Also, there are some “former crypto miner“ boards that are configured with SUPER wide slots for video cards exactly like this.

They’re great and cheap used because nobody wants them.

If I have to build a second one, that’s my next path.

I have some nvlinks on the way.

Sooooo I’ve got a friend that used pcie-oculus and then back to pcie to allow the cards to run outside the case, but that’s not what I do, that’s just the more common approach.

You can also get pcie extension cables, but they’re pricey.

I stumbled upon a cubix device by chance which is a huge and really expensive pcie bus extender that does some really fancy fucking switching. But I got that at a ridiculous price and they’re hard to come by.

If I do it right, I could host 10 cards total (2 in the machine and 8 in the cubix)

This also means that I’m running 3x 1600w psu’s and I’m most at risk for blowing breakers (adding in a 240V line is next lol)

I’m rocking 4 used ones from 4 different people.

So far, all good

You can’t buy 3090’s new anymore anyways.

4090’s are twice as much for 15% better perf, and the 5090’s will be ridiculous prices.

2x3090 is more than enough for basic inference, I have more for training and fine tuning.

You want epyc/threadrupper etc.

You want max pcie lanes.

Given the price of P40’s on eBay vs the price you can get 3090’s for, fuck the P40’s, in rocking quad 3090’s and they kick ass.

Also, Pascale is the OLDEST hardware supported……… for how long?

Also, you’ll want to look for strange specific things to host multiple 3090’s etc… on your motherboard You want a lot of pcie lanes from your chip and board. You want above 4g decoding (fairly common in newer hardware)

Or you can do the dns challenge for letsencrypt

Check what your docker subnets are, but that shouldn’t conflict.

That sounds like your local network IP’s are conflicting with the default docker IP addresses.

What is your routers subnet?

Ahhhh now you’re talking kubernetes.

I mean you can do it with 2 machines and docker compose, but yeah.

If you have a docker compose, you can just bring it to a new machine with the storage medium and hit “go” and it’ll go.

That’ll probably be enough for a home setup and have a 1 hr downtime in a failure.

If you want “always hot” kubernetes is basically “multi-node docker on cocaine “

Damn, that addiction is strong lol.

I’m happy to help where I can but it’s a FUCKTON of knowledge and setup to go far enough to kubernetes it.

Docker-compose is 100x easier and gets you 95% of what you need.

You are going to want a single larger server and docker

Much easier maintenance

If you’re crazy, you’ll go with Kubernetes . I personally recommend it., I love it.

But I also work in it every day, so there’s a convenience there, but the complexity is off the charts .

Not a bad idea, but I’ll warn you that the addiction for homelabs is strong.

The n100 will make you sad eventually as your self-hosting addiction soars.

An older i5 with onboard gpu or an nvidia card will make you happier and not pull THAT much wattage.

Your Minecraft server will thank you.

Not really, no.

But this thing could be beat with a raspberry pi.

On Reddit, wallstreetbets used to call everything “retarded” and they’ve stopped and moved to “regarded” as a way of “almost” saying an offensive word.

He mentioned that he wants to create Second level Nat, that will require new IP addresses and DHCP in the subnet

TrueNas scale is a little heavier because it is a kubernetes implementation, but it is amazing.

The capabilities and possibilities are spectacular, and the maintenance required to keep all of your applications up-to-date is so easy that I no longer think about it.

What’s your cooling solution?

Could it be throttling the cpu?

2gb memory is not nearly enough.

Get a decent machine and run true as scale.

All of these things can be installed as helm/k8s “apps” and they almost self configure.

They update aswell.

It’s spectacular.

Then throw drives at it.

I have one “pool” of storage that’s a raid 10 for things I care about and then a giant zfs “jbod “ that has no backups for just mass storage of things that I don’t really care about.

While this is conclusively stoned as “cpu” issues, in case anyone else finds this thread…

While your isp can’t read the data over the VPN, they CAN see that you’re using a VPN and intentionally slow down your connection with traffic shaping because you’re putting so much data through the vpn.

What do the links look like on the start page?

The problem is that Tailscale gives your server a “magic” ip, which isn’t the same one as on your local network. On your local network, do you access them by port? Or reverse proxy?

Machine:8080 or service.machine.localdomain