from their self help discussions & tips.;-)

Since I have some extra, and mine's going into an Addonics x4 adapter purchased at Newegg along with the Samsung 950 Pro (512GB), am going with the fan method first. As I was able to cool down a Crucial 256GB mSATA SSD quite a bit (those whom purchased Samsung EVO models reported the same heating issue). In my case, that was quite odd, as all of my other components were running lower than normal temps, so it was a surprise that the mSATA SSD ran x2 warmer than the GPU under normal workload, and 25C warmer than the CPU or MB.

Hopefully the fan trick will work for me, as I only purchased the 512GB Samsung 950 to prove a point on the Tech forum where I'm an Advisor 'my hardware is better & 3x faster than yours'. Plus I have so many SATA-3 SSD's, while the 512GB Samsung Pro I'm now on is no faster than a couple of the others, purchased for the same reason, and for me, peace of mind with the 10 year warranty.

Have a blessed day!


Posted on 2016-08-18 06:08:46

Seriously? >>>it is very possible that a heatsink may be no better than a bare drive
or in some situations may cause the drive to throttle even sooner.<<<

If ambient temperature is X, then the SDD temperature WILL become X as well, a heatsink MIGHT push it there slightly faster, but unless you have wildly fluctuating ambient temp inside the case, that "slightly" is utterly irrelevant and doesn´t matter. If ambient is warmer, then the drive WILL be warmer regardless of heatsink or not, but any temps above ambient will still be dissipated faster WITH a heatsink.

Posted on 2016-11-30 13:11:20

You're only considering ambient temps, not situations. Any heat sink strapped to an m.2 makes it thicker, which means less clearance and airflow in cramped conditions. Also surface area if you're making the heat sink larger than the m.2 footprint, which also means less air circulation. Now take a non-blower GPU and dump heat from it straight into your restricted airflow m.2 heatsink...

Posted on 2016-12-03 10:28:42
Chris Kennedy

Thanks for the work on this issue. I frequently transfer large files from my desktop to my laptop. I'm using an Ethernet cable transferring from a Samsung Pro SSD 256 to an Intel 120MB 540s M.2 on my MSI GE72 laptop (purchased 8-2016). File transfer starts out at 112 MB/s for 35 seconds (4.55 MB) then is reduced to 60 MB/s for the remainder of the 12 MB file. After reading this article, I installed heat sinks to my M.2 (see pics). This had no effect what so ever. I’m leaning toward the cache on the M.2 drive, information that I haven't been able to find yet. Here is a snippet from Tom's Hardware: “There’s no mention of native TLC write speeds. Intel isn't the only company hiding its native TLC write performance, which occurs outside of the SLC cache buffer". I'm replacing the Intel M.2 with a Samsung 960 EVO 250MB (13MB cache) next week, I hope this solves my issue. It will have better heat sinks installed this time :). Any comments on what I'm trying to do here would be very helpful, thanks! https://uploads.disquscdn.c... https://uploads.disquscdn.c...

Posted on 2016-12-12 13:11:18
Joshua I.

You're not measuring the read/write speeds with this situation, your bottleneck is the ethernet/adaptors. Unless you have gigabit controllers in both desktop and laptop and assuming both controllers are able to transfer at max, you're going to be limited to 100 mbps (megabits/s) or 12.5 MB/s (megabytes/s). Even with gigabit ethernet controllers you're limited to a transfer rate of 125 MB/s and then your second drive is still having to write (after it's being transferred). Basically your drive isn't throttling - your ethernet is bottlenecking you. You could upgrade to 802.11AC on both which has a max theoretical speed of 1.3 gbps or 166 MB/s.

Posted on 2016-12-29 20:45:12
Chris Kennedy

Thanks for the reply Josh. I'm aware of the top speed of my Ethernet. I was whining because the Intel M.2 drive was throttling after 35 seconds, from 112 to 60. The reviews give you the speed ratings of the first few seconds and not the real world transfer speeds. I recently installed the new Samsung 960 M.2 and it doesn't throttle down at all (12GB file). I'm happy now. The 960 controller peaked at 140F during windows install. 3000 MB/s read 1500 write. Windows is instant-on. https://uploads.disquscdn.c... https://uploads.disquscdn.c... https://uploads.disquscdn.c...

Posted on 2016-12-30 13:58:43

I have a 960 Pro 1TB that is getting really hot, idles at 65C and I've seen it go up to 95C. It's back mounted of course and not too far from the CPU. Do you recommend a thin fan between the motherboard and case like this Evercool or a copper heatsink held on with silicone bands? These are the two options I'm looking at. Not sure if either will fit in my Node202 or not.
I'm leaning towards the fan as I'm not sure as you mentioned in the article that without air moving, what good does a heatsink do? Was planning on keeping the fan in place with a few pieces of doublestick tape.


Posted on 2017-06-03 10:10:49

If you are idling that high, I'm not sure a heatsink is going to do too much since the M.2 drive is on the back of the motherboard. Without any airflow, a heatsink will slow down the drive getting hotter, but won't prevent it. If you would somehow get cool air (or at least relatively cool air) behind the motherboard, that is going to do more than a heatsink I think.

Posted on 2017-06-05 17:25:22

I had the same thought and looked into a very thin fan that would fit between the back of my motherboard and case. The slimmest I found was 10mm which isn't going to work.

I've recently increased my base CPU fanspeed and that has helped a bit. Under CPU temp of 50C my fan goes to it's slowest speed, at 51C it starts to slowly ramp up. I'm going to continue to increase this 51C+ base speed until I can notice it in desktop use and lower it just to that point. Right now my 960 Pro is idling at 60C with a low of 56C and max of 64C. I previously had an 1800X which is for sale, that I had a lot of trouble keeping cool n quiet thanks to XFR constantly ramping to 4.0GHz+ and a baseclock of 3.6GHz that I never saw downclock (Ryzen Balanced plan). Ryzen X-series are bad for a Node202.

I bought an Alphacool HDX SSD heatsink from Amazon that arrives today. I don't expect it to help very much, but should be better than nothing. I'm hoping it blocks a bit of the heat coming through the motherboard from the CPU and gives the heat that does get generated while the drive is under load a place to go. Most of the heat isn't actually created by the drive, but coming from my CPU. This is a Ryzen 1700 / 960 Pro 1TB / Geforce1060FE in a Node202.

Ultimately I'm going to swapout my Biostar for the inevitable Asus ROG AM4, which I hope has a front mounted M.2 like their mITX Z270 board and that should be more than enough breathing room.

Posted on 2017-06-06 18:17:28

Going through my Disqus notifications and figured I'd followup on what I said previously. That Alphacool HDX SSD heatsink didn't end up working- made it too thick to fit (or even be attached to the M.2 slot at all).

I'm still running the Biostar X370GTN / 960 Pro configuration, but ended up simply increasing the CPU fanspeeds to keep the heat off the CPU thus no time to linger through the CPU backplate to my M.2 drive.

Temps are good. Idling right now at 59C on my 960 Pro with a max of 62C. As long as I stay under 70C (Samsung's max recommended operating temp), I'm happy.

Still looking forward to a front-mounted M.2 AM4 mITX board. Asus ROG boards are the only ones I've seen with this, so I'm expecting one to land around the AMD Raven Ridge launch. Will swap out then. This Biostar has been problematic. Reflashed the BIOS a few times trying to resolve strange issues with the fan settings being ignored. Got it all running well at this point but still, no thanks to Biostar.

Posted on 2017-07-15 22:01:39

This post is several years after-the-fact, but might be helpful to some anyway. The Samsung 950 Pro Series NVMe drives are now long in the rear-view mirror, and have been superseded by the 960 and then the 970 series. And the former throttling issues with the 950 series have been somewhat addressed (Samsung added a (very) thin heatsink to the 960 and 970 series).

However, there is still an excellent reason to use a 950 Pro in Laptops. And that reason is: the 950 Pro series deliver far better battery life than either the 960 or 970 series (as verified by reviewer Chris Ramsayer @ TweakTown Storage). In some cases, as much as 30 minutes more.

Additionally, a 512GB 950 Pro with very few host Writes can often found on eBay for a very reasonable price. I purchased one a couple of weeks ago for $59.95 (freight paid/no tax) that had just 8GB of host Writes on it (it's endurance is rated at 400 host Writes -- so essentially new).

Using (the apps) Crystal Disk Info, and HWiNFO64 (which gives much more data) to monitor drive temperature, I tested the Samsung 950 Pro 512GB in my Laptop both with, and without a Copper heatsink and thermal pad.

Over a 30 Minute monitoring period (with the Laptop sitting at idle) the results were:

With no heatsink the drive temperature maxed out at 140 degrees F. and had an average temp. of 138F.
With the Copper heatsink and thermal pad installed, the drive Maxed out at 134F. with an average temp. of 132.

So simply adding the Copper heatsink and thermal pad reduced the operating temp. of the drive by an average of 6 degrees Fahrenheit -- and that's significant when you consider what that 6 degrees temp. reduction adds to the length of time before the drive throttles.

(Note: I did not remove the Samsung label on the drive before installing the thermal heat pad and Copper heatsink.)

I was able to benchmark the drive using "ATTO Disk Benchmark" and the drive never went into throttling during the entire test.

The Copper heatsink I installed, including 2 thermal pads and 3 silicon bands (to hold it to the drive), is available from Amazon here:


Posted on 2019-09-09 23:23:35