Greetings fellow travellers!

I'm new to unraid and ran in to a problem of me possibly having a bad config?

My Setup is a Ugreen DXP4800 plus with 2 Seagate Exos 22TBs and 2 Kioxia 1 TB NVMEs.

Network is 2.5gbs with a unify flex 2.5g and new cat6 cable. Standard Setup HDDs as array, NVME as pool devices, with all settings on defaults as set by unraid, no special config tweaks of any kind.

Now to my Problem, I'm getting really "underwhelming" write speeds, when copying to the NAS. Copying to the spinning rust array i get about 130 mb/s on disks that can do well above 200. Essentially

just a little over 1gbps network speeds.

I have btw no parity since it is just media which i have backed up on separate disks anyways, and I'm not gonna shell out around €500 for refurbished disks for essentially just uptime, for some movies that i can copy, or in worst case rip from the original media again.

So parity checks are not the limiting factor here!

And i checked this with rather big files above 30gb, not just little shit with tons of folders and files of

small size, then it gets even worse naturally.

Then I tested it with one of the NVMEs, put a share on it, copied a big file, got a peak of 156 mbs, for a few seconds and then it levels out at around 130 mbs just as the hdd array.

This is with the default network settings, tried different things like enabling jumbo frames, different MTU, turning on and an off pretty much everything in the settings to no avail, always the same result. And yes i know how to reset to default so there is no piling up of config errors.^^

I tested the HDDs & NVME in the NAS with the Diskspeed app and they are working fine, giving read/write speeds as advertised.

I tested the network with iperf and get 2.35gbs in both directions consistently.

Ah, and before I forget, the source of the copys on my Desktop is a NVME as well, which performs well in speedtests.

I expect write speeds well above 200mbs consistently, especially on the NVME, which i was ofc planning to use as cache.

What the heck am I doing wrong? Or is this what I can expect perfomancewise from unraid?

Any insight or tips would be highly appreciated!

Try copying to a disk share (or exclusive share)

need more info, how is your storage assigned in unraid? Array + Parity? Zpool? btrfs pool? combination?

Hi Jorge & MowMdown!

Thank you for the fast response.

OK, so I activated Global Shares and exported the NVME, that I tested before as pool device as a disk share and wrote to it directly:

I got around 190-195 mbs the whole transfer not just as starting peak with diminishing speed to 130 over the transfer as before, so it is an improvement of around 60mbs.

But really this is a lot slower than i would expext NVME to NVME over 2.5gb networking. The transfer should be saturating the connection getting something like 280+, shouldn't it?

My setup

As already stated it's a Ugreen DXP4800plus, Pentium Gold 8505, 16 Gig RAM, Ethernet AQtion AQC113 NBase-T/IEEE 802.3an

The two Seagate Exos 22TBs are in an array with no parity, formated in XFS.

The two NVMEs are assigned as single Pool devices formated in BTRFS. So two pools with one NVME each.

The array is not cached yet, was planning to use the first NVME for that after filling the array from my old Synology NAS.

The second NVME holds docker containers and is planned to contain my music folder, because I want to have the rust spin only when I'm actually watching a movie not just using an app or listening to music. This is a actually working fine, since browsing my media on moonfin on my TV doesn't make the HDDs start only when I start playback.

Appwise I got the following for now:

ipferf3, unassigned devices (including UD plus), krusader, diskspeed and jellyfin

Everything is working fine so far, just the transfer speeds leave more than a little to be desired.

That suggests to me a CPU performance issue, I would not expect to see so much difference with those numbers. Make sure the CPU is throttling up; also check Settings - Power Mode is set to Best Performance.

Hi Jorge,

Tips and Tweaks Power Saving Settings

Enable Turbo yes

Schedule Turbo no

Normal governor performance

Power saving governor power save

would love to set the latter to on demand or or at least balanced as listed in the tool tip, but these options are not available for me? The drop down menu is not showing them, only the tool tip.

Had a look at the BIOS, there is nothing to set there other then Turbo on or off, which is obviously set to on.^^

When looking at the actual frequency the cores are running at, it gets weird.

First I tried cpufreq-cli which showed really low numbers, like 2 digit something between 10 and 100 mhz for all cores idle, ramping up a little during file transfer never reaching 1000mghz on even one core, falling short in more like 700-800 mhz on the p-core. This seemed a little to off to me. So i just did some grep "cpu Mhz" /proc/cpuinfo while transfering a file and it looked quite different, more in line with what you would expect. Baseline idle around 400mhz with 2 or 3 cores ramping up to around 2-3000 mhz intermittently. While transfering a file at around 190mbs (still to disk share) it looked something like this:

cpu MHz : 4397.066

cpu MHz : 4400.211

cpu MHz : 400.000

cpu MHz : 3300.291

cpu MHz : 3298.574

cpu MHz : 400.000

I have no idea where the bottleneck is or how i could further kick the cpu's derriere to ramp up some more. I'm sure there could be something else to be done in some config file and I'm not afraid of command line or editing configs, but i tend to handle these things with the utmost care since we all know doing that is a really bad idea, when you don't really know what you're doing and have no guidance.

I'm at a loss here...

Still seems to me like some kind of bottleneck; definitely it's not an Unraid limitation, with a fast enough CPU and devices I basically get line speed with 25GbE (with exlusive shares, FUSE always add some overhead, but this is typically noticeable with 10GbE or higher only)