_cjd_

I'd run what you have and learn what needs improvement first - though you might miss transcoding potential from the 12400 you may still be fine. Experience before spending when you can always wins. That may also give you a little time to save if you find you want any parts slightly more $ - server builds can last many many years. Make sure you get pci lanes to give you room for everything you might need (e.g. CPU transcoding falls short and you need a standalone card). I feel like 16gb ram may be selling yourself short - though that depends so much where this journey takes you next. Cache size depends on your usage. Bigger drives usually have better write durability. I run 1tb for docker/VMs with good use ratio. 1tb is nearly always adequate for a write cache for me, but not quite (depends how multiple systems overlap or don't) - with other use cases since added that's now 4x2tb raid 1+0.

If you go Intel you may not need an external video card as the igpu is quite good in this scenario (transcoding). For that matter, even with AMD you may be fine (I've had no issues with limited jellyfin use to browser or Roku ultra, AMD 5600X) Do you need that much memory or CPU? What kind of usage do you see now? Lots of VMs? Suspect you're over-speccing but maybe not. Power supply may be fine - you may be able to find one more power efficient. They do fail, though it's hard to tell when they do till it's catastrophic.

Can you use more cores? 5900 or 5950. In 8 core maybe 5700? Understand the demands of the software, it'll tell you what hardware will work best.

How critical is data integrity? For media streaming and most common use cases it's probably overkill. And "probably" is hedging. A flipped bit is almost surely impossible to notice. If you're not running redundancy on the array cache I have to assume you've already decided even losing a full file isn't an issue, so no ECC is fine.

Do you need the video card? Otherwise I'd start by reading through the powertop thread. There are certainly ways to spend on new hardware to avoid spend on electricity too. Also, start with a good way to read power consumption so you can tell accurately what each change does.

Your hands aren't tied at all - you just have to plan the move and back up (and restore) the data. Very possibly the same thing you'll need to do if a pool drive ever fails.

Why would 192.168.30.1 work? It works of the card is installed but network cable disconnected, but not when you connect cable to the new card? Based on what little info is shared I can't help but suspect this is on the network setup side, not unraid.

You can't use 11.0.0.0/8 at all (or shouldn't, it's public/routable) Not sure that's your problem but thought I'd note. Do things work when only the onboard LAN is connected? I assume you're still trying to connect to the 192... address. Double check that it's using the same address (check your router).

I got the RSV-L4500U The stock 2nd row fan mount has sharp edges and adds gobs of noise... I didn't care since I replaced it to fit the larger fans. The case itself is decent thickness steel, no complaints. I've seen consumer pc cases from them and they're not as nice... but also much cheaper. On par with all the other rack mount cases I've used (which isn't many, to be fair.) I have mine mounted "tower" orientation to a wall with a support at the back. I did also test and verify the 2 IcyDock 3x2 I have lined up vertically (5.25" top to bottom) making me think you might be able to use other hot swap cages and do more of a mix and match arrangement. I can't recall if I could have easily used the 6x2.5" hot swap I have. I can't vouch for the Rosewill hot swap bays. My IcyDock cages are an older model with a full enclosure at the 5.25" width. The 2.5" is a current model. I would have 47°C temps on the SSDs with this when doing backups (10gbe so it saturated drive write speed) thus the decision to not use it.

I really like the Rosewill 4u I have. Solid case, lots of room. I opted for no hot swap bays to let me run 120mm fans pushing across the drives; I made a custom mount to run 3x140mm in the second fan row. I had hot swap before and never used it - I realized I will likely always power down to do maintenance. There are 3 cages easily removed to actually work on drives, whichever case variant. I have 10 drives - 5x16tb 3.5", 5x SSD (2.5") and room for 5 more. Drives max out around 35°C, nvme gets a bit hotter. My goal was to manage temps hoping to prolong disk life. Lots of variations available on this case including 12 hot swap bays.

I have Mellanox Connectx3 and Intel x710-da2. Both worked out of the box, whether copper or fiber DAC to a Unifi USW-Aggregation. I also have my backup Unraid and a windows system connected via cat6 and a Unifi 10gbe SFP+ adapter - though hoping to run fiber when we have to open up walls for a 50a to the kitchen as those adapters are far more power hungry... I had to tune the windows/rj45 (forget which Intel card I have there, x550 I think) but with jumbo frames set up properly I see transfer rates approach theoretical max in iperf3. Backing up Unraid to unraid is disk speed constrained - it doesn't need the 10gbe for the main array. Not using jumbo frames drops to ~6Gb I got the Intel x710 as it supports lower power states better.

rsync is the way to go but can be weird with linked files and such without the right flags. It does not require smb. It's super useful to be familiar though. Krusader is more familiar for folks used to a GUI. I haven't used it for this kind of transfer. Is the 18tb a choice for any particular reason? You only need 14tb for parity. Not at all familiar with what you describe in 2. Appdata cache plan makes sense, not sure about the other. You may want a write "cache" after the files are moved over if you need more than write speed of a spinning disc (or am I just misunderstanding your description here). Lots of ways to set things up though so do what suits your needs. As long as you confirm 10x over your data is transferred before destroying the filesystem and moving them from the nas to unraid, you can try and redo till you have it right.

I use a short 2u case and an old i3 I had sitting around... It's too slow to run some things (it was just burning CPU trying to cache directories so to avoid always spinning drives up... Useless on the backup system I just wasn't paying attention when I set it up) but zero issues for backup. No parity though. Case hangs between studs under the stairs in the basement. Conveniently out of the way but still accessible. I do have 10gb networking and can confirm it's overkill, though I can approach saturating writing to m.2 nvme drive. If I ever have to run some services from the backup (e.g. hardware failure on main) it'll be handy though. CPU may be the issue then.

From experience... Jump to 10gb networking. Assuming you're not using spinning disks for photo access etc... if you can do SFP+ DAC (not rj45) its almost cheap to get into. And far more stable. Video card will do it's thing wherever it is. Diskspeed docker may help on perf measurement numbers.

Is there room to fit anything in pciex_2 or 3 with the 1060 in the lower position? You're fundamentally up against the hardware not quite having what you need. Is 2.5Gb that important?

Mirrored is fine for some level of redundancy but leaves no road for expansion. Does that matter? I think you have to do some trickery still to not have an array (next version changes that). Either way Unraid doesn't care where the drives are plugged in. I swapped the same drives between a few different systems when I was getting started.

I believe it's the SpaceInvaderOne vid/script which includes actually swapping stuff around so the backup server takes over primary duty - more bouncing between two than pure backup really. That might be useful if you plan to walk away with the primary.

You won't have memory error correction. Skimming threads leads me to believe most unraid users don't bother with ECC but I have no way to know if that's accurate. Unraid requires far less memory in the first place for a given data capacity. It's simply a risk to weigh. You haven't used it so far for your Plex server right? It's not any different continuing. I have it in my primary server but not my backup server- though some day I'll swap hardware in the primary and upgrade the backup and it'll have ECC then. And that's the magic - it's not hard to swap in the future if you decide to.

I also have a backup unraid server - WoL and sync via User scripts weekly, waits if it tries to start while parity is running - still using fewer drives than the main server but it works great. Hardware all salvaged except the drives. It's in a short 2u case and is technically portable. Handles and all that...

That's unfortunate! If you have any non-ECC to test it might be worth it just to be sure. That's well into unusual interactions but glad you found more info. Makes me wonder if it's just the 5950 or also the 5900 - and/or just on the x470 (I have an x570d4u with 32gb ECC at xmp/3200 and a 5600x - stable as can be. I also don't need more cores)

2 systems updated 6.12.10 > 6.12.11 no issues. Thanks.

As far as I know, yes. You can always format and test before committing to actual use in this way. If it's USB, I've read in a few places that can cause issues but perhaps only if an array device; you'd be mounting an unassigned device I think.

You can't move a multi-drive setup with a software raid solution between systems. It would have to be hardware raid, and the controller dependencies that brings along for the ride. Individual drives are not independently readable in either case. Unraid array behaviors are limited to an unraid array. The drives are still readable independently, just not directly in windows. But... A livecd/USB boot of a Linux distro (which includes unraid) in a pinch should let you mount and share. So you'll have to weigh your requirements.

Unless it's set to xmp speeds I'd be surprised to see ram an issue. If it is xmp, it's technically overclocked, certainly not default settings. Assuming a good CPU, heat, insufficient power, or a timing/clock related setting mismatch are the most likely issues that come to mind. When you swapped back to the 3900 was the thermal paste coverage good? What cooler are you using? The new CPU runs hotter and demands more power. Heck, just check all the power supply connections to be sure they're seated, maybe reseat just to be sure. And make sure that supplemental power is a full 8 pins. Speaking of reseating, try that with ram as well just to be sure. If you have more than one stick, you could try dropping to one at a time... I'd be really surprised to see this be unraid related -unless a VM starts immediately which causes issues over drivers or something. I bounced between incredibly different hardware setups early on getting everything sorted and no issues.

You on the latest bios? Have adequate CPU cooling? All the supplemental CPU power on the motherboard attached? Have you reset bios to defaults? It sounds a bit like overheating or power instability but that's wild guessing. It could also easily be overclocking settings not quite sitting right with the new CPU if you did any of that.