SER9 AI 9 HX370 NVME drive disappeared

thorpemark

I have the SER9 AI 9 HX370 with a Samsung 4TB 990pro as the OS C: drive and a 4TB 990pro w/heatsink (shelled) as the D: drive. It set up fine with WIn11pro and some apps. Lastly, I added ChannelsDVR to use the D drive to store recorded shows. That also went fine. I used Samsung’s Memory Manager and tested and performance rated.. all worked. I tested using Channels DVR software from a TV using the Channels app. It recorded and played back content fine. I had to step away from that project for a week or so but the system was on the whole time and recorded maybe 2 new shows in the interim.

Then today I tried to access the shows from a TV. It looked like it should.. listed the shows etc. But what was happening was it could no longer access files on the D drive. I assume some data is stored with the app on the C drive and it could tell me what shows were out there.. but trying to play live video or a recording meant accessing the D drive and that failed. It showed as “connection” error.. which, it turned out, was misleading.

Using the self-test inside channels everything looked great except for a red flag for recording to the D drive. I checked teh properties of the folder on the D drive.. it showed as read-only (which I figured, change that and we’re good.. but the recursion of files and folders turned up missing file after missing file. Saw a suggestion to reboot to bios and check the drive status there before continuing.

BIOS no longer sees the D drive.

So.. either the shelled-“990pro with heatsink” (that’s how the BIOS had shown it) is bad, or it really shouldn’t be used without its heatsink, or the NVME slot itself is bad. But since that slot looked fine when I set this all up, I am uncertain how to proceed.

I will try replacing 1TB SSD as the D: drive (one of the 1TB SSDs that came with the SER9) but it took days for this failure to show up. I could have Channels recording to it. I have another non-heatsink 4TB SSD on the way (this is getting expensive) but testing that will be in the same cycle of many days of testing.

I will try the possibly-faulty SSD in an on-order external enclosure or try to find a PC here that has an empty NVME slot to try it with .. and hopefully one where the Samsung heatsink can be put back on for the test.

Supposedly, removing the heatsink voided my warranty, but I cannot imagine that the heatsink 990pro is any different internally than the non-heatsink version. And yet, BOIS knew I was using a heatsink stick.. so I have no idea.

Does anyone have any advice for me? I’d love to be able to tell what was faulty.. the SER9 or the NVME stick.

thorpemark

update: after cold shutdown and some time.. just rebooted to bios and now it sees the 2nd nvme drive. while still in bios I am running all the self tests.. the short ones passed and the long version running now.

thorpemark

Last update (I hope).. I grok’d a few questions and I got it to suggest something interesting… that is in the Power Plan settings.. advanced settings.. it showed setting for turning off hard drives (set to 0 or NEVER) and under PCI Express there is Link State Power management and you set that to OFF.

This second part was interesting because the symptoms it suggested was exactly what my problem was. My OS NVME drive was functioning perfectly.. but I only used the 2nd drive, the D: drive, so far for DVR recordinings… and after several days of no recordings.. somewhere in there it it turned off for power savings and did not come back on. So, the settings above should solve that… but when will I know for sure? In a few days maybe. Anyway, I was able to cancel my 4TB 990pro non-heatsink replacement so I hope this is it.

Guess I have to come back when I know. Maybe someone else will encounter the same issue and this could help them not jump off a cliff.

support6

thorpemark
Great detective work! The PCIe Link State Power Management setting is a sneaky culprit here—NVMe drives (especially secondary ones) can sometimes get “stuck” in low-power states if the system aggressively manages power, leading to them dropping off the bus entirely.
Your fix makes total sense: Disabling that setting prevents the PCIe link from entering low-power modes that might confuse the drive, and setting “turn off hard drives” to “never” ensures the OS doesn’t treat the NVMe like a traditional HDD that can be idled.
To confirm it’s resolved, keep an eye on the drive over the next few days—if it stays detected in BIOS/OS and Channels DVR records/plays back without issues, you’ve nailed it. If it fails again, you can circle back to testing the drive in another system or enclosure, but this power tweak feels like the right call given the symptoms.
Fingers crossed this is the end of it—let us know how it holds up!

thorpemark

support6 Just replying to you to maybe trigger a review of my latest on this

I should also add that yesterday it failed in less than 2 hours of inactivity.. when I did a cold reboot after finding teh error the previous night I quickly st up some recording for 2 hours away and sometime between setting that up and the recording I lost access to the drive.. so recordings failed and, again, I couldn’t play live TV (which would buffer to the missing D drive).

And that brought me to asking Grok about tracing the system logs and teh message above this one.

This is what Grok says I should set in BIOS.. of course Beelink’sBIOS language is different and it can be difficult to figure out where everything is..

Next Steps to Confirm Cause:
BIOS NVMe Power Settings:
Reboot into BIOS/UEFI.
Check for NVMe-specific power settings (e.g., ASPM, PCIe low-power modes, or NVMe idle timeout).
Set PCIe/NVMe to maximum performance and disable any low-power states (e.g., L1.2 substates).

Okay.. current state: The drive remained accessible for a long time now so, again advised by Grok, I stopped the powershell script that kept accessing D and replace the log with a Windows ADK command.. wpr -start DiskIO -filemode

Supposedly, that looks for the system making calls to D: but does not keep the drive active on its own. I will check for access from a TV Channels client every so often until it fails.. which I fully expect will happen.

thorpemark

Well.. the OS changes may have helped.. but I had to get into BIOS as well. If it works this time I’ll review the BIOS settings I changed and post.'

Yes.. this means the D: drive access disappeared again in exactly the same way. Since computers was on, the software thought the D drive was there.. its info on stored recordings was on C: but pointed to D:. And live viewing would want to buffer to D: (I assume) because that failed as well. And, like last time, warm restart to BIOS saw BIOS not see the 2nd NVME drive, only listing the 1st OS drive. But a cold reboot afterward returned both NVME drives to be seen by the BIOS. I changed 2 BIOS settings.. one in ACPI and anuother, iirc, off

Manual PPT limits set to (they started at 0.. grok suggested a non-zero value might be best and recommended teh below based on the CPU):
Fast PPT to 50000mW
Slow PPT to 35000mW
Slow PPT Time Constant to 5 seconds

also
ACPI Auto is disabled
Hibernation disabled (had been enabled)

I think that is the only 2 I changed but grok did suggest I look for any other Sleep Mode settings.., wanting me to make sure S3, S4 and S5 sleep modes are disabled and said somehting like is S3 is enabled it may trigger elevation to S4 or even S5 in some settings.

In my software tests, I recorded 4 shows at teh same time around noon and they all worked, but this was while I was setting things up with teh new BIOS settings, so it probably did not have a lot of time to fall asleep. I set up more recordings for about 4 hours from now, 7PM my time.. so sometime later tonight I will go fire up a TV app and see if it can access the new recordings.. or the old ones or live TV since they will all require access to D: drive.

I have no idea what kind of performance I can expect if someone is using the PC to browse or read emails while it is trying to record shows and someone TV in the house is playing a recording.. but I have been impressed. I was worried seeing some choppy video but that seemed to be totally related to a software setting in playback.. something about shared tuners. I really like what I have seen from the miniPC so far, so I think you have a winner here.

Also FYI… really like your BIOS.. you guys have made an effort. While I don’t have any minisforum minPCs I do have a couple of their miniITX mobos.. I am using one to type this now and another is in a Truescale NAS (which I may try to get running ChannelsDVR but it is mostly for personal photo and video backup. Their BIOS is weak sauce. Example: Samsung Magician wanted to update FIRMWARE for the NVME drives I have in this one and I couldn’t do it. Some trusted computing settings were involved where I had to do a round-about update through booting a USB stick.. that is normally somehting to update a mobo BIOS.. not a memory stick or hard drive FIRMWARE. And minisforum was no help. It seems like minimal effort is put in to their bios so that is why I decided beelink was worth a shot and I have to say it has worked out so far.

thorpemark

Unfortunately, this too, has failed. Same result. I was waiting a few hours for Channels to record somehting at 7pm and the drive must have gone inaccessible before then.

The NVME drive passes all BIOS NVME tests and passes the Samsung Magician tests.

I suppose there must be some kind of event logging I could use on Win 11 pro to determine why the drive dispapears but it is beyond my knowledge level. Guess I’ll try to grok that.

thorpemark

Oh boy…. latest test a success so far, but I fear the tool used to observe the problem is causing the problem to not happen.

I asked Grok for a method to track when I lose access to the D: drive so I can better target system logs. This was because its advice for system logs centered around the ERROR (trying to access D: and not being able to) rather than tracking commands to tell the drive to hibernate or sleep. So when I used the suggested log filters to find the error I only saw what I knew already.. which was that the ChannelsDVR app attempt to record programs at 7 pm.. after maybe 5-6 hours of inactivity.. fails. As did any further attempt to view live TV thru Channels (which would buffer video to D:) .

So Grok suggested this admin-level Powershell script:
$logFile = “C:\nvme_log.txt”
while ($true) {
$status = if (Test-Path ‘D:\’) { “Accessible” } else { “Inaccessible” }
Add-Content $logFile “$(Get-Date): D: drive $status”
Start-Sleep -Seconds 10
}

Pretty simple.. checks if it can access D: and writes to a log file on C:

After setting this up I checked teh log file regularly and saw it growing every 10 minutes with the line that says it can still access D:.

I set up recordings in the near future.. maybe an hour way.. and then some for 3AM.. mimicking the hours of inactivity followed by ChannelsDVR recording 4 TVE (TVEverywhere) feeds at one time.

It worked.

Good news, right?

Well.. what if the powershell script checking teh D: drive every 10 minutes is the reason it never went to sleep (assuming the sleep followed by failure to awaken is the problem)?

Yes, if the D: drive remains accessible at various times all day today then I am fairly certain we can say it is the sleep/awake issue and not some hardware failure.. hmm.. unless the drive controller is a hardware problem in failure to wake up.

I cannot just run that powershell script on boot up every day to force this to work. There has to be a better way to make sure the drive is never asked to sleep… again, assuming that is the issue.

Anyone have a complete list of BIOS settings and OS settings to prevent such sleep modes for NVME drives?

support6

thorpemark
1.Windows Power Settings:

Open Power Options > Change plan settings > Change advanced settings:

PCI Express > Link State Power Management > OFF

Hard disk > Turn off hard disk after > NEVER

USB settings > USB selective suspend > DISABLED

Processor power management > Minimum processor state > 5%

2.Device Manager Tweaks:

Under Disk drives, right-click each NVMe > Properties:

Uncheck “Allow the computer to turn off this device to save power” (Power Management tab)

Update drivers > Browse my computer > Let me pick > Select Standard NVM Express Controller

Samsung-Specific Fixes:

Install Samsung Magician to:

Check/update firmware (critical for 990 Pro stability)

Run Full Diagnostic Scan

Enable Over Provisioning (7-10% of drive capacity)

3.Validation Test
After applying all above:

Stop all monitoring scripts

Simulate idle:

powershell
powercfg.exe /requests /output C:\powerreport.txt
Check powerreport.txt for background processes preventing sleep

Let system idle 6+ hours, then attempt D: access

4.If Issue Persists
Hardware Isolation Test:

Swap the two NVMe drives (move OS to secondary slot)

Test with a different drive brand (some 990 Pros have L1.2 state bugs)

Use M.2 to USB enclosure ($15) to rule out slot issues

HenryPalos

Having drive detection issues too, huh? I once had a similar problem after removing an NVMe’s heatsink for a tight fit, causing thermal throttling and eventual failure. Did you check if the drive felt excessively hot? Anyway, off to make a sundae. I need to get back to Papa’s Freezeria .