NVMe Burn-In Testing?

[joeschmuck]

While I consider myself quite knowledgeable on the topic, I still like to ask questions to gain other perspectives and opinions.

I have six 4TB NVMe modules coming in and I want to test them specifically for thermal properties to ensure I have no issues overheating. I do not want to perform a significant write test but I'm perfectly fine doing many read test. So the question is, what options are there to generate the maximum heat in an NVME?

I plan to run a SMART Long Test on each drive a few times, maybe I should put it into a looping script and that will generate some heat. I am also thinking badblocks in non-destructive mode might be an option, but I don't yet know it it is actually a read only test. If it is a Read, Write, Read test then it's a nogo for me.

All of the NVMe modules are PCIe 4.0.

Appreciate the feedback.

 

[probain]

I'm guessing that you don't wish to use writes, due to "wearing out" the drives?
But if all you want to do is generate heat, then running badblocks or similar. Shouldn't make any real difference. If the drives are new, then they should be able to write that much data for years. A day or so of testing doesn't really matter in the grand scheme.

But this is my very humble opinion. And I totally see where you're coming from in your reasoning. And if by chance I am way off in mine, I'll happily take a bite out of that humble pie :)

 

[joeschmuck]

I'm guessing that you don't wish to use writes, due to "wearing out" the drives?

Correct.

If the drives are new, then they should be able to write that much data for years.

Eh, I don't think like that. Here on this forum we have tried to discourage running badblocks on a SSD. Running badblocks through one cycle can use as much as .02%, it's more than I want to give away.

A day or so of testing doesn't really matter in the grand scheme.

While I agree with you for the most part, it is not a big issue in the grand scheme of things, the only reason I personally would perform a write test on the NVMe drives is if I questioned if the drive was functioning properly, mainly for an RMA exchange.

However my goal with this question was to find out what would make them heat up without writing to them, besides using a blow torch.

My other observation will be when I place all my data on the NAS, if the NVMe does not record Min/Max temperature (I haven't seen one that does yet), I will have to record it using a script I have. This will should give me a better idea but if I can get a good test from reading, I'll be happier.

Why am I going through all of this in the first place?
When I build this computer, I typically customize the cases to ensure proper airflow across the components that need the cooling. I will cut holes into the case, they end up looking like the case was manufactured like this. I will add mounts for fans if I have to. But I'd rather build it, test it, pull it apart and modify the case one time only. Then put it all back together. The goal is to have a silent power sipping TrueNAS machine in a nice small Mini-ATX case.

You can call me an overthinker. I have time on my hands and I try to think things through to the end so I do not need to do any rework. Plus I laid out a lot of money to build this and it's just a hobby to me. I like to build and fly RC Airplanes, another hobby that requires attention to detail, and costs a lot of money, but it's a lot of fun.

So everyone has their own opinion and yours is not wrong at all. One thing you will get here are a lot of opinions and to me, it is okay to disagree as long as we are sharing, learning, and being civil about it. No humble pie for you today, sorry.

 

[winnielinnie]

I am also thinking badblocks in non-destructive mode might be an option, but I don't yet know it it is actually a read only test.

badblocks defaults to a read-only test by default if you do not provide it the -w or -n flags. This might be a way to "heat up" the NVMe.

Example:
badblocks -b 8096 -c 65536 -p 99 /dev/nvme1

CTRL + C to cancel, or let it run through 99 red bal-, passes.

 

[joeschmuck]

But I like the Destructive Testing

 

[probain]

0.02% was more than I expected though. Interesting.
And now you also made me curious of your case design too

 

[joeschmuck]

I will post photos. I don't believe this case will need to be modified. I forgot what I purchased. I have the ability to add fans where I will possibly need them. But yes, I have modified cheap cases. Use a dremel tool, cut a circle, deburr it, add a nice chrome or black fan grill. The thing with most cases is they puncture a lot of holes where a fan would mount and think that is good enough. It isn't. The metal left is a restriction, could be one third or more of a restriction. Remove it to provide unobstructed airflow but do a professional job so it looks like it could have been factory made. Also the fans do not need to run as hard which lowers fan noise.

I just got the pieces in a few hours ago so it's going together very well. Unfortunately I'm a little ill so my brain is not firing on all cylinder's so I'm taking my time to not mess it up.

 

[winnielinnie]

Unfortunately I'm a little ill so my brain is not firing on all cylinder's so I'm taking my time to not mess it up.

Get well, sir. Rest up.

I like to think of my body telling me to "slow down" whenever I get sick. I listen to it. (It also helps that I'm already lazy to begin with... )

 

[joeschmuck]

The NVMe's I purchased have a built in heatsink, but they do not fit on the PCIe adapter card flush as they should. Arg, need to find a way around this and keep the heatsink. I'll figure it out, might take a few days. But the computer is otherwise operational. I need to update the BMC and BIOS firmware and I will be done with this project for the evening and figure out the heatsink problem. I may not need them however if I can keep them, I want them.