AOOSTAR WTR Max vs Minisforum N5 Pro

After seeing Wendell’s video I was about to click buy on the AOOSTAR WTR Max, but then I discovered the Minisforum N5 Pro. Both support ECC, Oculink, have a 10GB nic, and a few M.2 slots. The N5 has one fewer 3.5" bay but adds a PCIe slot. For me, the most significant differences are the CPU and NICs.

Here’s what I’m trying to figure out: How should I expect the idle power draw of the two to compare?

The N5 Pro has a more powerful CPU with more cores (Ryzen AI 9 HX PRO 370), but AMD lists its default TDP as much lower (28W) than the AOOSTAR’s CPU (45W Ryzen 7 PRO 8845HS). The N5 Pro has a single Marvell AQC113 copper 10GB Ethernet port (and a Realtek 5GB) instead of the dual Intel X710 SFP+ cages (and dual Intel 2…5GB) on the AOOSTAR–surely skipping the SFP+ saves a lot of power at idle.

NASCompares did a video and writeup about the two, but the power draw comparison is not particularly helpful (he compared idle spinning drives in one machine to no drives in the other or maybe I’m missing something). A comparison with the same M.2 and 3.5" drives in both would be more definitive.

I also like that the N5 Pro doesn’t put the M.2 in a hotswap bay. My homelab isn’t in a closet so someone could theoretically walk up and say “ooh what does this eject” and crash the OS (less of an issue for the SATA drives).

The NASCompares video implies that the screen on the AOOSTAR is accessible via an internal network switch or something and it has its own fixed IP. Maybe it shows up as a USB NIC in the best case, but if it really is somehow switched I don’t like the idea of adding a mostly unknown IoT device into the mix.

I would be interested to see Wendell’s take on the two of these head to head. Of course the Minisforum is more expensive but it might be the best compact-homelab-in-a-box (when you need both compute and storage) I have yet to find.

the display is USB in fact. the x710 wasn’t designed to idle quietly but it’s the high tier option for 10gb vs aqc IMHO.

I like the ai max CPU more even at slightly lower power but they’re pretty close real world. the aoostar might have a slight edge in all core heavy workloads tbh.

by default the tips for using indicates that the software control interface can be accessed by a fixed IP but until I installed the sub-screen monitoring software app and ran it I was unable to access the control interface.

Here is the instructions for it if you want to read it.

Operating instructions are currently being further refined…txt (2.0 KB)

1 Like

Here are the instructions in the file… translated

Linux Usage Instructions

  1. Unzip the file, navigate to the AOOSTAR-X-linux folder, right-click in an empty area, open a terminal or cd to the AOOSTAR-X-linux directory. Note that the path cannot contain special characters or Chinese characters.

  2. Install and start the service. In the terminal, type the following (root privileges and password required):

pve requires additional commands:

sudo apt update
apt install sudo

Start the secondary screen service

chmod +x ./create_service.sh
sudo ./create_service.sh

  1. Uninstall the secondary screen service. In the terminal, type the following (root privileges and password required):
    chmod +x ./remove_service.sh
    sudo ./remove_service.sh

  2. Install lm-sensors + it87 driver (displays fan speed and other sensor information, optional)

Install lm-sensors

sudo apt update
sudo apt install lm-sensors

Install dependencies

apt update && apt install -y pve-headers proxmox-headers-$(uname -r) dkms flex bison

Install it87

dpkg -i it87-dkms_1.0.63-1_all.deb

Note: If errors occur, you need to resolve dependency or conflict issues.

Uninstall lm-sensors

sudo apt remove lm-sensors

Windows Usage

Double-click AOOSTAR-X-Setup.exe to install. Windows or Windows Server must have all drivers installed; otherwise, errors will occur and the program will fail to start.

General Instructions:

Import the theme from the compressed package; no decompression is required. Windows themes use different sensors than Linux themes, so you need to modify the sensor settings yourself.

The admin port is 5123. The default username and password are admin 123456. Please change the password promptly to avoid security risks.

Display Home Assistant Data:

Open the Monitor3.json file and modify

“ha_url”: “192.168.1.1:8123”,

“ha_token”: “eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiI5ODRmYzJlOTliNzE0MzI4OTA4ZjhlMmZkZTdkY2Y0MSIsImlhdCI6MTc0OTM1NzU2OSwiZXhwIjoyMDY0NzE3NTY5fQ._rAaiSRhUKQx2X7U7KrdlxYkD0VBx2t6tcPfEQLoeTo”

Save this, and rerun AOOSTAR-X. You should then find data starting with HA_ in the sensors.

ha_url is the Home Assistant’s external address, and ha_token is the token.

To display AIDA64 data: File → Settings → Hardware Monitoring Tools → External Programs → Check “Allow Shared Memory” → Select all data at the bottom → Save

1 Like

Don’t confuse x710 da2/4 (sfp+) with x710 t2/t4( copper).
sfp+ will run full speed at <4W
copper will run at ~15W just keeping link up
both should and do idle(no sfp+ adapter nor cabble plugged in) at ~1W.

aqc113 is exceptionaly efficient but still draws ~8w with 10G link up and approx 10-11W at full speed.
and yes x710 da2 is tier above aqc113, for being fiber :wink:

on a side note:
JMB585 (sata controler usen on n5 pro is infamous for failures under heavy load. Bizzare choice if you ask me, asm1166 costs $2 more and is much more reliable. deal breaker for me.
I would love to see n5 pro tested with 16TB HDD resilvering in full raid10 or even raidz2 array.
my 2 cards based on that chip failed during resilver. in normal low usage used to throw crc/dma errors once a week.

2 Likes

this is good input – I too have had failures with that chipset but also successes in some long-running boxes. I should set that up and test it

@GigaBusterEXE gab some 16tb drives, fully populate and install proxnox on the n5 please

so far the Minisforum n5 pro seems solid:

this is resilvering onto a blank drive after the pool being filled to about 13tb/37tb:

 zfs list
NAME   USED  AVAIL  REFER  MOUNTPOINT
tank  13.1T  37.6T  13.1T  /tank

c1:00.0 SATA controller: JMicron Technology Corp. JMB58x AHCI SATA controller

This sata controller does run hot, and will go sideways if it gets too hot, and is not as good as other JMicron controllers, but so far knock on wood hasn’t done anything untoward. This is the second back-to-back resilver with a fresh drive in a z1 pool. @Sasquatch1

4 Likes

Long time lurker, but I figured I’d reply since I’m using the Minisforum N5 Pro and having issues with the JMB58x AHCI SATA controller. I have a pretty simple setup - using 4 SATA 20TB drives in a mergerfs configuration with the 5th as parity (snapraid). Every so often I will see this kernel message:

[24591.459657] ahci 0000:c1:00.0: Using 64-bit DMA addresses

Followed shortly by something similar to this if I try to access data on any of the mergerfs drives:

[28399.549008] XFS (sde1): Metadata corruption detected at __xfs_buf_ioend+0xc7/0x160 [xfs], xfs_inode block 0x7ba7f92c8 xfs_inode_buf_verify
[28399.549167] XFS (sde1): Unmount and run xfs_repair
[28399.549180] XFS (sde1): First 128 bytes of corrupted metadata buffer:
[28399.549195] 00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549214] 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549233] 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549253] 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549274] 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549295] 00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549318] 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[28399.549340] 00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................

Each time this happens, a different drive will complain, so I don’t think it’s an issue with the hard drives. Plus if I use these drives in a different server I have (using a Gigabyte B650 motherboard) I never see these issues.

I’m currently using the Minisforum N5 Pro with CachyOS. Not sure what else I can do at this point - at the moment I’ve ramped up the SSD cooling fan to see if that has any effect. I’ve only had the N5 Pro for about a week or so at this point. Have you been seeing issues on your side?

1 Like

so far it’s still steady but this is exactly the sort of thing I’d expect if the controller is a bit sus

Is there anything to mitigate such issues? I’ve tried ramping up the cooling to no effect. Tried setting libata.force=noncq (shot in the dark - figured I’d at least try it). I’ve taken out each drive, cleaned the backplane, etc. I’m not even doing that much activity on the drives - just downloading a 60TB file at about 5MB/s and yet within a few hours I’ll see the issue come back up.

When it happens I seem to lose access to ALL the SATA drives (the nvme drives are fine). The only way to get the drives back is to reboot. This only seems to happen when doing writes to the disks, not when doing reads (used it for days without writing and was fine).

If there’s no way to mitigate this, I think I’ll have to return mine.

1 Like

Sorry that should say 60GB file, not TB.

have you tried the usual things like disabling pcie power management (pcie_aspm=off) etc?

Yeah tried pcie_aspm=off today (not sure what else I could turn on/off apart from what I already posted earlier) and unfortunately it still errors. It usually can manifest anywhere from 30 min up to 4 hours. Tried the LTS kernel as well.

I’m not sure what else to try…can’t believe they would sell this as a NAS system if the SATA controller is that finicky. Are you doing your testing with their OS? I tried looking in the BIOS to see if there were some options to toggle, but not much there related to the SATA controller that I can see…

No, I was using linux natively with a recent kernel. When you RMA your device, be sure to tell them you think the sata controller is flaky and if they can just swap that, it would be ideal. The PCB the sata controller is on is maybe easy swapped too, so they might just send you one of those.

You might be able to eyeball it and see if you see any sus soldering and/or post some pictures here. If it is the controller it is not an uncommon failure mode. Also the gold fingers where the motherboard slides into the connector maybe worth looking at to make sure those are clean and shiny. A bad connection could also lead to a similar type of signal integrity error.

I reached out to support and someone named Jason replied asking if there was any PCIE device installed on the server, and if so, if I could remove said device and test again.

I do have a low profile dual 10G NIC installed in the PCIE slot. The NIC ports are passed to an OPNsense VM I have running on the box for WAN and LAN, and then I use one of the server’s onboard LAN ports as a means of connecting the server itself to the LAN network (plugs into a switch).

I’m a bit worried because they seem to be signaling that you can’t use the PCIE slot AND have consistent behavior with the SATA controller? I didn’t see anywhere in the documentation that this was a no-no though.

In any case I can test that this weekend and see if the behavior changes. But even if it does “solve” the problem, I can’t use the server this way as I was trying to have a single all-in-one solution for the storage and network functionality (my old NAS server did this fine with the Gigabyte motherboard).

I suppose I could split out the OPNsense VM to its own dedicated device, but then it kinda defeats the purpose of having this beefy server imo - I was really hoping to take full advantage of the hardware and keep the footprint small (I don’t have the room for a dedicated rack in my house).

Is there something I missed in the docs? Am I “doing it wrong”…?

could try explicitly setting the pie device to gen3. could be interference. it’s worth experimenting. if it does work without the pcie card it’s not necessarily the end of the line for the pcie card

I won’t be able to take out the NIC until later this week, but in the meantime, do you know which BIOS option controls the main PCIe slot? The ones I see seem to be for the NVME drives. Here’s what lspci shows for each port: LnkSta: Speed 5GT/s, Width x4 (downgraded)

Tried to search the net for documentation on this but not finding anything…

Found out via manual testing that the first PCIe device listed under PCIe options is the main one. Despite that, I took out the PCIe network card and after testing a bit this evening, I still see the same error manifest. Sigh…will see what support says about this but I think I’m at the point where I’m just going to try and return the device. I’m pretty bummed out at this point.

Thanks for all the assist on it.

MINISFORUM takes this issue very seriously.

MINISFORUM has also received this feedback and will attempt to replicate the problem on our development team using the same OS version.

Could you please provide your system version(OS, Version, Linux kernel Version), what HDD and other PCIe devices you have plugged in?

So wanted to do some more testing before I pointed at the N5 Pro as being the culprit.

I took my old NAS system out of the closet (let’s call it the Scrappy NAS - it’s a custom mini ITX build), installed CachyOS on it just like I did on the N5 Pro (same kernel, etc), and then proceeded to set everything up as I had it setup on the N5 Pro.

Took all the hard drives in the N5 Pro and moved them over to Scrappy, set them up with the same mount options, etc, just like I had them in the N5 Pro. I didn’t even reformat or anything - same filesystems, same existing data on them, etc. I then proceeded to start downloads of some 60+GB files and left it running all day/overnight.

Came back this morning, and…all is well. All the files downloaded (each file downloaded at around 2-15MB/s rate - my internet sucks), no loss of data, no complaining from the filesystem of the SATA drives, no loss of access to the SATA drives.

I DO see the same message in Scrappy’s kernel logs that I saw in the N5 Pro:

[15270.954311] ahci 0000:0f:00.0: Using 64-bit DMA addresses

But in Scrappy’s case it’s just a normal kernel message - nothing bad happens. In the N5 Pro’s case, every time I saw that message, I’d lose the SATA drives.

It should be pointed out that if I downloaded smaller files, say 8GB and below, on the N5 Pro, then I’d never see the issue manifest. It’s only when I get into larger file sizes that I would see issues with the N5 Pro.

Anyway, I think at this point it’s pretty conclusive that I have a defective N5 Pro. I COULD ask for a replacement, but honestly I think I’ll ask for a refund for a couple of reasons:

  1. The SATA issue of course…

  2. Lack of control over some things in the BIOS and with the hardware config - it’s not BAD or anything like that with the N5 Pro, but when moving back to Scrappy, I have a lot more control over things, esp the hardware config (since I built it), and I didn’t realize how much I missed that.

  3. Do I really need such a beefy server like the N5 Pro? (the answer is no, but I will miss it…)

  4. The fan noise of the N5 Pro is a slight issue imo - you can’t completely eliminate it. It’s what you get for cramming all those things into such a small form factor

One silver lining out of all of this is that I have new motivation to move my OPNsense router to a dedicated small NUC that I have laying around. My family was getting tired of the network going down every time I was troubleshooting with the NAS server. I saw the recent video over making a HA OPNsense setup…I might just try that (not that I need such a setup, but it would be fun to experiment with).

Once again thanks for all the assist for those who chimed in. Shame that I have to send the N5 Pro back - I wanted it to work out. It’s a neat little box in concept.

1 Like

@buddwm
Hi Where are you from?
I means. What is your current AC voltage? And what kind of HDD(model number) you using?