_cjd_

Drives that big I wanted two parity - one lost is a lot of data (I'm only at 16tb drives at the moment) That said - they're not quiet, but walls or any other physical obstructions will help reduce noise levels a ton. Isolation mounted drives will help a little. My home office is in the room next to my server and I can hear drives spin up. Nowhere loud enough to wake me though. In fact, an upright freezer is notably more obnoxious - how is the fridge noise for you? It is the kind of noise I suspect you'll get used to ignoring. That said, sensitivity to noise while sleeping is so different for each person, it's hard to know.

https://pcpartpicker.com/products/motherboard/#A=1 Any good parts site also lets you filter on capabilities. I have no idea why I'm even going this far for you - the fact you're coming here asking for others to hand you answers makes me doubt any build will go well. Bifurcation you have to check manuals, but server/workstation boards are likely to have support - desktop, check. I still see 133 options on Newegg among server boards with ECC.

Why exactly are you looking at zfs at all here? For media sharing a vanilla unraid array seems a better answer.

Plenty of options either Intel or AMD. Can't say I've seen 4 m.2, but an add-in card works if the pcie slot supports bifurcation (best approach, anyway). That's still loads of options.

The power supply will limit quite how efficient this can go. if you really want to sip power a different direction and no ECC is probably in order. If you haven't, read through the powertop thread. I'm not sure how the CPU will do with all the sharing and future maybes - I have an i3 (granted, a nit older) in my backup server and it can't handle a "cached directories" plugin (also doesn't need it). I'm not experienced with this specific hardware or all the things you may run enough to be sure. And finally - parity is not a guarantee, just a probability. Two parity drives improves the odds. Proper backup to multiple systems/locations is as close to guarantee as you can get.

At a guess it may have been used for something like pfsense. If your network is segmented with vlans you can run those on different hardware, pair them for redundancy, or dedicated hardware for a bunch of VMs... Are the 1gbe non-removable? If so they may have just been ignored (I do that, dual 10g SFP+ and unused onboard 1gbe). Whether it's worth a "maybe" keep or sell will still have to be something you sort out.

If the power supply is the cause it would be surge related, and possibly hot surge vs cold (assuming all drives spin up at boot). It may get worse. How old is it, and what wattage? Did you change anything in how drive power is delivered? (Different power cable arrangement?) Drives, I'd explore a couple at a time SATA to motherboard direct, if possible. Just see if they show up, maybe self test. And maybe on a temp os (temp unraid trial?). Card failure seems less likely but not impossible. I lost a couple HDD (fortunately retired 500gb, I was testing some stuff planning a backup unraid server) not long ago to a perfectly functional power supply - 3 SSDs were fine, but 3 spinners (even two) and only one would work - and the initial failure toasted two drives. It was an old Corsair PSU. Caps and solder joints and who knows what do age out and need attention - in this case just a new PSU.

Wild guesses: * Cables not fully seated or perhaps even damaged during the swap * Power supply pushed over the edge, possibly now unstable If nvme/pcie share lanes, devices simply won't show up. I've never seen it myself but esd damage is technically a possibility also.

Just use smaller disks for the entire array. You can't split a big disk into two partitions and read from both without a performance hit.

Custom means you're on support, not to mention being responsible for picking parts and assembling. It can be exactly what you want though. You're really the only one who can answer this question for you.

Even with logging for home power consumption, untaid, and unifi (to influxdb) and multiple years worth of data, a win 98, 10, and 11 VM (3 total) I'm around 300gb... Only problems have been with a pihole-unbound docker going sideways and filling up docker allocation with logs. I don't do the torrent stuff though. I do use a different pool for write cache.

Pay attention to actual number of pcie lanes available - that's why you don't see more than 2 usually; using both typically makes them x8. Sometimes the slot size is 16 but actual lanes x4. This one is claiming 1 pcie 4 x16, 4 pcie 3 x16... Maybe? I'm skeptical. Realtec lan would have me stay away, but there is seemingly room for an add-in card if it's trouble. And onboard video would depend on your CPU choice.

The difference is how it reaches 8 ports - that may impact power consumption. Hard to know and expensive to validate. I'd agree either is going to be a solid choice - though supermicro may be in trouble? So perhaps not the better option anyway. In the end, set your goals, commit, and don't look back.

Supermicro MBD-X13SAE series might be an alternative if you're unsure about the Asus. More onboard SATA, but IMO the best pcie layout is the -F-O which brings IPMI (something I'd want, you may not). More $ though, and a PCI slot... Not sure if you wanted 2.5g onboard but that's another difference.

I see 2-3w power consumption on the switch from SFP+ to rj45 adapter when I fire up a computer with 10gbe (two different adapters, two different computers). Less than half a watt for similar but fiber DAC, and even less for copper DAC. Thanks for the PSU links. Will see how they hold up over time, that's where I've had stuff go south. With attention you can certainly tune AMD to be reasonable. My network stack (including server) as a whole is still a huge chunk of my annual power bill - not every day, but it's constant.

I can sympathize with the struggle. I run an asrock x570-d4u/5600x/32gb ECC and wonder if ECC and IPMI are worth keeping... idle for me is a mean of 46w, low of 38w. It used to be worse though (HBA card changed asm1166 expander, settings tweaks, fan curves the big improvements for me). I could lose fans at the expense of drive temps (still safe levels, maybe I'm too worried there) but the big issue is I can't idle to stopped fans. Your Intel picks wouldn't idle as low as a different chipset, but then you couldn't ECC. 8w adds up across everything, but I bet with the AMD you can ditch the HBA so that'd be a power trade. Maybe you'd be fine with a b650 or whatever is the right one here - just will be fewer pcie lanes I think. Skeptical on the power supply too - any documentation/independent testing? I've been burned so many times by PSU brands I don't recognize - long ago started being a lot more cautious.

Losing the Intel for transcoding may not be worth the trade, especially looking at power efficiency. Then again, it may be you'd need separate graphics for 6 transcode streams anyway. Not sure as that's not something I deal with. And I say that as someone who would only reluctantly buy Intel (CPU - no one better for network cards)

It's definitely better than older Intel chipsets, and even in my AMD system (x570/5600x) slightly better than mellanox connectx-3. I'd guess yes, but don't have data to back that up. At worst it won't be worse.

If you can find them, I believe the 750w seasonic titanium rated PSU is among the better ones sort-of still available at low power levels, and the 850 only slightly behind. Also top tier supplies. You will spin up all the drives at the same time at boot. For power efficiency, consider sticking with intel x710 family SFP+/10g. Well priced on the used market. May not be significant with what is noted about c states here though. Also evaluate whether you need that HBA card or can get away with onboard SATA and a more boring sata adapter or two - it may also keep power consumption down. I use a 6-port asm1166 m.2 adapter and shaved over 10w vs 8 port 9200-8i, no observed loss in throughput. Only 5 spinners for me, but 5 SSD for file shares (1 is just music for my squeezeserver, 4 are raid10 for array write cache and active/fast shares. Shares are backed up to the array still (plus more layers of backup)

Epyc may be a better option on price, especially in the used market. Hard to beat when a big need is lots of pcie lanes and x16 slots. You won't necessarily know if ECC would help unless you're monitoring for bit drift etc. in other ways. If something only lives in memory, you may just never know. And you may be fine without. Case... Just make sure you're going to be able to move enough air for multiple gpus and everything else. I have a rackmount on a wall (bottom of the case to the wall, airflow left to right). So much easier to work in and SSD temps down 20° (previous case was an older lian li midtower)

SFP+ is still super rare onboard, so pcie it is - and even onboard 10Gbe is rare and pricier (comparing same board, one with onboard 10Gb, one without) than pcie. I do also use onboard on my backup server, but for WoL.

Which adapter? I use one with asm1166, no issues at all. It's not with the mainboard you're using though.

I agree, sounds like a power supply issue. I have gotten so picky about what brands I'll buy any more - I've just seen too many issues over the years. I have an old (near 20yr) Corsair power supply which did the same thing, though in my case it only supported one spinner - could get 3 SSDs just fine. It also fried two drives in the process of testing.

Rosewill makes a handful of 4u rack mount cases flavors - I use a RSV-L4500U (15 drive spots, no hot swap) and have swapped the stock fans out (including a handmade mount for 140mm fans behind the drive bays). Not swap costs more but one of the variants is 12 hot swap. The stock mid-chassis fan setup can be quieted by flipping the fan mount, and I can't speak to stock fans noise. I previously had hot swap bays and found the rare use case where I used them didn't justify the cost. If I'm actually doing maintenance I'll have the system powered down, and they add airflow restriction. The 5 drive cages are pretty quick to remove (cables are the trick). I use server cable bundles (typically 4 or 6 in a bundle, no coating just emf wrap so much less bulk - same as you'd get if using an HBA with SATA). Extra 3 bays this way. Only 10 drives in mine, they stay low 30s temp doing parity check and fan speeds still slow.

I run music on an unassigned device - same SSD I've used for years, just in a new server. I simply back it up to the main array to save myself time on what I'm sure will be the eventual required recovery to a new drive. At that point it'll likely head to a pool.