LinuxQuestions.org
Share your knowledge at the LQ Wiki.
Go Back   LinuxQuestions.org > Forums > Non-*NIX Forums > General
User Name
Password
General This forum is for non-technical general discussion which can include both Linux and non-Linux topics. Have fun!

Notices


Reply
  Search this Thread
Old 08-22-2022, 10:30 AM   #1
doumamuzan
Member
 
Registered: May 2021
Posts: 35

Rep: Reputation: Disabled
Question Is heat sink on NVMe hard disk necessary?


I recently saw an NVMe drive with its own heat sink, and of course, it is priced a lot higher than the same NVMe without its own heat sink

I have never seen an SSD or NVMe with its own heat sink in my life

I have never heard of SSD or NVMe drive experiencing issues because of "overheating"

I would like to hear others' perspectives

In this photo, the NVMe with its own heat sink is the one below

https://kin-phinf.pstatic.net/202208...tSinkyesno.jpg
 
Old 08-22-2022, 12:30 PM   #2
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,395

Rep: Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078
I have a degree in electronics but it's from LONG before any storage cost under thousands of dollars, but I do try to keep up and I build my own PCs as well as for friends and family. I own perhaps a dozen SSDs half or more of which are NVME. Recently I watched a video.. I can't recall which tech'y channel it was (I habituate quite a few) but it was one of the channels that owns and uses serious test gear and graphs out results.

Apparently, and I have to admit I was surprised, the thin, small caps that come on most motherboards perform as well as any extra cooling unit provided of course there is decent air flow AND it was found that performance suffers if NVME drives run too cool! The temperature at which this occurs is substantially lower than most people's ambient, so it's nothing to worry about. I am a cooling freak. I was taught that heat is the enemy of all electronics but that was back in pretty serious vacuum tube days and such tests have me rethinking cooling solutions, at least for NV ME.

I currently have a number of NVME drives mounted on PCIe sleds that came with a fairly massive heatsink and my Main has 9 fans. However I don't intend to reduce cooling but I will definitely seek to pay less for units sans heatsinks.
 
Old 08-22-2022, 01:48 PM   #3
Timothy Miller
Moderator
 
Registered: Feb 2003
Location: Arizona, USA
Distribution: Debian, EndeavourOS, OpenSUSE, Void
Posts: 3,873
Blog Entries: 14

Rep: Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424Reputation: 1424
For most entry level PCIe 3 drives, no. They're simply not fast enough to generate sufficient heat to really take advantage of a heatsink.

For most high end PCIe 3 drives, it can help prevent performance throttling under heavy usage, but for light day to day usage, they don't help.

For most high end PCIe 4 drives, yes. They will generate sufficient heat fairly quickly to cause thermal throttling, and so really do need the heatsinks to achieve maximum performance for more than a short duration. The next generation of drives will begin to eliminate that need, such as the SK Hynix P41 Platinum. Using the latest nand & controller, it's able to achieve the same type of performance as SN850, 980 Pro, etc. while only having 4 lanes on the controller instead of 8. The nand is also capable of higher bandwidth, so this all goes together to allow the P41 Platinum to achieve absolute top of the line performance while generating SIGNIFICANTLY less heat than it's competition.
 
Old 08-22-2022, 04:01 PM   #4
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,395

Rep: Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078
I seem to recall a newer one reiterating this test but this (2020) is what I was writing about with SSDs requiring a specific level of temperature above ambient.

https://www.youtube.com/watch?v=lQmI5A27Iv8
 
Old 08-23-2022, 05:09 AM   #5
fatmac
Senior Member
 
Registered: Sep 2011
Location: Upper Hale, Surrey/Hants Border, UK
Distribution: Mainly Devuan/Devuan based - with - OpenBSD - NetBSD - Fatdog64 - TinyCorePlus
Posts: 4,641

Rep: Reputation: Disabled
I recently put a 128GB M2 SSD into an external metal case, & filled the disk with my files, about 60GB, & that drive case was very hot to handle, when finished. So I conclude that if worked hard, they get very hot, but whilst under normal desktop type of usage, no problem.
 
Old 08-23-2022, 07:27 AM   #6
kilgoretrout
Senior Member
 
Registered: Oct 2003
Posts: 2,933

Rep: Reputation: 378Reputation: 378Reputation: 378Reputation: 378
Quote:
I was taught that heat is the enemy of all electronics but that was back in pretty serious vacuum tube days and such tests have me rethinking cooling solutions, at least for NV ME.
I'm of the same generation where repairing the TV consisted of popping off the back, discharging the high voltage capacitors to ground, pulling out the suspect vacuum tube, taking it down to the local hardware store which had a do it yourself vacuum tube tester and testing the tube. But I wonder how much of this handed down folk wisdom about the dangers of high temps to electronics is still relevant today.

If you are staying within the manufacture's temperature specs and not getting into throttling territory do lower temps result in increased longevity for today's electronics? It would probably depend on the device and would require empirical testing to determine. Unfortunately, to run devices at various temps within spec long enough to determine if there is a significant effect would take so long the device would be obsolete by the time you get significant results. Such is the pace of modern technology.

I know of one test on hard drives done by Google a number of years ago. Google looked at the failure rate of 100,000 hard drives in their data centers over a 9 month period. One of the variables studied was the effect of temperature on failure rates. They found that temps over a broad range had no effect on failure rate except at very high temperatures and then the effect was slight.

https://storagemojo.com/2007/02/19/g...re-experience/
 
Old 08-23-2022, 07:59 AM   #7
boughtonp
Senior Member
 
Registered: Feb 2007
Location: UK
Distribution: Debian
Posts: 2,795

Rep: Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049
Quote:
Originally Posted by kilgoretrout View Post
I know of one test on hard drives done by Google a number of years ago.
Where that number is fifteen, because the data was released in Feb 2007, and the hard drives in question were PATA and SATA, so it's far from clear ifthese results are relevant to the different form factor and technology involved in NVMe drives.

(It's also possible there are differences between brands, and that even the exact same drive used in an always-on server farm results reacts differently when subject to the different environment and usage profile of a home machine.)

 
Old 08-23-2022, 10:33 AM   #8
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,395

Rep: Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078
Just considering the power supply connections it should be obvious that the energy required to create digital storage of bits is always substantially more when mechanical energy is required (which also results in heat) compared to no moving parts. It is also useful to note that the replacement of electro-mechanical drives with SSDs rapid ramping up in Enterprise server farms is n ot just about throughput but also heat, a very substantial component of cost with it's required removal, especially when one may have many terabytes of storage in a very small confined space. Have you heard a Petabyte server startup? The fan noise is crazy, even in, liquid cooled systems.

My conclusion? Heat is still an enemy that bears scrutiny but pure Solid State digital is less vulnerable than electro-mechanical, and NVMEs benefit from above ambient temperatures... not hot, but also not too cool. They have an ideal range.
 
Old 08-23-2022, 11:05 AM   #9
kilgoretrout
Senior Member
 
Registered: Oct 2003
Posts: 2,933

Rep: Reputation: 378Reputation: 378Reputation: 378Reputation: 378
Quote:
Where that number is fifteen, because the data was released in Feb 2007, and the hard drives in question were PATA and SATA, so it's far from clear ifthese results are relevant to the different form factor and technology involved in NVMe drives.

(It's also possible there are differences between brands, and that even the exact same drive used in an always-on server farm results reacts differently when subject to the different environment and usage profile of a home machine.)
Of course, but where's the empirical data demonstrating deleterious effects of within spec high temps in today's storage tech? You accept on faith that heat is somehow bad without empirical data to back it up but try to explain away this Google result which is based on empirical data.

We've been told "heat bad" forever in electronics and in vacuum tube tech it probably was. I'm not saying that the Google study is relevant to today's storage tech. But back when that study was done, when it came to old PATA, SATA spinning hds, everyone thought "heat bad" and the results of this empirical study study surprised everyone. It points to the need to test old beliefs against empirical data.
 
Old 08-23-2022, 11:15 AM   #10
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,395

Rep: Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078
Similar to the fact that Einstein did not supplant Newton (Newton still works within fairly widespread confines) the idea that "heat bad" is still a useful concept within specific environments, but we have ventured into environments by 2022 not imagined since, for example, the 1950s.

Please note that the electronics onboard JWST had to be specifically designed to operate at extremely cold temps yet a few pieces have heaters to raise their temps, insulated from those that require near absolute zero.... the right tool for the specific job.

As "right tool, specific job" applies to this thread, thermal throttling can be observed and quantified. So far measurements confirm the useful temperature range for NVME drives and it appears both High and Low should be capped for a solid balance between performance and longevity

Last edited by enorbet; 08-23-2022 at 11:18 AM.
 
Old 08-23-2022, 11:54 AM   #11
sundialsvcs
LQ Guru
 
Registered: Feb 2004
Location: SE Tennessee, USA
Distribution: Gentoo, LFS
Posts: 9,928
Blog Entries: 4

Rep: Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587Reputation: 3587
I'd say: "if the engineers thought it necessary to include 'an expensive additional metal part' in their design, presume that they knew what they were talking about."
 
Old 08-23-2022, 12:33 PM   #12
wpeckham
Senior Member
 
Registered: Apr 2010
Location: Continental USA
Distribution: Debian, Ubuntu, RedHat, DSL, Puppy, CentOS, Knoppix, Mint-DE, Sparky, VSIDo, tinycore, Q4OS,Manjaro
Posts: 4,531

Rep: Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219Reputation: 2219
Semiconductors change properties and can encounter a condition called "thermal runaway" leading to catastrophic failure.
Yes, that is at least as bad as it sounds.
Their construction is doped silicon: basically dirty glass. Glass is a liquid, and it does flow (although VERY slowly) at room temperature. Guess what kinds of things happen if it gets really hot.
I agree with sundialsvcs, and can add another justification. A heat sink adds mass and parts, making the assembly more complex and cost more to produce. No idiot would add a heat sink to the requirement unless there were significant risk in not exporting heat form the device faster. IF they put it on, it was not just for looks.

Now: new question, did they need a heat sink because the device was that fast and powerful, or because it was so poorly designed that without the sink it accumulated heat without improved performance? I would assume it was a good device until I see a proper and objective evaluation, but I would LOOK for that evaluation!
 
Old 08-23-2022, 01:06 PM   #13
enorbet
Senior Member
 
Registered: Jun 2003
Location: Virginia
Distribution: Slackware = Main OpSys
Posts: 4,395

Rep: Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078Reputation: 4078
There is a 3rd reason... the market in After Market Hot Rod Performance Gear, like fans and heatsinks is multi billion dollars in 2022.

I saved hundreds of dollars buying an i5-10600K instead of say an i9-10600K or 10900K but immediately spent most of that savings on Fans and Heatsinks to overclock the i5 from a base frequency of 4.1GHz to All Core 5.0 GHz. I got some of that balancing act back by being able to get by with 16GB RAM, since more cores requires more RAM. Most of what I ask my PC to do benefits more from Clock than Cores... and I'm an inveterate hot rodder.

Now if I can just wait for the 4000 series GPUs to come out to retire my GTX 1070 Ti....

Last edited by enorbet; 08-23-2022 at 01:13 PM.
 
Old 08-23-2022, 01:15 PM   #14
boughtonp
Senior Member
 
Registered: Feb 2007
Location: UK
Distribution: Debian
Posts: 2,795

Rep: Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049Reputation: 2049
Quote:
Originally Posted by kilgoretrout View Post
You accept on faith that heat is somehow bad without empirical data to back it up but try to explain away this Google result which is based on empirical data.
???

You need to re-read the thread and pay closer attention to what is being said, and by who.
I have made no statements about heat, nor can I see any post for which your response actually makes sense.

The 2007 data was probably useful to hard-drive manufacturers back then wanting to increase reliability. It may even have informed the design of NVMe which began to emerge later that year, but that data is not relevant to doumamuzan's question of whether a heatsink on an NVMe drive is a necessity.

 
Old 08-24-2022, 05:01 AM   #15
doumamuzan
Member
 
Registered: May 2021
Posts: 35

Original Poster
Rep: Reputation: Disabled
Thanks for your responses!

Last edited by doumamuzan; 08-24-2022 at 05:02 AM.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off



Similar Threads
Thread Thread Starter Forum Replies Last Post
How do I find out why my laptop shutoff while compiling? (bad RAM, heat sink?) Quon Linux - Hardware 3 06-18-2012 05:29 AM
[SOLVED] Core 2 Duo heat sink found with double stick tape needsleep Linux - Hardware 7 05-11-2012 02:45 PM
How to open IBM/ Lenovo T41 to clean the heat sink noir911 Linux - Laptop and Netbook 2 08-16-2009 07:07 PM
Heat Sink and Power Supply question arnuld Linux - Hardware 52 12-16-2008 08:41 PM
Heat sink, or not to heat sink ? Pres Linux - Hardware 4 07-13-2003 03:49 PM

LinuxQuestions.org > Forums > Non-*NIX Forums > General

All times are GMT -5. The time now is 05:18 AM.

Main Menu
Advertisement
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
Open Source Consulting | Domain Registration