HN user
I own a dedicated 48vCPU with even 160GB RAM.. its not that expensive, check ebay, maybe now with mem prices it will be a bit more steep but as a hobby it's not crazy to think one owns such a piece of hardware. My dual GPU setup was more expensive I think.
I bought two RTX3080s with 20GB during my holiday in china (set me back 700euros) I'm getting 800-1000 input tps and 60-100tps output with Qwen 3.6 27b Q8 (MTP, P2P, 200k context) this feels like opus4.5 level while coding (pi harness). Also easy to just host your own openai compatible api from home this way and still use your MacBook as dev station.
Ok, i went through the P2P rabbithole, and incase anyone reads this, just use Fedora Rawhide, and run that repo install.sh mentioned in the blog post :) Now i have:
GPU0 GPU1
GPU0 X OK
GPU1 OK XGraphene is NOT a jailbroken/rooted OS, its a real secure unrooted, bootloader locked OS, and MS Authenticotor works just fine. If anything does not work its related to dependency of the App maker on a certain attestation google play services grapheneos.org/articles/attestation-compatibility-guide
I'm using bazzite on my ai-rig just because it has the gpu-optimized things setup (also nvidia-open). Looking at P2P seems to be available only for 90-versions of the nvidia rtx gpu line, not 80, and some versions of 50xx? (apparently the 5080?). Anyways, i downloaded that uncensored model and tweaked those kv settings etc. still getting 60-80tk/s but im able to get my context on 180224 now, used to be 131072 which gave me some trouble, this is already a win :)
$ nvidia-smi topo -p2p r
GPU0 GPU1
GPU0 X CNS
GPU1 CNS X
i guess not, i use llama.cpp with:
--spec-draft-n-max 3 --spec-type draft-mtp --split-mode tensor --tensor-split 1,1
and my (gen) tk/s are between 60-80 tk/s
will test this uncensored model and ngram added as well this weekend
btw, i also set my powerlimit to 220watt per card (with nvidia-smi) that will cost you around 1 tk/s but safe you a LOT of power and heat :)
I bought two 3080/20gb and one of those MACHINIST X99 mainboards as well (one with two full x16 pcie slots) those boards come with a xeon cpu included (for the pcie lane support) it set me back 800 euros total (had a spare psu, ssd and mem in a drawer) and now im also happily running 80tk/s Qwen 3.6 Q8 (MTP).
l'm curious if Bartosz will actually contact him for that final casting, i know i would definitely love to own one. And mainly because I love mechanical watches, especially what's inside, but I don't like to wear or really use them, this is a great solution for this "problem".
How does bitwarden’s backend being build in a gc language matter? There is also an api compatible project in rust called Vaultwarden if gc matters. Is 1password and lastpass’s backend better in that regard? would be nice if he could elaborate on that point.
And to add to this, i learned to not use the 'bluetooth' icon to manage the AirPods, instead use the 'triangle with broadcast signal'-icon found next to the volume in the control centre or on many other different locations on your ipad, iphone or apple tv.
Just turn off the auto-connect feature and start using that icon, all the mentioned 'annoyances' will go away :)
- Frontend: Angular 7+
- Backend: NodeJS with Hapi and Mongoose / Mongodb
- Auth module in Hapi (JWT)
- All in Typescript and VSCode
These libraries have been stable for a while now. Tried GraphQL / Apollo for a while but the amount of breaking updates made me go back to the stack mentioned above. Currently i'm exploring dotnet core (in C#) and blazor.
This is exactly what I was looking for. After searching on ebay for the HP 7550/7475a i noticed they are generally only available in the US. Since the shipping costs and import duties to the EU are really high for US-ebay products and most of them did not guarantee it would work I started looking into alternatives.
So I tried my luck on the 'Drawing Machines'... But they are (in my opinion) very slow, suffer from the same 'do your own maintenance/support' like current 3D printers and brands like AxiDraw, Eleksmaker or Mackerblocks are crazy expensive compared to the 3D printers which contain much more electronics and material.
Also looking into their 'boards' I noticed they use different firmwares and mostly leaned on SVG/Inkscape/Python and i'd rather have the HP-GL to mess around instead of getting lost into this firmware/library rabbit hole.
Then I stumbled on the Silhouet Curio while looking through youtube videos. It has so many options.. whole 'home' businesses are build around this device. Looking closer into the electronics I noticed it uses a proprietary language/driver. However it does have a small active tech community that was able to reverse engineer the protocol and created an Inkscape plugin.. but that will make it on par with the drawing machines, and I'd rather have a documented protocol from the get go (like HP-GL).
I almost gave up on my dream of owning a plotter and tinkering with fractals and the HP-GL on a rainy Sunday afternoon. Until I decided to read this thread that started it all once again and read your comment about the USCutter, this thing is exactly the right price and seems to actually use HP-GL, awesome, my search continues.
I've lived there for five years, and like you, I'm always very surprised by all the articles and comments made by people about the Bijlmer (about it being the worst place ever..) But all these people tend to have one thing in common, they have never lived or even visited the Bijlmer, or indeed tend to be a bit racist. It's sad, because it is not a failure at all. In retrospect I even believe it's one of the most unique places The Netherlands has to offer.