Mini HTPC with Thunderbolt for QNAP DAS connection (Getting reboots on multi-GB file transfers with current setup).

[RimBlock]

Hi,

I have a QNAP 4 disk DAS with Thunderbolt and an Intel i7 NUC but the NUC is struggling with heat management (ambient here is around 32 degC) during large file transfers and 4k movie playback.

The NUC currently runs Windows 11 Pro and whilst I am not locked in to this OS, I like the fact that if anything goes wrong I can pull a disk and troubleshoot it on my main Windows desktop.

I am thinking of replacing it with a Tiny PC with better cooling and am hoping the community can point me in the right direction.

Key requirements;

• Prefer NVME for OS.
• Thunderbolt 3 connection.
• 1GbE Lan (faster will be unused as the home network is 1GbE).
• Bitstream pass-though app compatible to my Marantz HT Receiver via HDMI.
• Tidal app compatible.
• Low power as it is always on for NAS duties.
• Low noise but does not need to be silent as we are next to a fairly noisy busy road anyway.
• Fairly cheaply sourced on EBay (2/3/4 gens old is fine if it matches the above at a reduced price).
• No gaming functionality is required.

What would be a good starting point, ideally with minimal hacks / fixes to get past vendor lockouts .

Again, happy to look at alternative OS's outside of Windows if they meet the requirements above although would prefer something without a subscription or significant setup work (cover art, bios etc) as a 'media player' type front end adds very little value for me. I already have a Nvidia Shield Pro that I use for Netflix / YouTube etc and scrolling through hundreds of pages to find a movie to watch feels like a waste of time compared to browsing to a folder and double clicking. Happy to listen to alternative viewpoints though .

Thanks.

 

[Vorwrath]

Have you considered just improving the cooling on your existing NUC? The Akasa fanless cases might be an option, depending on exactly which NUC you have - some of them reportedly perform a lot better than the stock cooling.

You might also just be able to lower the package power limit a bit in the BIOS, to make the CPU run less hot. Although that will reduce performance accordingly.

Most mini PCs tend to run on the warmer side, so you might otherwise end up swapping to something that isn't much different.

 

[RimBlock]

Thanks for the suggestion.

I believe it is a NUC10i7FNH. Unfortunately Akasa only does one case for it and it is too tall, even on its side. One of their slim wide cases would have been great. I would like to reuse the NUC if possible as it would be a waste otherwise.

I was hoping one of the mini micro PC's would still be small and have better ventilation / cooling if I had to switch it out. The HP HP Z2 G9 looks good but is very expensive for what I need and no idea if it has thunderbolt 3. Its size would probably work ok though.

 

[Vorwrath]

Under the "Power" settings in the BIOS, make a note of the original values, and then try setting: -
Package Power Limit 1 (Sustained): 25
Package Power Limit 2 (Burst Mode): 25
Platform PL1 Time Window: 0

The default settings allow the processor to boost well above the normal power limit for a little while - this disables that and limits the power to 25 watts, which might help quite a bit with heat generation. If it still runs too hot, you could also experiment with a lower value than 25 (some lower power devices run the i7-10710U that's in your NUC at 15w, which is the base spec for the part, so it can certainly go down to at least that with somewhat reasonable performance).

I don't think the HP Z2 G9 has Thunderbolt by default, but there may be an option to add it using the Flex IO port. It's a similar story with the Lenovo Tiny models (they're popular here for server use and there's a good reference thread on them) - they don't have Thunderbolt by default, but it can be added using a specific riser and PCI-e card.

Some of the Minisforum models definitely come with Thunderbolt (might be listed as USB4, which is pretty much the same thing in practical terms). The MS-01 is the most "server-ish", but I don't think that one is the best choice for you, as I've seen people complaining about cooling issues with it and it's top feature having is SFP+ ports which you don't need anyway.

 

[ttabbal]

You could probably clean dust out of the NUC and re-paste the CPU. I've had to do that with some older setups, the paste dried out over years in use. Doing it on my Xbox One dropped fan speeds from max all the time to silent under load. Try some of the tricks the overclockers use for cooling, like better quality paste etc.. The latest hotness seems to be PTM7950. I like that it doesn't dry/pump out and can last basically forever, but performs up there with the best non-conductive pastes. I won't use the conductive stuff, I don't trust it.

Thunderbolt hasn't really been popular. I think Intel charges for it, so some manufacturers have avoided it. I would question if a QNAP can even saturate a 1Gbps link, or even a USB2 480Mbps one. In any case, if it can do USB3 or 4 you get the increased compatibility and no speed drop.

 

[RimBlock]

Thanks. The connector on the back for the DAS is TH3 / USB3 so can go the USB route if needed.

I cleaned out the fan which had a little dust but mot excessive. I will have a go repasting as suggested and see if that helps at all.

Transferring around 20GB from one DAS disk to another DAS disk caused the NUC to reboot after a few GB. This has happened when doing the same from my PC to the NUC in the past as well but not today for some reason. Guess it was running cool when I started. Feels like the thunderbolt chipset is generating a lot of head under load. Maybe just moving to the USB3 port may rectify the issue.

 

[Vorwrath]

If you haven't tried it already, it's worth updating your NUC to the latest BIOS just in case it's some kind of Thunderbolt compatibility issue rather than a heat problem. I think the link for your model is NUC10I7FNH - Support which was a nuisance to find (seems like the support stuff for the NUCs is on the ASUS site rather than Intel's now since they acquired the brand).

 

[RimBlock]

Under the "Power" settings in the BIOS, make a note of the original values, and then try setting: -
Package Power Limit 1 (Sustained): 25
Package Power Limit 2 (Burst Mode): 25
Platform PL1 Time Window: 0

Thanks will also give that a go.

Agreed with the Minis machines. They are also fairly expensive.

 

[RimBlock]

Ok, seems I maybe going down the wrong rabbit hole. It may not be temps that are the issue....

Ran HWInfo on the NUC and dumped the data to my PC.

Used remote desktop to connect and did a local copy from one DAS drive to a 2nd DAS drive (both HGST SATA). Copied maybe 45GB and then the RDP connection died and the feed from HWInfo stopped.

No temps reported on anything inside the NUC (NVME Boot, CPU, PCH etc) went over 62 DegC.

I tried changing to USB3 on the NUC side (DAS only has a TH3 / USB-C socket) and same problem. No temps from the DAS / its drives but not sure why issues in the DAS would kill the NUC and cause a hard reboot.

It still feels like a thermal issue but the readings are saying not. Short of taking it out of the case and blowing air at it from a standing fan, not sure what else to try.

Still investigating but any thoughts ?.

 

[Vorwrath]

I wasn't 100% convinced it was a thermal problem after you posted more details about the exact symptoms, which was why I made the previous post about updating the BIOS. I've got some other ideas, but there are several things it could be really, so you'll probably need to do some further investigation.

• Which DAS is it exactly? How is it powered? Could it be pulling too much power from the NUC ports and causing a reboot due to that?
• Some of the QNAP units have their own firmware, so you might be able to do a firmware update on the DAS itself.
• It could possibly be that the power supply on the NUC is failing, but nothing else you're doing with the system is putting it under enough load for this to show up.
• Is anything recorded in the Windows event log (most likely under "System") just before the NUC reboots?
• Run a test for a few hours with MemTest86 on the NUC to verify that the memory is operating correctly and no errors are recorded. It's not uncommon for weird and hard to diagnose issues to come down to memory instability.

 

[RimBlock]

Thanks for the suggestions.

Yeah, when looking through the readings from HWInfo I did wonder if it was a power brick issue. I didn't notice anything in the stats but didn't really look closely.

The QNAP unit is a TR-004 (USB-C not TB3 it seems - removes the most difficult requirement).

Updated the QNAP firmware from 1.0.2 to 1.4.

Same issue occurs copying the data .

Will check for BIOS updates tomorrow. The thing is this issue never used to occur.

In the event view there are a few errors.....

Failure status: A device which does not exist was specified.

Device GUID: {acdd875f-4f81-de3d-ed3e-d0eca3ed8359}
Device manufacturer: QNAP
Device model: TR-004 DISK01
Device revision: 6108
Device serial number: 5132313242323634
Bus type: USB

Application pop-up: Windows - No Disk : Exception Processing Message 0xc0000013 - Unexpected parameters

Looks like the NUC looses access to the DAS, errors with messages and then bluescreen reboots. Strangely the flush errors didn't appear on the last reboot when copying. I will do the copy from the actual PC tomorrow as well rather than via RDP.

 

[RimBlock]

Ok, so I think I am getting to the end of this rabbit hole.

The drive I slotted in to the DAS is a shucked WD80EDAZ. It seems these are likely air cooled and pretty sensitive to heat. Copying from one drive to another int he DAS box when the drives are next to each other is causing the loss of connection and, strangely, the NUC to restart.

Copying from the source to (HGST 10TB drive) to another HGST in the DAS works fine (tested with 200GB or so). Temp is around 43 degC after the copy finishes (QNAP software drive details).

Although not used for the test, the 8TB WD80EDAZ is sitting at 47 DegC...

Starting the copy back to the 8TB drive had the NUC reboot after a minute or two. After the NUC comes back the drive is reported to be at 49 DegC.

I have a 10TB Ironwolf drive sitting around so I think I will change this 8TB drive for that one and put the 8TB in my rack server with much better cooling.

Hopefully that will fix my woes...

Thanks for everyone inputs.

 

[Vorwrath]

I'm not sure it's going to be the drive temperatures. Most hard drives are rated to run up to something like 60C in their manufacturer specs. That's not particularly recommended and higher temperatures do cause an increased chance of failure over time. But the drives shouldn't really be dropping out at the kind of temperatures you're mentioning.

I think a good move would probably be to test the DAS connected to another machine and try the same copy that you're usually able to reproduce the problem with. See if you observe similar weird behaviour with the DAS dropping out or anything rebooting. That should at least determine fairly conclusively whether the problem is with the DAS or with the NUC.

 

[RimBlock]

The problem is still persisting although it takes longer to materialise now.

Ambient is currently 32 DegC. Drives are running at around 43 DegC with zero use in the QNAP DAS box.

I have now tried:

• Reversing the rear fan so it pulls air from the front and pushes it out of the back.
• Removing the metal on the inner rings of the fan vent (including the part covering the central fan hub) and replacing with a wire fan guard.
• Fully cleaning the DAS internals (bit of dust but not choked at all).
• Installed Hard Disk Sentinel for checking the disks and monitoring temps.

The drives are now sitting around 2 DegC cooler than before with no use.

Issue still persists.

Max temps for the drives (ever) are 47 DegC for the NVME OS drive in the NUC. Highest HDD in the DAS has been 57 DegC.

I think a good move would probably be to test the DAS connected to another machine and try the same copy that you're usually able to reproduce the problem with. See if you observe similar weird behaviour with the DAS dropping out or anything rebooting. That should at least determine fairly conclusively whether the problem is with the DAS or with the NUC.

Good call. Will try with my desktop PC this afternoon which also has USB 3. It is in a better ventilated area but should give a good idea.

TBH, if it is a cooling issues then fine but what worries me the most is that the NUC hard reboots... Surely Win 11 should be able to manage a disk dropout gracefully during a file transfer.

Fallback plan is to us a server in my rack which has an E3-1470v3 (25W) in it but it is a little more noisy as you could imagine.

Edit: Changed the thread title a little to reflect current understanding.

 

[RimBlock]

The plot thickens....

When hooked up to my main PC, copying data from one disk to another (around 117GB) killed explorer.exe after maybe 20 minutes. The machine still ran but I had to use task manager to start a new explorer process.

Drives didn't go above 47 DegC so, I guess, not a drive overheating issue.

A Bit Defender scan of the drives also caused the same issue.

So, current situation...

• Copying multi GB of data (drive to drive or from a 3rd non-DAS drive to a DAS drive) causes explorer.exe to crash after some time.
• Explorer.exe crash timing seems variable (sometimes within 5 min, sometimes up to 20 min).
• Drives don't go over 47 DegC on best case testing.
• This is a new(ish) issue (Multi-GB files have been copied over in the past without a problem).
• Unit is clean (well dusted with compressed air and contact cleaner).

Remining possible causes...

• Faulty USB-C cable (probably unlikely and don't have a spare to test with).
• DAS USB chipset overheating under load.
• Power supply not quite up to the load required any more.

Short of getting another USB-C cable and testing with that, I am not sure how much further I can go with this before calling it a day and going another route.

To be fair, the unit works fine serving movies / music / docs via the NUC (shared drives). Just moving large chunks of data is problematic.

 

[Vorwrath]

Yeah, it seems like a weird one. I think you're correct on the remaining things to check. It is possible that it could be a cable issue, to me that seems worth testing - it is feasible that a badly put together or damaged cable that isn't quite up to spec might cause intermittent errors.

You're right that the NUC shouldn't be hard rebooting just because a drive drops out or a USB device disconnects. For that reason I'm leaning towards it being something power related. If too much power is being pulled from the USB port for some reason, that would probably trip protections and cause a reboot. But why it would be doing that I don't know. The DAS PSU not providing enough power? Or the DAS getting stuck in some kind of weird failure state? Or some kind of intermittent short circuit happening in the DAS or the cable as it warms up?

The only other thing I thought of is to make sure you're not connecting the DAS through any kind of extra device like a USB hub (I'm guessing you're probably not since it's not been mentioned so far). Some of those aren't properly designed and can put too much load on the bus.

 

[RimBlock]

Thanks for the reply. It is always good to have others to bounce thoughts around with with these sorts of issues.

A few extra notes and observations.

• The connection to the DAS is direct with no hub in between.
• The cable is a short Anker USB-C cable and looks to be undamaged. There where no issues I can recall with moving large (100GB+) chunks of data around but I don't do that often at all.
• The issue as exhibited by my PC when I tried the DAS with it is that explorer.exe crashed but the machine did not hard reboot. The impression was that the NUC hard rebooted as it dropped the RDP connection and I had to login again after some minutes.
• One other thing popped up as I copied over on my PC and that was the Bit Defender alerted on Couch Potato installer (in my software repo). This is probably quite an old version but would have been sourced from the CP website. Not sure if they had any drama about viruses in their exe's before. The NUC had Norton on it previously when it was showing issues and I am not sure if it alerted etc. Copying other sections of my data excluding that part still show the same issues so probably not the cause.

I will probably move to the rack server for a NAS. I was thinking to go the TrueNAS or UNRaid route but the data is on NTFS formatted drives and my understanding is that neither of these can make it available so I can copy it off before building pools with these disks (I have some other spares to use as temp storage). Without pulling my PC apart and to connect those drives internally and then copy 44TB of data over a 1Gbe network link, Windows Server maybe a better option (I have available licenses).

Would love to hear the above is not correct if so. I have used TrueNAS in the past and am fine to go that route again if possible without a lot of hassle. I have a couple of external USB 3.5" cases a couple of the drives were originally shucked from knocking around so could try that but not the preferred option.

 

[Vorwrath]

It should be possible to mount the NTFS drives on Unraid using the unassigned devices plugin. TrueNAS used to do that as well via the "Import Disk" feature, but it sounds like this was removed and they now recommend mounting the disk on a Windows client and copying over the network instead. I expect there's probably a way around that using the normal Linux mounting methods, but might be fiddly and require extra packages etc. Either way you would need to first make a pool with some other drives, then copy the data off your NTFS drives, before you can do anything with those drives. And at least for Truenas you won't be able to then just add the drives you freed up to the original pool due to ZFS limitations.

I think the option of connecting the drives to your PC internally and copying over the gigabit network is actually not a bad one, and is probably what I'd do. At least that way the drives should be read reliably and you'll avoid whatever the DAS problem is. The network is limiting you to 125MB/sec, but you'd struggle to get more than about 200MB/sec on average out of an individual hard drive anyway (or much lower if there's a ton of small files). So the copy is going to take you several days whichever method you use.

If you set up the new network share that you want to move the files to, you can run a command on your Windows machine like

robocopy "D:\Files" "\\SERVERNAME\Files" /MIR /R:10 /W:10 /NP /FFT /LOG:c:\copy.log

to mirror the files to the new location. That's a better option than just moving files with the GUI, since if something goes wrong you can just run the same command again and it will pick up from where it left off.

 

[RimBlock]

Thanks,

I have been playing with Unraid all afternoon and have finally got an array created, added 2x 3.5TB U.2 caches and mounted (via terminal) one of the 4 DAS drives. I finally mounted the array and am no copying the data over locally between the drive and the mounted array share (SMB).

I also got one port on my 10GbE adaptor working (sort of auto working but in a pcie v2 x4 slot when it is a pcie v2 dual port x8 card).

Server is a bit noisy but it is working for now. Ended up using 2x 14TB drives for parity which is a bit painful but have quite a few 8 --> 12 TB spare drives which used to be used for Chia so at least they are doing something of use.

Thanks for all the help...

 

[RimBlock]

Sooo.....

Tried Unraid but whilst I could mount the NTFS disk and then copy some data over, it still failed after around 20 minutes. The 17h parity setup (2x 14TB drives for parity) has put me off. The server I have is also pretty noisy reducing any Wife acceptance factor.

I decided to move over to TrueNAS Scale on my main (currently unused) storage server (dual Silver 4210s and 512 GB ram). The server is X11 based and much quieter. After installing (had to use the IPMI Java console and remote storage as the USB's built with Rufus in wither ISO or DD modes would not boot) and building some vdevs and datasets, it came to light that the dev team removed mounting ntfs disks in the current version.

Had a spare 10GbE controller which I put in the server and then pulled my PC apart and connected a drive which refused to spin up. Remembered something about drives not working with PC sata connectors but had an adaptor which then worked. Tried copying via Windows explorer which also failed after maybe 20 minutes.

Finally the fix was robocopy and the command you supplied above. Has been working like a champ with no failures for multiple TB now. Thanks so much for that.

Now looking at apps. Liked the Unraid app availability and interface but 17h before I can first time safely use the storage is just too much for my use case.

Suspect all will now work out and although this is no longer a low power option, it is far better for redundancy.

Thanks for all the previous suggestions.