The table you’re referencing leaves out CUDA/ tensor cores (count+gen) which is a big part of the gpus, and also not factoring in type of memory. From the comments it looks like you want to use a large MoE model. You aren’t going to be able to just stack raw power and expect to be able to run this without major deterioration of performance if it runs at all.
Don’t forget your MoE model needs all-to-all communication for expert routing
These are just off the top of my head. Best case scenario the blocking does work and the teen never tries to bypass it. They’ll still just move onto “wasting” time on something else. This is treating the symptom and not the root cause.
Pihole can set up “groups” for different blocklists. You specify client by IP or MAC address so it doesnt matter what the dhcp server is, so long as there’s a static IP or static MAC address. My pihole server doesn’t have dhcp set up and I’m able to do this fine
Though from personal experience this just becomes a game of cat and mouse, and if you have a motivated teenager then they will find a way to circumvent this. For example android can rotate MAC addresses, and IP addresses are trivial to spoof as well.
Haven’t used all of those but my recommendation would be to just start trying them. Start small, get a feel for it and expand usage or try a different backup solution. You should be able to do automatic backups for any of them either directly or setting up your own timer/cron jobs (which is how i do it with rsync).
I submitted a response but if i may give some feedback, the second portion brings up:
I am willing to pay a substantial amount for hardware required for self-hosting.
This seemed out of place because there were no other value related questions (iirc). Such as:
I’m sure you could also think of more. But i think it’s pretty important because between cloud service providers and any non-free apps you want to use, it can be quite costly compared to the cost of some hardware and time it takes to set things up.
The rest of my responses don’t change but if you’re wanting to understand the impact of money in all of this, i think some more questions are needed
Best of luck!
You’re not connected to wifi or vpn from the looks of it. jellyfin is hosted on your local network. You need to be connected to that network for any device you want to access it. The most direct way is to connect via wifi. If you want access from outside your house you’ll need to look into opening a remote connection via something like cloudflare tunnel
Logseq to some extent, but it’s set up to be a journal/ meeting notes where you tag pages, add documents, etc. it would be up to how you’ve tagged things. Does have a graph view of your pages and whiteboard feature.
Personally it wasn’t exactly what i wanted out of a PKM but it is really powerful. It’s intended to handle taking notes efficiently from meetings and then somewhat self organizing the notes as long as you tag stuff.
Foundry was the 2nd thing i started self hosting (the first being pihole). Have had it running for 5 years now.
Other than that i only recently started expanding my self hosting:
I would use cloudflare pages (or any forge ‘pages’ feature) before using tunnels for a static website
From a user experience its a social media site, like reddit.
And an ELI5 for the technical parts:
What is the relationship between Radicle and the Radworks ($RAD) token?
Radicle is a true peer-to-peer protocol. It doesn’t use nor depend on any blockchain or cryptocurrency.
Radworks, the organization that has been financing Radicle is organized around the RAD token which is a governance token on Ethereum.
From the FAQ in case it’s relevant to anyone
This is what i did. There are many static website generators that can help. I use Hugo which let’s me write in markdown, download themes (modify if i want), and it builds the site which can be hosted for free on codeberg/cloudflare/gitlab/github ‘pages’ feature. All support letting you use custom domain if you have one.
Codeberg pages comes to mind (for a simple personal site anyway)
Looks fairly impressive, including live collaboration
Im not familiar with doku wiki but here’s a few thoughts
tldr - lesson learned. buy a new domain and move over to it.
but for those who want to learn something new - you are only renting your domains. If you fail to pay by the registration date then you generally get a grace period to pay more money to renew it. If you fail to pay before that period expires then the domain will be released. Some companies like godaddy will automatically buy the domain for another year (or more). But even if Godaddy doesn’t then it still goes up on a list of expiring domains and there are backorder services that will try to buy the domain or auction them off.
So in the end it doesn’t really matter what registrar you use. If you do not pay, it goes back to a list where people can see it is expiring and then you’ll get some people who either want to legitimately use that domain or more likely they are wanting to try to sell it to you or someone else for more than they buy it for.
And I saw someone mention file a complaint. I’m sorry to say that if you did not have money to renew the domain then you aren’t going to be able to do that either. This is called Uniform Domain Name Dispute Resolution Policy (UDRP) and the fee is between $1500-4000 for 1 to 5 domains.. Additionally, just because you file a complaint does not mean the issue will be resolved favorably or timely. These complaints can last years, and there is no guarantee you will get the domain back.
This is why you should always pay your domain rental fee. And if you don’t, then you need to either be willing to pay a ton of money to get it back or you will need to move on. Sorry its a tough lesson to learn but if you’re just a student then you probably weren’t using this to run a business or anything so in the end you are quite fortunate.
Tflops is a generic measurement, not actual utilization, and not specific to a given type of workload. Not all workloads saturate gpu utilization equally and ai models will depend on cuda/tensor. the gen/count of your cores will be better optimized for AI workloads and better able to utilize those tflops for your task. and yes, amd uses rocm which i didn’t feel i needed to specify since its a given (and years behind cuda capabilities). The point is that these things are not equal and there are major differences here alone.
I mentioned memory type since the cards you listed use different versions ( hbm vs gddr) so you can’t just compare the capacity alone and expect equal performance.
And again for your specific use case of this large MoE model you’d need to solve the gpu-to-gpu communication issue (ensuring both connections + sufficient speed without getting bottlenecked)
I think you’re going to need to do actual analysis of the specific set up youre proposing. Good luck