Ever seen NVMe under FLIR? Samsung PM961 512GB NVMe SSD Thermal Analysis

Thought I'd share this with you guys, offers some interesting insight into NVMe cooling. I've seen people on various forums looking into water cooling blocks etc, and found it quite interesting to see if it's actually needed or more of a snake oil remedy. I mainly focused on the connector side of the module as the NAND chips themselves really don't seem to do much in the way of heat at all. Rather, they only gradually 'warm' up as the board itself gets saturated by the NVMe SSD controller.

Interesting phenomenon is the flutter on the NVMe controller at the start of the video, this is actually me moving my mouse over icons on the desktop before starting up applications and downloads to give a real world stress test.


Part 2 w/ Benchmarks

 

@Dialup David cool! What’s your take on the necessity of nvme cooling though? I’m not too knowledgeable? But the CPU hitting 90 degrees is troublesome. Do SSDs hit a temp where their performance is altered by “throttling” of the ssd? I don’t think hitting 60 degrees causes chip damage—computer implementation design to downgrade performance in response.

 

I did a little testing using more old school tech.

I have 2 PCIe Gen 3 x 4 slots on an MSI Z370 M5 mainboard desktop.Only one slot has an included heat sink and thernal pad.

I placed a thin wire sensor between a Samsung 512GB 960 Pro and heatsink and powered the digital display from a SATA power connector.

Starting the Windows10 Pro x64 OS in the morning displayed a startup temperature of 23.6C that rose to 39.9C with only background processes running.

Opening Firefox browser and live streaming ustv247 dot com fox news raised the temperature to near 44.0C

The real test comes when running Crystal Disk Mark v6.0.2 that's set at 9 passes.

A high temperature of 61.6C was observed and recorded during the Sequential write and 4K QD16 tests

 

I'd love to see this SSD (1TB version ofc) under FLIR. They always report 46C temperature, making me very wary of purchasing despite reasonably good benchmarks.

 

I have that one, and its even hit 110c once, and the thing just gets hot even if the drive is completely empty. Its funny compared to my 8200 pro, which never goes past 30's, but thats the benefit of a cooling plate.

 

Software cannot measure temperature anywhere near as accurately as a hardware device that is designed for that purpose.
I place no faith in software readings that are not a result of a hardware sensor placed on the spot to me monitored.

 

@Starlight5 @Casowen wow. That’s a really good price for a 1tb M.2! I just got my addlink s70 1tb for 144$. But 126$ is a steal...with that said 101C does come across as concerning unlike the 60C measured by whatsthepoint. @WhatsThePoint you should see about using your hardware measurement on the Asgard ssd id be curious too.

 

@Joedoewoe 101C is a Samsung. It seems a faulty drive to me, I believe @Casowen better return it.

Asgard's real temps remain a mystery, we need someone brave enough to order. I won't be surprised if different batches of Asgard come with different NAND, however.

 

I have no desire to purchase an Asgard ssd or any others at the moment.

I still have an OCZ Vertex 2(Intel chips)Vertex 3 Max Ops,SanDisk Extreme,Crucial MX500,a few 860 Evo 2.5" plus NVMe 960 Pro and a 970 Pro.

The NVMe SSDs are boot drives while all the others are storage and testing W10x64 19H1 builds.

I hope the next I buy is for PCIe Gen 5.

If I was tempted to buy another it would probably be an HP EX920 or EX950.