One of my three drives will go in an empty 5,25 bay. I know there are several brackets on the market, but I was hoping that people who already use this could give some recommendations?

Thanks in advance!

I wasn't able to find any low load tests for this PSU before I bought it, so I thought this post could be useful for anyone thinking about getting the HDPlex 250W GaN PSU. To help with this, I compared it to the Be Quiet! Pure Power 13 M 550W, which is already recognized as a reliable option for power efficiency. You can find a test review for the latter here. The most interesting figure related to low load efficiency is this one:

Obviously, it reveals that the Be Quiet PSU has an efficiency of 69.6% at 10W load and 77% at 15W load. As I don't have professional measuring equipment, I simply compared the low power loads for a time span of 10 minutes starting 20 minutes after booting. Both PSUs used the exact same hardware, that is:

• Unraid 7.2.2

• Mainboard: Gigabyte C246M-WU4

• CPU: Intel i3 9100

• Memory: 2x 32GB Kingston DDR4-3200 ECC RAM

• Cachedrive: 1x WD Black SN850x NVME M.2 SSD

• No hard drives!

• no running docker containers

• no active case fans, only the CPU fan was running at 400rpm

• the system reaches C10 state

The results look like this (measured with Fritz!DECT 200):

The HDPlex consumes approximately 2,2 W less on average in my idle scenario. Utilizing the same hardware as previously mentioned, along with the HDPlex PSU, my hard drives (5x Toshiba MG08 16 TB SATA HDD, 3x WD WDC161 Gold 16 TB SATA HDD), and running Docker containers, I achieve around 12.8W load during idle when all hard drives are spun down. This is an impressive outcome, IMO.

Plus, I can confirm that the HDPlex PSU can power my eight (!) hard drives without any problems if you don't need the PCI-E power port on the PSU. Four drives are powered by the power cable connected to the SATA power port on the PSU and the remaining four drives are powered by a second custom cable that uses the orange pins in the below figure. The +5V is splitted from the SATA power port, which is no problem since the PSU is a single rail one.

Before I go buy one of those Realtek RTL8127AT cards and just use another SFP+ to RJ-45 tranceiver in the switch to do 10G over RJ45 ethernet, has anyone experienced this? The link works and I can access the dashboard, ping, and all that. I've bridged the two SFP ports on the Unraid server together as br2...here's just a snippet.

By the way, each port on the X710-DA2 is in its own IOMMU group - 18 and 19 - straight by itself with no PCI bridge. I have this card in my GPU slot - so PCI-Express 4.0 16x, more than capable of handling both power and link of the PCI-Express needs.

Jul 12 19:18:31 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Down

Jul 12 19:18:31 PLEXARY kernel: br2: port 1(eth2) entered disabled state

Jul 12 19:18:31 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None

Jul 12 19:18:31 PLEXARY kernel: br2: port 1(eth2) entered blocking state

Jul 12 19:18:31 PLEXARY kernel: br2: port 1(eth2) entered listening state

Jul 12 19:18:33 PLEXARY kernel: br2: port 1(eth2) entered learning state

Jul 12 19:18:35 PLEXARY kernel: br2: port 1(eth2) entered forwarding state

Jul 12 19:18:35 PLEXARY kernel: br2: topology change detected, propagating

Jul 12 19:26:24 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Down

Jul 12 19:26:24 PLEXARY kernel: br2: port 1(eth2) entered disabled state

Jul 12 19:26:25 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None

Jul 12 19:26:25 PLEXARY kernel: br2: port 1(eth2) entered blocking state

Jul 12 19:26:25 PLEXARY kernel: br2: port 1(eth2) entered listening state

Jul 12 19:26:27 PLEXARY kernel: br2: port 1(eth2) entered learning state

Jul 12 19:26:29 PLEXARY kernel: br2: port 1(eth2) entered forwarding state

Jul 12 19:26:29 PLEXARY kernel: br2: topology change detected, propagating

Jul 12 19:40:48 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Down

Jul 12 19:40:48 PLEXARY kernel: br2: port 1(eth2) entered disabled state

Jul 12 19:40:49 PLEXARY kernel: i40e 0000:01:00.1 eth2: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None

Jul 12 19:40:49 PLEXARY kernel: br2: port 1(eth2) entered blocking state

Jul 12 19:40:49 PLEXARY kernel: br2: port 1(eth2) entered listening state

Jul 12 19:40:51 PLEXARY kernel: br2: port 1(eth2) entered learning state

Jul 12 19:40:53 PLEXARY avahi-daemon[2750123]: Withdrawing address record for 192.168.1.13 on br2.

Jul 12 19:40:53 PLEXARY avahi-daemon[2750123]: Leaving mDNS multicast group on interface br2.IPv4 with address 192.168.1.13.

Jul 12 19:40:53 PLEXARY avahi-daemon[2750123]: Interface br2.IPv4 no longer relevant for mDNS.

Jul 12 19:40:53 PLEXARY kernel: br2: port 1(eth2) entered forwarding state

Jul 12 19:40:53 PLEXARY kernel: br2: topology change detected, propagating

Jul 12 19:40:56 PLEXARY ntpd[2749596]: Deleting 4 br2, [192.168.1.13]:123, stats: received=0, sent=0, dropped=0, active_time=1919 secs

Jul 12 19:41:04 PLEXARY avahi-daemon[2750123]: Joining mDNS multicast group on interface br2.IPv4 with address 192.168.1.13.

Jul 12 19:41:04 PLEXARY avahi-daemon[2750123]: New relevant interface br2.IPv4 for mDNS.

Jul 12 19:41:04 PLEXARY avahi-daemon[2750123]: Registering new address record for 192.168.1.13 on br2.IPv4.

Jul 12 19:41:07 PLEXARY ntpd[2749596]: Listen normally on 5 br2 192.168.1.13:123

It seems to drop the state, go down for a brief second, and then go back up, every like 10-20 minutes right now. I haven't tried the other SFP+ port on the Unraid's X710-DA2, but the fact that its working fine right now makes me curious - why is the link going down at all? my pfSense box is the same in every way - except running pfSense, I mean hardware. Its link doesnt go down at all, ever.

Been rock solid.

If someone can help me, my city has started doing a lot of work on the roads and I had just in the last week 3 outages and in the last parity check I had my first error. I`m looking to buy a cheap UPS that has enough capacity to just turn off my nas safely .

I run a i5 12400 , 4hh , 1ssd and no GPU

Cheap options that i have are :

UPS nJoy Keen 600 USB 600 VA

and

UPS nJoy Token 600 600 VA

Does someone know if they work with unraid?

If not i will probably go with a UPS APC BX750MI-GR 750 VA as i`ve been reading that APC UPSs work with unraid but is double the price of the other 2

I was recently gifted a NetApp DS4246 and I’m looking for the most budget-friendly way to rackmount it.

I may end up stacking up to 3 of these units, so I’m thinking I’ll need at least a 12U+ rack solution. I’m not looking for anything fancy — just something affordable, sturdy, and practical for a home lab setup.

For those who have experience with these shelves, what would you recommend? Are there any cheap rack options, DIY solutions, or things I should look out for when choosing a rack?

Thanks in advance for the help!

In Unraid I have activated it as APC and USB. But unfortunately it is not recognized. Has anyone managed to fix this?

1. Single 12 volt rail.  A subsequent figures refer to the 12 volt rail.

2. The minimum capacity that can power your build. Any more will just waste power. All drives will be in use during startup, shutdown, parity check, parity build, failed drive emulation, and drive rebuild, but startup requires the most power.

3. 2 amps (24 watts) per green drive and 3 amps (36 watts) per non-green drive on the 12 volt rail.

4. 5 amps (60 watts) for the motherboard on the 12 volt rail.

Please make comments below and I will incorporate changes in to this post.

Does anyone run a higher drive count server that's low power? I'm not even sure what's reasonable when it comes to power draw.

The current server is 4U, has 32 drive bays and is currently running:

2 x AMD Opeteron 6380

64GB Memory

20 x 3.5" drives set to spin up individually and spin down when not in use.

3 x SSDs

NVIDIA Quadro P620

2 x 1400w redundant power supplies

It's primary use is Emby and unless I'm doing a parity check, it rarely spins up more than a couple drives if we are all watching separate movies.

I haven't put a meter on it yet, so I can't even say where I'm at right now, but with all those fans spinning 24/7, it can't be low.

I forgot that my UPS shows usage. I have one power supply plugged into the UPS right now, because oddly the other failed yesterday, and it's showing 385 watts. I have a replacement on order.

GitHub

I love my Asustor Lockerstor 6 gen 2 (AS6706T) for my primary NAS (running the ADM OS (not unraid)), however with drive prices being expensive nowadays I decided to use another Asustor AS6706T as a secondary offsite NAS but I wanted to use a bunch of the older drives I collected before populating my primary NAS with drives instead of paying the high prices of new drives today.

After installing Unraid 7.3.1 I noticed some of the common problems others have experienced regarding LCD panel not usable out of the box and blinking green led on front and none of the 6 drive bay led lights working, also lack of a fan curve based on hardware temps. This depends on the Asustor Platform Drivers Unraid plugin being installed.

I thought I'd take a stab at "vibe" coding up some scripts that enable some of that functionality back to working under Unraid. I'm new to Unraid so this is my only experience with it and I LOVE the boot from internal storage now instead of the external USB thumb drive. There is a 'claude-persist.sh' script that is disabled by default and is only a helper in the event you want to persist claude cli on your system after reboot to assist with troubleshooting. I would not recommend running an AI model on your storage NAS unless you know what you are doing

Just a FYI this was a personal project and try this at your own risk, but it was helpful to install Claude Code CLI after Unraid install to help find all the led sensors and help define a fan curve.

I only have my specific hardware and not sure if there are different hardware variants of this model, so your mileage may vary.

Hopefully if you have this hardware and run Unraid it will be useful to you.

Thanks!

Any experience with this ? Any well known vendors out there ? I can't seem to find anything....

Corsair Rme 750w ATX PSU with jumper pin --> Power to HDD

if i'm, not mistaken, shouldn't Corsair PSU and Lenovo share common ground in this scenario? ( should probably also change out the corsair to a meanwell.)
And if they are supposed to be grounded, any tip on how other's have solved it?

I did buy new hardware for my current Unraid server, part of this was new motherboard (ASUSTeK COMPUTER INC. TUF GAMING B550-PLUS WIFI II)

and disks (see screenshot)

While performing parity the diskspeed was ~220Mb sec.

There is an onboard 2.5Gb NIC (realtek) onboard in the server.

When I copy data from the Unraid server to my NAS (attached to the same 2.5Gb switch) I get ~80Mb sec

When I copy from external hdd attached to the Unraid server to my NAS I get ~130Mb sec

I used large files to copy to get best results

So when I copy data from the array it is slower then when I copy form attached external hdd

I did also try with an intel based NIC but this does not solve the issue either

Is there anything that I am missing? A setting or configuration somewhere?

Thanks in advance

FUBorg

The old motherboard had PCH overheat problems so I've swapped in a new one,
I've set the boot order to the boot disk.
Set rest of bios to the same as the old one (to the best of my memory).
It either sits on black screen for ever or the HP page. No signs of the unraid boot page.

Should I reinstall unraid and restore from backup?
Any suggestions gratefully received,

https://a.co/d/03j3FK63

Wanted to post here and get some recommendations from other Unraid users as well.

Note: I know they all get hot and with spinning drives I’ll never achieve 10G speeds but I’m hoping to improve network performance with multiple in/out connections.

I recently migrated my Unraid server from a Ryzen desktop system to a Dell PowerEdge R730, and since then the system feels noticeably less responsive overall.

Symptoms

• Unraid web GUI feels sluggish/intermittently laggy

• Browser often shows:

  • “Waiting for 10.10.0.250”

• WordPress Docker container feels slow/sluggish

• VM creation in Unraid GUI feels slow

• Windows 11 VM GUI sometimes stutters/hitches

• Overall responsiveness feels inconsistent compared to previous hardware

Hardware

Server

• Dell PowerEdge R730

CPUs

• 2x Intel Xeon E5-2630L v4

• 20 cores / 40 threads total

RAM

• 128 GB RAM

Cache / VM Storage

• Samsung 990 PRO 2TB NVMe SSD

• Connected at PCIe Gen3 x4

• Used for cache pool + VM storage

Array

• 4x 2.7TB HDDs

NICs

Integrated:

• Broadcom BCM5720 Gigabit NICs

Additional:

• Intel 82599ES 10GbE NIC not in use at the moment

Current Unraid Version

Linux 6.18.29-Unraid

Current Findings

CPU Governor

The system was initially running with:

schedutil

I manually switched all CPUs to:

performance

using:

for cpu in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do echo performance > $cpu; done

This slightly improved responsiveness.

Dell BIOS Power Profile

Current BIOS profile:

Performance Per Watt (OS)

Storage Performance

VMs are running on the Samsung 990 PRO NVMe SSD, so storage should not be the bottleneck.

Localhost curl response appears fast:

curl -o /dev/null -s -w '%{time_connect} %{time_starttransfer} %{time_total}\n' http://localhost

Result:

0.000425 0.043714 0.043811

Network / NIC Details

Management interface currently uses a Broadcom BCM5720 NIC with the tg3 driver.

ethtool -i eth0:

driver: tg3
firmware-version: FFV22.00.6 bc 5720-v1.39

ethtool -c eth0:

rx-usecs: 20
tx-usecs: 72

I suspect the Broadcom tg3 NIC / interrupt moderation / latency may be contributing to the sluggishness.

System Information

CPU

Intel(R) Xeon(R) CPU E5-2630L v4 @ 1.80GHz
2 sockets
20 cores / 40 threads

Memory

125Gi total RAM

Storage Layout

Samsung 990 PRO 2TB NVMe mounted as cache pool
4x 2.7TB HDD array

Questions

• Is this known behavior with Dell R730 + Broadcom BCM5720 (tg3) NICs?

• Could tg3 interrupt moderation affect Unraid GUI responsiveness?

• Would moving management/network traffic to the Intel 82599ES NIC likely help?

• Are there recommended BIOS settings for Dell enterprise hardware running Unraid?

• Any recommended tweaks for improving UI/network responsiveness on dual Xeon systems?

Thanks!

fat-phoque-diagnostics-20260526-1229.zip

2xIntel® Xeon® CPU E5-2650 v4 @ 2.20GHz

RAM 96 GB

2x14 Tb WD Gold + 8Tb WD Gold + cache SAMSUNG SSD 512 Gb

I have installed UNRAID version 7.2.5 , and when I shoutdown the system it will sort of not complete the shutdown..... I wait something like 1hour and I had to force shoutdown and consequently at switch on parity restarted (meaning 14-16 hours waiting). What I'm doing wrong? How to proper shoutdown the system?

Moreover when doing PARITY the APP are not working, and already PLEX and KRUSADER also give error message once Parity completed. I need some help.

Thanks

There's a reset-button right underneath the errors counter-screen... 😁

A few months ago one of my pool drives gave up, causing of course a lot of read errors.

I since replaced the drive with a brand new one. But now, with updating to 7.3, my pool indicated errors. I checked and indeed, thousands of read errors.

Could it be these remain from the failure? If yes, how can I reset it? Since replacing the drive, everything worked perfectly fine. So I'd be surprised if the errors are current events...

Thanks a lot!

Why the A2000 Stands Out

Budget-to-Performance King:

$50 tier: The Quadro P400 handles 1–2 streams (4K HDR/Dolby) but struggles beyond that.

$100 tier: T400/T400 4GB offer ~50% better performance (3 streams).

$??? tier: The modded A2000 dominates with AV1 support, 8K/4K HDR transcoding, and unmatched efficiency.

Space-Saving Beast: Single-slot design fits even cramped NAS/Server cases (like my U-NAS A810). No more sacrificing drive bays for a GPU!

unRAID Compatibility Hurdles (and Fix!): The stock NVIDIA driver failed to recognize this modded card, but the open-source driver saved the day. Pro tip: Add options nvidia NVreg_OpenRmEnableUnsupportedGpus=1 to /config/modprobe.d/nvidia.conf—boom, full detection!

Specifications

Real-World Performance

Tested with:

8K AV1 → 1080p 10Mbps

4K HDR AV1 → 1080p 10Mbps Result: Buttery smooth transcoding, GPU utilization under 50% (see screenshots below).

Drawbacks (Minor!)

13W idle power (slightly high for a low-power rig).

Fixed fan speed (noisy at full tilt, but the seller includes a quieter replacement fan).

Annoying screw placement (one hidden under the heatsink—patience required).

Final Verdict

This modded A2000 is unmatched in unRAID transcoding:

✅ AV1 future-proofing

✅ Single-slot flexibility

✅ Raw power for 5+ streams

If you’re building a compact media server, this GPU deserves your attention.

PCPartPicker Part List: https://se.pcpartpicker.com/list/bPvqph

CPU: Intel Core i5-12500 3 GHz 6-Core Processor (kr2627.18 @ Amazon Sweden)

Motherboard: Gigabyte Z690 UD DDR4 ATX LGA1700 Motherboard (kr2916.00)

Case: Fractal Design Define R5 ATX Mid Tower Case (kr1224.33 @ Amazon Sweden)

Custom: be quiet! System Power 7 300W (kr217.09)

Custom: Fanxiang M.2 NVME SSD 1TB (kr1520.29)

Custom: Fanxiang M.2 NVME SSD 1TB (kr1520.29)

Custom: LSI 9211-8i (kr580.39)

Custom: G.SKILL Ripjaws V 16GB DDR4 (kr1400.83)

Total: kr12006.40

Prices include shipping, taxes, and discounts when available

Generated by PCPartPicker 2026-05-05 18:02 CEST+0200

Thanks

Joe

I bought this NIC because I’m a specs nerd and I couldn’t help myself. My mobo only has one remaining viable pci e slot left and it’s a 4.0 4x slot. Yes I could have gotten an older stable card but I’m a sucker for the latest and greatest and I liked the fact that it’s still Ethernet and claims to not exceed 5.1w under full load.

Plus it’s real intel and real server grade. Anyways…

It pops up in Unraid system devices as:

8086:57b0] 08:00.0 Ethernet controller: Intel Corporation Ethernet Controller E610 10GBASE T

But it won’t bind to a driver. I’ve spent hours with chat gpt and Google ai trying so many different options to get it to stick to either .ice or .ixgbe. Neither one works on the latest Unraid build.

I flashed the latest nvm to 1.40 when I got the card. Intel says it’ll work fine if I can get on a modern ixgbe driver but unraid is locked and I am REALLY liking Unraid overall.

Any ideas outside of running an unsupported kernel or waiting for Unraid to ship a modern ixgbe driver?

-Unraid 7.2.3

-Fractal Meshify 2xl case

-Intel 13700k

-32gb 7200 DDR5 RAM

-MSI Mag Z790 Tomahawk WiFi

-9305-16i HBA with 3D printed shroud and 60mm noctua fan

-Intel e610-XT2 NIC

-8x HDDs currently. Planning for 14

Components

Totals

Now, im fine with the wattage but the whole 12v rail thing confuses me, especially because I think that is what is causing me issues now. Can anyone recommend a PSU that has strong rails and wont randomly drop drives?

It is reccomending Corsair RM1200e 1200W PSU but I really wanted to get some non-ai opinions.

I’ve been running Unraid since last year (in my living room, since it’s not easy to run network cables there).

The whole setup is currently running on a Dell laptop with an i5 processor, 16 GB of RAM, and a 1 TB SSD (used as cache).

I have a USB 3.2 hub connected with 5x 4 TB 2.5-inch USB 3.1 HDDs (1 parity drive + 4 drives in an array).

So far, so good, and everything is running smoothly. However, I’m thinking about making the hardware a bit more “reliable.”

Key requirements would be:

- QUIET! (It needs to keep running in the living room)

- Energy-efficient!

- Expandable

- It should be able to run a few Docker containers.

- Small UPS if necessary.

- Transcoding in Jellyfin should work smoothly

- Be AFFORDABLE!

I don’t want to spend 1,000 euros or more right now.

Do you have any suggestions for what such a setup might look like?

AND how do I go about migrating to new hardware? Especially if the HDDs need to be replaced as well?

I'm having a strange issue I can't figure out and I'd appreciate any advice. 

I have bought an APC UPS, connected it to unraid and set up the daemon fine, if I unplug the UPS from the mains it shuts down cleanly at the battery level I've selected. When I then plug the power back in the UPS status light turns solid green but the server doesn't power up. If I press the power button it starts fine. 

However, I have the bios set correctly to boot when power is supplied. If I plug the server into the mains without the UPS and flick the power at the wall outlet it powers up and boots unraid. 

The controller card is an LSI 9305-24i, with 6 SFF 8643 -> SFF 8087 cables.

In the front of the chassis I can see that the drive is actually getting power, but does not appear in Unraid GUI at all. I have tested with several other drives, but have so far ignored the problem as I've had spare bays. Now the bays are all full except for the last one which is simply not working.

I would assume the only potential culprit would be the backplane? If it was the power, controller card or cable I would assume the other bays in the same row would have similar issues?

Any tips of what other things I potentially could check would be appreciated.

It a GMKtec K12 AMD Ryzen 7 H 255 with 8 core/16 thread, integrated 780M graphics should be a beast if unraid can actually use it... Got it barebones still waiting on ram arriving will be 32gb of 4800 ddr5 which is shared over the cpu and gpu in that processor setup. Has anyone run this setup before. Noticed some threads on the minis forum N5 which is the same chip and people mentioning needing custom drivers and now im panicking

My plan is initially at least run a kioxia 512gb ssd in this box then have some spinning drives in another room out of the way, they're all 6tb enterprise grade stuff probably running raid 5 got an old terramaster nas for them to go in but my plan is that nas will act as dumb storage mounted to unraid does this seem like a good idea was sure it was fine as I was buying the parts now I'm not so sure I've done right 🤣😅

I’m planning to migrate from a QNAP TS‑453A to a Lenovo P710 workstation.

I have 4 x 8 TB WD Gold drives installed, plus a 5th identical drive that I cannot mount internally in the P710, and I’m wondering whether it makes sense to use that fifth drive externally and, if so, what setup you would recommend.

I also have a 3 TB Iomega backup drive that uses FireWire 800, but it seems difficult to find a FireWire 800 PCIe card that both works reliably and is known to be compatible with unRAID. Would also you recommend to have a separate HDD externally ? I wish to have separate pool for this and install no critical data.

The QUADRO P4000 8G has an add value ? I’m going to use it mainly for Plex and Backup.

What kind of DVR would you recommend to install on this machine that would be easy to be compatible with UNRAID and Plex ?

Thanks for support

Currently I run mainly Plex, NextCloud, and a torrent client.  I'd also like to run a VM or 2.

It seems some tech influencers are excited about this but I'm not sure if it's genuine or not.
https://nascompares.com/2025/01/08/minisforum-n5-pro-nas-revealed/

I’ve been spending way too much time lately looking at the upcoming hardware shifts, specifically the jump to PCIe 5.0. It’s pretty wild to read that we’re moving from the 7GB/s speeds we saw with PCIe 4.0 into this new territory where 10GB/s+ is becoming the standard for consumer drives. I’m currently running my Unraid setup on an older 10th gen Intel system, and while it’s been a workhorse, I’m starting to map out a migration path to an AM5 or Raptor Lake platform to take advantage of these insane NVMe speeds for my cache pools.

One thing that’s been bugging me, though, is the power side of the equation. As I’ve read more about drives like the Corsair MP700 reaching those 10,000 MB/s sequential reads, I’ve noticed they seem to be getting a lot thirstier and significantly hotter. In my experience, once you start pushing the performance envelope this hard, the margin for error with your power delivery gets a lot thinner. I’ve always just used a high-quality single PSU, but with the increased wattage demands of modern CPUs and now these high-speed drives, I’m starting to wonder if my current power strategy is sufficient.

In the past, I had a drive drop out of an array due to a minor voltage ripple on a cheap unit, and it’s made me a bit paranoid about stability. When you’re moving data at 10GB/s, a power hiccup feels like it could do a lot of damage very quickly. I’m seriously considering whether it’s time to look into a chassis that supports redundant power supplies, just for that extra peace of mind, or if a high-end Titanium-rated unit is enough.

Has anyone here started experimenting with PCIe 5.0 drives in their Unraid builds yet? I'm curious if you found the power draw and heat to be a significant factor, or if the "real world" impact on your PSU overhead was less than the spec sheets suggest?

Anyway, back in 2015 or so there was a bunch of us buying up these server MB's for our UnRAID server so I was curious if anyone has upgraded since then and if so to what? Thanks.

unRAID will happily boot of just about any of these which means that the bit that usually breaks (the memory card itself)  can be replaced without requesting a new license.

The problem is that most of these adapters have one GUID per model/line meaning that you cant license them with unRAID.

There is no way to know if the GUID is unique without physically having a device to test and it is not practical for me to buy 50 of them to test.

This thread is a long term effort to find the perfect unRAID USB to memory card adapter. Ultimately it will make unRAID even more reliable.

Update:

Summary listed here http://lime-technology.com/wiki/index.php/USB_Card_Readers

Feel free to edit and fix the formatting.

https://medium.com/@aletik/the-impossible-nvidia-gpus-ca48ac425efc

how many sata drive can i connect? will a ASUS hyper m.2 drive work? how will a m.2 to oculink adapter benefit me to add more drives using a x4 Pcie lane?

Your ASUS ROG Crosshair VIII Dark Hero is actually a very good base for a DIY NAS because it has a lot of SATA ports and flexible PCIe lanes. Let’s break down what you can realistically connect and how expansion options like Hyper M.2 or OCuLink would work.

1. Native drives you can connect (without adapters)

Your motherboard provides:

• 8 × SATA 6 Gb/s ports

• 2 × M.2 slots (PCIe 4.0 x4 NVMe or SATA) (@ROG)

So directly on the board you can run:

• 8 SATA drives

• 2 NVMe drives

➡ Total = 10 drives without any add-in cards

Importantly:
Unlike many boards, using both M.2 slots does NOT disable SATA ports on this model. (Tom's Hardware)

2. PCIe slot layout (important for NAS expansion)

Main slots on this board:

• PCIe 4.0 x16 (CPU) – GPU slot

• PCIe 4.0 x16 (CPU) – runs x8 if both used

• PCIe 4.0 x16 (chipset) – x4 electrical

• PCIe 4.0 x1 slot (@ROG)

If you already have:

• 2 GPUs

• USB expansion card

• Ethernet card

You’re already consuming several slots, which limits expansion options.

3. Will an ASUS Hyper M.2 card work?

The ASUS Hyper M.2 x16 Card V2 can hold 4 NVMe drives, but it only works properly if the motherboard supports PCIe bifurcation (splitting x16 → x4/x4/x4/x4).

Your board does support bifurcation, but there are caveats:

Best case

If installed in a x16 slot with bifurcation enabled

You can run:

• 4 NVMe drives on one card

If installed in a x8 or x4 slot

Then you get:

• x8 slot → 2 NVMe drives

• x4 slot → 1 NVMe drive

So with two GPUs already occupying the main slots, the Hyper card will likely only run 1–2 drives.

4. Using an M.2 → OCuLink adapter (very useful for NAS)

This is actually a great NAS trick.

What it does

An M.2 NVMe slot (x4) is converted into OCuLink (SFF-8611).

Then you can connect:

• OCuLink → 4-port SATA

• OCuLink → 8-port SATA

• OCuLink → U.2 / NVMe backplane

Bandwidth

An M.2 PCIe 4.0 x4 slot ≈ 8 GB/s

Typical SATA drives:

• ~200 MB/s each

So:

• 4 SATA drives ≈ 800 MB/s

• 8 SATA drives ≈ 1.6 GB/s

➡ Even an x4 link easily handles 8 drives

5. Example NAS configuration with your board

Possible layout:

Motherboard

• 8 SATA drives

M.2 slots

• 2 NVMe cache drives
  or

• 1 NVMe + 1 OCuLink adapter

PCIe x4 slot

• 8-port HBA (LSI card)

Realistic drive counts

Without crazy mods:

Total possible: 20–30 drives

6. Best NAS expansion method (recommended)

Instead of Hyper M.2, I strongly recommend:

PCIe SAS HBA

Examples:

• LSI 9211‑8i

• LSI 9300‑8i

Advantages:

• 8–16 SATA drives

• True enterprise controller

• Perfect for TrueNAS / Unraid / ZFS

7. What I would do with your exact system

Given:

• 2 GPUs

• USB card

• Ethernet card

I’d do:

Onboard

• 8 SATA drives

M.2 slot #1

• NVMe cache

M.2 slot #2

• OCuLink → 4-8 SATA

PCIe x4 slot

• LSI HBA (8 drives)

➡ Total ~20 drives possible

💡 One important NAS tip:
Gaming boards like this lack ECC support with some CPUs and IPMI, but performance is excellent.

✅ If you want, tell me:

• Your CPU model

• Case size (how many drives physically fit)

• Which NAS OS (TrueNAS, Unraid, Proxmox)

…and I can design an optimal 20-30 drive layout for this motherboard. It’s actually a surprisingly powerful NAS platform.

if I do nr 7 i lose the usb card that i pass through to a VM and the same with the ethernet card

You’re running into the classic limitation of consumer boards used as NAS/virtualization hosts: not enough PCIe slots once GPUs are involved.

Let’s map the ASUS ROG Crosshair VIII Dark Hero slots first so the options are clearer.

1. Your motherboard PCIe slot reality

Slots available:

1. PCIe x16_1 (CPU) – normally GPU

2. PCIe x16_2 (CPU) – GPU (runs x8/x8 with slot 1)

3. PCIe x16_3 (chipset) – x4 electrical

4. PCIe x1

So if you're using:

• GPU #1

• GPU #2

• USB controller (for VM passthrough)

• Ethernet card

You’ve already filled all usable slots.

That means:

• No slot left for an HBA

• No slot left for Hyper M.2

So expansion must come from M.2 slots or lane sharing.

2. Best option for your situation: M.2 → SATA/OCuLink

Use the two M.2 slots as your expansion ports.

Option A – M.2 → 6-port SATA controller

Example chips:

• ASM1166

• JMB585

Each M.2 adapter can give:

• 5–6 SATA ports

So:

➡ Total = ~20 drives

No PCIe slots used.

3. M.2 → OCuLink (better than SATA controllers)

Instead of SATA controllers, you can convert M.2 to OCuLink.

Then connect to:

• OCuLink → U.2 NVMe

• OCuLink → SATA backplane

• OCuLink → SAS HBA

Advantages:

• lower latency

• higher bandwidth

• enterprise cabling

An M.2 x4 link can comfortably run:

• 4–8 SATA drives

So:

➡ 16–24 drives total

4. Why Hyper M.2 won’t help you much

The ASUS Hyper M.2 x16 Card V2 requires:

• a x16 slot

• PCIe bifurcation

But your two x16 slots are already used by GPUs.

So it would force you to remove either:

• GPU

• Ethernet

• USB controller

Not ideal for your setup.

5. Another trick: USB controller via motherboard headers

If your USB passthrough card is just for extra USB ports, you might not need the card.

The ASUS ROG Crosshair VIII Dark Hero already has:

• multiple USB controllers

• USB headers

Often you can pass through the onboard USB controller to a VM instead of using a card.

This frees a PCIe slot.

6. Best layout without losing your cards

PCIe slots

• GPU 1

• GPU 2

• USB passthrough card

• Ethernet card

Storage

• 8 × motherboard SATA

• M.2 → SATA controller (6 drives)

• M.2 → SATA controller (6 drives)

➡ 20 drives

7. Ultimate trick (if you want crazy expansion)

You can use a PCIe bifurcation riser:

• x16 → 4 × x4 slots

Then one slot can hold:

• USB card

• Ethernet

• SAS HBA

• etc.

But it requires BIOS bifurcation and physical space.

💡 One more important question that changes everything:

What CPU are you running?

Example:

• AMD Ryzen 9 5950X

• AMD Ryzen 7 5800X

Because Ryzen CPUs only provide 24 PCIe lanes, and how those lanes are used determines the best expansion strategy.

If you want, I can also show you a very clever NAS trick using one M.2 slot that gives 20+ drives while keeping all your PCIe cards. Most homelab builders with this board don’t realize it’s possible.

i'm using a 3900x cpu. show me the nas trick with one m.2

With the AMD Ryzen 9 3900X and the ASUS ROG Crosshair VIII Dark Hero, you actually have a very powerful trick available: turn one M.2 slot into a full SAS storage controller without using a PCIe slot.

This works because an M.2 NVMe slot is just PCIe x4 lanes. If you convert that slot to a PCIe slot, you can run an HBA exactly like a normal server.

The “1× M.2 → 16+ drives” NAS trick

Step 1 — Use an M.2 → PCIe x4 adapter

This adapter plugs into the motherboard M.2 slot and gives you a PCIe x4 slot on a cable or riser.

What it does:

M.2 slot (PCIe 4.0 x4)
        ↓
PCIe x4 slot

Now you can install a real SAS HBA.

Step 2 — Install a SAS HBA

Typical cards:

• LSI 9211-8i SAS HBA

• LSI 9300-8i SAS HBA

These provide:

• 8 SATA/SAS drives directly

But the real magic comes next.

Step 3 — Add a SAS expander

A SAS expander multiplies ports.

Example:

• Intel RES2SV240 SAS Expander

Connection:

M.2 slot
  ↓
M.2 → PCIe adapter
  ↓
LSI 8-port HBA
  ↓
SAS Expander
  ↓
16–24 SATA drives

Now a single M.2 slot can run 24 drives.

Why this works well

Bandwidth calculation:

M.2 slot (PCIe 4.0 x4):

• ~8 GB/s

Typical HDD:

• 150–250 MB/s

Even 20 drives only need ~4 GB/s, so the link is still not saturated.

So performance is perfect for NAS workloads.

Your full system layout

With your current hardware:

PCIe slots

• GPU #1

• GPU #2

• USB passthrough card

• Ethernet card

(no changes required)

Storage

Motherboard:

• 8 SATA drives

M.2 slot #1:

• NVMe cache drive

M.2 slot #2:

• M.2 → PCIe adapter → HBA → SAS expander

Result:

➡ 24–32 total drives possible

Why homelab NAS builders love this trick

It solves the biggest problem with consumer boards:

• no PCIe slots left

Yet still gives enterprise-style storage expansion.

It’s commonly used in TrueNAS, Unraid, and Proxmox builds.

One more trick (specific to your board)

The ASUS ROG Crosshair VIII Dark Hero has two M.2 slots.

If you really wanted extreme storage:

• M.2 #1 → HBA (24 drives)

• M.2 #2 → NVMe cache

or

• M.2 #1 → HBA

• M.2 #2 → second HBA

That could theoretically run 40–48 drives from a gaming motherboard.

✅ If you'd like, I can also show you:

• The exact $15 adapter that works best for M.2 → PCIe

• How to power the SAS expander without a server motherboard

• How some people run 30+ drives on this exact board

Those tricks make a huge difference for NAS builds.

→ HBA → SAS expander. how does a sas expander connect with a HBA

A SAS expander works a bit like a network switch for storage. Instead of each drive connecting directly to the HBA, the expander lets many drives share a few SAS links from the HBA.

Let’s walk through how it connects.

1. Basic connection layout

Example hardware:

• LSI 9211-8i SAS HBA

• Intel RES2SV240 SAS Expander

Connection looks like this:

CPU / Motherboard
      │
M.2 → PCIe adapter
      │
LSI HBA
      │ (SFF-8087 cable)
      │
SAS Expander
   ├── Drive 1
   ├── Drive 2
   ├── Drive 3
   ├── Drive ...
   └── Drive 24

2. The actual cables used

Most SAS HBAs use Mini-SAS connectors.

Typical ports on the LSI 9211-8i SAS HBA:

• 2 × SFF-8087 internal SAS ports

Each port carries 4 SAS lanes.

So one cable provides 4 drive channels.

HBA → Expander cable

You connect them using:

SFF-8087 → SFF-8087 cable

HBA port
   │
SFF-8087 cable
   │
Expander port

This creates an uplink from the HBA to the expander.

3. Connecting drives to the expander

The expander has many ports.

Example Intel RES2SV240 SAS Expander:

• 24 SAS lanes total

• multiple SFF-8087 ports

Each port can break out to 4 SATA drives using a breakout cable:

SFF-8087 → 4× SATA

Expander port
     │
Breakout cable
 ├── SATA drive
 ├── SATA drive
 ├── SATA drive
 └── SATA drive

4. Bandwidth explanation

One SAS lane (SAS2):

• 6 Gbps

One SFF-8087 port:

• 4 lanes

• 24 Gbps total

If you connect two HBA ports to the expander:

HBA port 1 ─┐
            ├── Expander
HBA port 2 ─┘

You now have:

• 48 Gbps uplink bandwidth

That easily supports 20+ HDDs.

5. Powering the SAS expander

The Intel RES2SV240 SAS Expander can be powered two ways:

Option A (server style)

Plug it into a PCIe slot (for power only).

Option B (better for your case)

Use a Molex power connector.

So it can sit anywhere in your case.

6. Why this setup is popular for NAS

Advantages:

• One controller manages dozens of drives

• Works perfectly with TrueNAS / ZFS

• Very reliable enterprise hardware

• Uses very few PCIe lanes

7. Example setup for your system

Your system could look like:

M.2 slot
   ↓
M.2 → PCIe adapter
   ↓
LSI 9211-8i
   ↓
Intel SAS expander
   ↓
24 HDDs

Plus:

• 8 motherboard SATA drives

➡ 32 total drives possible

💡 Important tip:
Flash the LSI 9211-8i SAS HBA to IT mode so it acts as a pure HBA (no RAID). That’s required for TrueNAS, Unraid, and ZFS.

If you want, I can also show you something even more interesting:

A cheap 36-drive SAS expander setup that uses only one cable from the HBA.

Most homelab NAS builders miss that trick, and it works perfectly with your 3900X system.

(For 6 bays I found the Terramaster F6-424 Max, but it is only 6 bays... Plus: Is Terramaser any good?)

Motherboard ROG MAXIMUS X HERO (WI-FI AC) - with i7 8700k and 16GB DDR4

Photo of the SMART test result (You may run a Seatools test on the drive or use the free trial of HD Sentinel to check its health.)

Is it possible to run either of these things in Unraid?

When i try and run either the of the self tests within unraid, i get an unknown error (Diagnostics attached). I am unaware it's not a recommended controller, i do intend to change as part of a hardware upgrade shortly.

sap1ent_diagnostics-20260311-2321.zip Sap1ent-smart-20260311-2319.zip

I looked into extensions like CableMod but fast found out they dont make power extensions.

All I could find was like 1 to 1 sata extension cables, that sounds a bit messy.

I’m planning a quiet Unraid NAS for my home office and would appreciate feedback before purchasing the parts. The system will initially start with 4-6 drives, but the goal is to eventually grow the array to 18 drives, so expandability and long-term reliability are important.

The NAS will sit in a very quiet office, so noise matters a lot. That’s why I chose mostly Noctua cooling components and plan to spin down all drives when not in use.

The main workload will be Jellyfin (mostly Direct Play) and file storage.

Planned Hardware

Notes on Some Choices

PSU

I am aware that the TX-1600 Noctua PSU is massive overkill for the build. I chose it intentionally for the following reasons:

• excellent build quality

• extremely quiet fan profile. I have used it in another build and it is effectively inaudible

• I dislike fanless PSUs because long-term heat can degrade components

• A total of 6 SATA power connectors

The Seasonic cables have 3 SATA connectors each, so with 6 SATA power connectors I can power 18 drives safely without questionable power splitters. I want to use only the original Seasonic cables.

Design Goals for the Motherboard

I am still not fully happy with my motherboard selection.

Ideally the board should have:

• Socket 1700 for i3-14100T

• DDR4

• 2.5 or 5 Gbit LAN

• 1× PCIe x16 slot (for HBA)

• 1 additional PCIe slot for future expansion with enough space between it an the 16x port so that I can place 2x 80mm Fans in a PCI-slot bracket in between them to cool the HBA.

• internal USB 2.0 header for the Unraid USB

• HDMI and/or DisplayPort

• No on-board GPU (there is already an iGPU in the i3-14100T)

• ASPM settings in BIOS

• good Linux compatibility

• minimal extra controllers / RGB

Basically: as few gimmicks as possible.

Questions

1. Realtek LAN concerns

Both candidate boards use Realtek network controllers.

I have read that Realtek NICs can sometimes disable ASPM which prevents deep CPU package C-states thus increasing idle power. Has anyone experienced this on these boards with Unraid? Would an Intel NIC board be significantly safer for low idle power?

2. HDDs

• Are there quieter enterprise drives above ~24 TB that you would recommend? Optimally with published Benchmarks that you can link to?

• Due to write speed I want to avoid SMR and instead go for CMR or HAMR. Thus Seagate Exos 30 TB would be the largest Exos I consider, due to 32 TB and larger being SMR.

• Are Exos drives acceptable in a large 18 drive array in terms of vibration? For some WD Red drives I found entries how many drives can be used in the same case, due to vibrations of the other drives being an issue. Is there any information out there regarding this for the 28 TB Exos? I did not find it.

• Can enterprise drives be shut down or should they run 24/7? I wanted to wake the NAS by LAN when starting my main PC and put it to sleep either when shutting the main PC down or via a command from a tablet within the local network (the one used to watch Jellyfin content when not sitting at the PC).

• I am also open to other drives. The reason I am so focused on the recertified 28 TB Exos is their "low" price. The current HDD market is nuts.

3. HBA compatibility

I plan to connect all drives through the LSI 9306-24i to avoid mixed controller issues.

• Is the 9306-24i known to cause issues with Unraid? As far I see it the HBA is already flashed in IT mode from the manufacturer, thus this should not be an issue.

• Does it prevent deep C-states or low idle power with certain mainboards? Can you recommend some mainboards and/or HBAs that fit my build and don't have this issue?

4. USB header → single USB port

I could only find USB header splitters (1 → 2 ports). Does anyone know a high-quality adapter that converts a USB 2.0 header to a single USB port? Looks cleaner in the end.

5. Fan hub

Since the system will use multiple case fans and mainboards with few controllers tend to have few fan connectors, I still need a good fan hub. Recommendations for a reliable PWM fan hub would be appreciated. How to pull HBA heat information and inform the fans with it would be even more appreciated.

Any feedback is welcome. I’d like to get the platform right before committing to the hardware.

Thanks!

I have a storage case from inter-tech, where there are 24 HDD bays. I have WD Red disks. I will power them with a Seasonic Focus GX-750W power supply. There are backplanes in the case. 1 backplane = 4 HDDs with 1 molex for power. A total of 6 backplanes and 6 molex connectors. Since I didn't have enough molex cables, I ordered cables from cablemod. 1 cable with 3 molexes. I have 3 of these cables. Can I power 8 HDDs with one cable with 2 molexes?

Can I power all 24 disks with 3 cables? Each cable will use 2 molex connectors.

The disks consume 2A (24*2=48A) when spinning. The Seasonic Focus GX-750 has 62A for 12V. So will the power supply handle it?

Runs a few dockers.

• sabnzbd, some *arrs

• Plex & tautulli

• fr24feed-piaware

• paperless-ngx

• immich (don't really use it)

• nextcloud (don't really use it)

• rustdeskserver-AIO

• luckybackup

and 3 VMs

• Home Assistant OS

• Windows XP (old backup from my father in law, he access it every now and then via VNC) - no network in the VM

• Windows 10 VM with ETS programming (most time offline)

Storrage

• Array

  • 3x Toshiba MG10ACA20TE with 20 TB each (25,4 TB used, 14,3 TB free)

• Pool

  • 2x Kingston 2TB nvme (cache mirror) (holding app data and VMs)

  • 1x Gigabyte 512 GB nvme (download temp for sabnzbd)

Most of the storage is used for "Media" (plex), TimeMachine (Mac Backups) some for personal Backup like Pictures and Files (they are also in OneDrive).

I have still got an DELL i7 OptiPlex 7060 microSFF with an i5-8500T, 16 GB Ram on the shelf. I could put in the 512 GB nvme from my current build and a 1 TB SSD 2,5" in the SATA slot for an unraid installation (with no Array) just to hold the Dockers and the 3 VM (whereby 2 of the 3 are more in stand by mode anyways) and would migrate the 3x 20 TB Toshibas and the 2x 2 TB Kingston nvmes to a Unifi UNAS 4 Bay, with all my Network gear I am already on the Unifi Universe.

It would out of the box do my storage, my pictures sync from all my mobile Phones, my Time Machine. Running the 3x 20 TB Toshibas in Raid 5 would give me the same space and protections as I do have right now and I could get an other 20 TB in the future to extend it to "about 60 TB useable storage". The UNAS would hold the 2x 2TB Kingston as a read/write cache and network would be connected via 2.5 GBit/s to my Switch, while the "good old" unraid Optiplex would be on the network via 1 Gbit/s.

Dockers on the downscaled unraid would only be

• sabnzbd, some *arrs

• Plex & tautulli

• fr24feed-piaware

• paperless-ngx

• rustdeskserver-AIO

• luckybackup

Whereby only sabnzbd/*arrs would write to the NAS (SMB share mounted) on completion and Plex would read from the NAS (SMB share mounted).

Luckybackup would write the Paperless files and backups of the VM/Dockers to the NAS periodically.

The CPU load of the current server (when Plex is used) is less than 20% - videos are mostly direct streams but I also have a Plex Pass which enables hardware transcoding on the iGPU - which also the DELL OptiPlex has.

Would that kind of make sense? Would it be lower on power consumption?

Here is what I think I know:

Document all the drives serial numbers and their corresponding IDs in Unraid

Take array offline

Power down

Move USB to new hardware - does this work? does it account for the license?

Move drives over

Power up

Check array

Assign drives the proper IDs

This seems to simplistic so any help is appreciated.

I would like to build my own server (motherboard/CPU/RAM) vs, getting one with used components. As a result it would be much easier in a 3U or 4U case as those will support commodity hardware/coolers better. I really want a full depth case with rails, not a shallow screw mount rack case. I also would like to have more than my current 12 hot sway bays. 16+ would be ideal.

Any recommendations for cases? Preferably used. The Supermicro 836 comes up a lot on forums, but I want to consider all options.

SIDE NOTE:

This would be my first go a 'rolling my own' in a server chassis. Any concerns with consumer CPU coolers in a 3U case?

I've 5 20Tb Seagate SAS X20 drives - ST20000NM002D. 4 of these are successfully spinning down using the SAS Spindown plugin from doron. However one of these will spindown, but immediately spins back up again. I have removed this from the array and it's now an unassigned device that has been wiped and formatted. It still refuses to remain spun down. Are there any commands or firmware upgrades that I can do to make this behave correctly?

Thanks in advance.