Interesting! please keep us updated.
I might try this myself as I haven’t deployed the Beelink as yet and just want to see how it will perform with 6 x 1TB 970 Evo plus
Interesting! please keep us updated.
I might try this myself as I haven’t deployed the Beelink as yet and just want to see how it will perform with 6 x 1TB 970 Evo plus
My system is still humming along with TN CORE and no problems to be seen. I identified what might hurt folks with SSDs that have higher power consumption and therefore also higher cooling needs.
To recall: I run 6x Crucial P3, 1 TB each. They have a max power consumption of 5.5 W:
root@truenas[~]# nvmecontrol power -l nvme0
Power States Supported: 5
# Max pwr Enter Lat Exit Lat RT RL WT WL Idle Pwr Act Pwr Workloadd
-- -------- --------- --------- -- -- -- -- -------- -------- --
0: 5.5000W 0.000ms 0.000ms 0 0 0 0 0.0000W 0.0000W 0
1: 3.0000W 0.000ms 0.000ms 1 1 1 1 0.0000W 0.0000W 0
2: 1.5000W 0.000ms 0.000ms 2 2 2 2 0.0000W 0.0000W 0
3: 0.0300W* 5.000ms 2.500ms 3 3 3 3 0.0000W 0.0000W 0
4: 0.0025W* 8.000ms 40.000ms 4 4 4 4 0.0000W 0.0000W 0
Maybe you folks running into problems check that - I assume there is a similar tool for Linux.
Looking closely at the hardware the really clever design - you have to give them that - mounts the components to be cooled on the outside with a giant cooler and a ventilator in the centre. Similar to Apple’s infamous “trash can”, I guess, at a fraction of the price.
This shows how the SSDs are mounted on two of the four sides:
But if you tilt the assembly another 90 degrees and look closely - the SSDs are supposed to be thermally connected to that cooler via thermal pads, but:
After tightening the screws they clearly bend a little bit outwards - I tried multiple refittings pushing them down in the middle to no avail.
And, even if they were perfectly flat they would still not have good contact to that thermal pad.
So that might be one reason for SSDs overheating and failing.
I’d be interested in other users’ numbers with different SSDs.
Mine run at 46 Celsius idle up to a maximum of 55-ish, never to or above 60 so far.
Kind regards,
Patrick
P.S. Maybe changing the thermal pads to ones twice as thick would indeed help. One still needs to keep in mind that thermal paste and pads are only second best. They are supposed to fill the tiny (in the case of paste) air gaps between components. A close fitting of metal on metal would beat paste/pads any time.
I’ve actually used flat copper heatsinks on the outer side of the drive also. I’ll do some pics tomorrow. Got to go to bed as it’s now 03.23am here lol!
I had the same problem with proxmox. I got it to install on the mini ME thanks to this forum post:
https://forum.level1techs.com/t/beelink-me-mini-proxmox-on-mmcblk-mini-howto/231244
In a nut shell, you break out to a command line after it asks for the time zone by pressing CTRL+ALT+F3. Then:
cd /usr/share/perl5/Proxmox/Sys
Then:
nano Block.pm
Search for “nvme” then add the following statement after the “return” line after the “nvme” line:
{ elsif ($dev =~ m|^/dev/mmcblk\d+$|) {
return "${dev}p$partnum";
Save and exit the editor. Then run the command “proxmox-tui-installer” and a text version of the installer will begin and let you complete the installation.
The guy who wrote the article said that he didn’t think that mmc was as bad as it was years ago. Anyhow, I hope this helps. I had to manually type everything in. In retrospect, it might have been possible to telnet or ssh into the box, but I didn’t try it. If you have problems installing, check out his post for a more detailed response.
In my case, I did order some kingston 1GB drives but I don’t have them yet. So the only thing running is proxmox with no NVME yet. I do plan on putting TNCE on one of my NVME’s rather than putting it on the MMC.
As a side note, I know that Linux in general likes to pound on the /var/log directory with writes. I’m not exactly sure just how harmful these are to MMC. Years ago, there was a similar concern with Raspberry Pi’s doing the same thing to the SC card that it was installed on. The solution in that case was to map the /var/log/ directory to a RAM drive. This meant that the log files were volatile unless you had a cron job to copy them over to persistent storage whenever they got too full or on a shutdown.
Other than this, I don’t know enough about proxmox to say anywhere else that it would pound.
Absolutely - install to the eMMC from the installer, boot, done.
No idea, sorry.
I use TrueNAS because it is based on FreeBSD. Now that that changed I’ll rather switch products than OS.
I run a single CE installation for docker based applications - I like the “apps” experience, but if that ever proves to be a dead end I can always spin up a VM with Dockge or Portainer.
I do not “hate” Linux - I use it where it matches the requirements. I just consider FreeBSD to be vastly superior.
Kind regards,
Patrick
I have never seen a thermal pad for NVMe M.2 that was really squishy. I have had different thicknesses and I had to install a thinner one so the chips would lay flat and transfer heat better. I’ve also 3D printed a very small spacers for the screw post to increase the gap slightly so the card doesn’t bend. That M.2 slot is under a lot of pressure with the bend. I have no idea if the solder pads will break free of not which is why the spacer came in handy. I think I used a .4mm spacer. It may not be much but it helped me.
Maybe over time the material conforms better?
If you have any links to the squishy or the really good stuff, please share. I’m actually curious.
Thanks you for sharing. I have also moved from truenas scale to truenas core for Beelink mini after reading your post. I was not even able to initialize the pool with “truenas scale” with raidz1. After moving to truenas core I was able to initialize pool. Have not yet tested with full load of data but currently I have uptime of 5+ hours. But so far it’s looking good.
Currently I am using 6xcrucial p310 4tb in raidz1
That’s good to hear. I’m new to Truenas. Been using Qnap QTS for a while. I’m considering doing the same and installing CORE on the eMMC to free up Bay 4.
The only issue for me would be setting up Plex server.
Would appreciate any pointers if you’ve installed it.
I have
1 x KINGSTON OM8PGP41024Q-A0 as system Bay 4 (TN Scale)
3 x Samsung SSD 970 EVO Plus 1TB in RAIDZ1
2 x Timetec 2 TB MS12 in Stripe
Uptime 28 hours with no issues so far
FYI…I sat a 120mm fan set on (M) setting on the top of the Beelink blowing in.
Temperatures below are from Reporting on Scale
CPU - 28 deg C
1 x KINGSTON - 26 deg C
3 x 970 EVO Plus - 33 deg C
2 x Timetec 2 TB - MS12 - 31 deg C
Not under load however. I will try that later.
Please keep us updated on how you go. Thanks for sharing
Cannot comment on 970 EVO Plus cards but my Mini ME is running TrueNAS Scale version 25.04.2.1 with (6) Crucial P3 2TB PCIe Gen3 NVMes with no problems. I should also point out I’m booting from the eMMC.
My array is a stripped mirror and my peak power usage when maxing out both NICs is < 30 watts. My average temp across the NVMe and CPU is 45C. So far Its been running fine for 2 days now.
The only thing I have not been successful at is using LAGG to aggregate my NICS when connecting to a Unifi Pro 24 switch. It works but usually only for about 5 minutes then I loose connection.
All in all thus far I’m happy with both TrueNAS and the Mini ME.
Can you let me know how you installed the verison “25.04.2.1”? I tried installing it and i have the error “failed to find partition number 2 on mmcblk0”. I was only able to install 24.04 version and then update it. Are you able to install “25.04.2.1” version directly?
I ordered these and they are really soft and mold well to components of various height. They are not elastic, though, won’t bounce back into their old thickness once used. Like play-dough.
That pack of four is enough - you can cut them to length and then use two pieces of the leftovers each for the last two SSDs.
I also removed the paper sticker Crucial put on top of the components.
Right shows PCB side - outside when mounted. Center is component side with sticker, left is component side without sticker.
I have now run 10 concurrent dd jobs with compression disabled - CPU maxed out - and the temperature does not go above 46 Celsius compared to 55 before. Idle temperature is below 30 Celsius now and was 46.
Quite an improvement for 10€
You can also ramp up the ventilation in the BIOS setup changing from “automatic” to “full”.
While “automatic” is whisper quiet on my desk “full” is definitely noticeable but still less noisy than my Supermicro systems. Depending on where you place the system, giving as much ventilation as possible is probably a good idea.
Thank you for that information. I looked into it and these do look very good. Ordering some today.
I got that error as well on each of the initial test installs I did. Just powered down, rebooted and it usually installed on the second attempt. Somewhere there is a thread on how to solve the issue. I didn’t follow it as the reboot was fairly quick and did the job.
The thread on how to solve the issue is titled “Installation failed on eMMC Odroid H4+”. I would consider the last comment on said thread if you dont want to mess with editing py scripts. I didn’t follow it as a power off/on reboot was fairly quick and did the job.
Proxmox was hard coded to reject mmc installations. Several reboots got me the same results. In my case, I don’t actually have my nvme sticks yet, so accidental installation on those was impossible. In your case, did you have your nvme installed when you installed proxmox? Are you absolutely certain it installed on your mmc and not your nvme?
Yes NVMes were installed at the time of truenas installation and yes os is running from eMMC.
Thanks for the info. I’m about to experiment.
I have my TN Scale installed on a Kingston in bay 4 as OS with 2 x Pools.
There are 2 x Timetec 2TB in Stripe (pool-1) and 3 x 970 Evo 1TB in RAIDZ1 (pool-2)
It’s been running fine for 2 days no issues.
I will first destroy the RAIDZ1 (pool-2) but leave the 2 x Timetec 2TB in Stripe (pool-2) intact.
I will copy the existing TN installed on the Kingston in Bay 4 onto the eMMC and replace it with another 1TB 970 Evo plus. That will give me 4 x 1 TB drives to use in RAIDZ1.
All of this will depend on whether my existing TN install will boot from the eMMC.
Im going to experiment tomorrow myself by removing the mPCIe wifi/bt card which I dont use and installing a mPCIe to M2 NVMe adapter with a crucial 1 tb 2230 card. I’ll post the results.
Good Luck
Please report and if successful link to the adapter used. TIA!