• 4 Posts
  • 135 Comments
Joined 1 year ago
cake
Cake day: June 12th, 2023

help-circle
  • This is a good way to do it.

    I went one smaller with the Node 304 which only can do 4 HDDs with a GPU inserted. Going used for consumer desktop CPU is the most powerful play for the money I think.

    This is a good path forward OP for a pretty powerful server

    • Node 804
    • Used AM4 motherboard ( microatx B550) (can be around 150€)
    • used 5700X or similar (seen as low as 100€)
    • new 500W power supply
    • 32GB DDR4 3200 ram in 16GB sticks
    • WD red plus 10TB helium filled for balance of noise and performance and price. My 10TB drives are as quiet as my 4TB. My scheme is ZFS mirror of 4TB (2 drives) for important docs, and 10TB drives for non critical data. Drives are by far the most expensive unless you get good second hand drives
    • if you want to do Jellyfin media server, pick up an arc A310


  • You said it right there in your comment.

    Sleep mode, (and other effectively off modes) where it is functionally useless, it can do.

    MSP430 can do 140uA/MHz. That is ~7 times the power that this application supplies, and that is not counting any single other chip quiescent current or chip that actively provides useful data. You would have to have a battery anyway or a big cap to provide the needed current for on-states. Or you could run it extremely low frequencies like you said, but those tend to not scale linearly at all with per MHz power ratings. Quiescent currents tend to catch up fast at that scale. I would be extremely doubtful that 150kHz would scale perfectly and wouldn’t have already exponentially decayed to around its lowest possible on-state consumption for the chip. I would definitely have to see tests on that.

    The smallest of batteries like the VARTA tiny cells in TWS’s are infinitely more useful and practical and it would take this application months to fill a single cell, discounting all losses.


  • It is the Mac of network hardware in my corporate - entered experience.

    It is aesthetic hardware, marketing, and everything software related looks polished on the surface, but is buggy (particularly their access which is the worst thing to be buggy) with the least possible configurability, completely obscured debugging resources, and proprietary ways to make you reliant on their support services.

    That being said, I am still using them because I got a 30€ UAP-AC-SHD from my company’s old stock when we switched to Cisco hardware. And their cloud gateway ultra is a good value. My whole house setup with prosumer hardware will be 140€ and where my internet comes in is the worst place in the house to put a wireless router.


  • I don’t think people realize how extremely little 50uW is.

    For a standard 3.3V microcontroller assuming a 95% efficient voltage regulator will be a current of 14.4uA. Just having the HSI master clock enabled on one of the low power STML0 chips is 15uA. This will literally only the clock. That is 0 sensors, 0 communication, 0 IO, nothing useful at all. For reference, reading SPO2 with a very efficient maxm86161 takes 10uA by itself in ultra low power mode with low accuracy and not counting the max leakage current of 1uA. For full operation, you need about 1000x-10000x that amount for short bursts.

    “Oh but it can cHaRgE tHe BaTtErY”

    Let’s say the device has a standard 100mAh battery (apple watch had a 228mAh or more). At 100% efficiency with absolutely not one millijoule being used by any other electronics (which would never ever happen, it would at the very least need a boost converter), it would take around 277 days to charge up that tiny tiny battery.

    Let’s take another example of an even smaller battery. To charge one side of the airpods 3rd gen (0.133Wh battery), it would take 110 days per ear

    This is one of those free “energy harvesting” fad BS based in nothing but wishing and marketing. It is an interesting learning project for wireless antenna beginners, but that is the extent.




  • That is a different usecase though. That is simply syncing local musical with a server.

    I do that too because i have an SD card. Just use Syncthing for that. Much faster and less hassle. You can use any music player on your phone that you want, not just one that works with jellyfin.

    If you aren’t streaming music in real time for the majority of time, then do a phone sync, not a streaming server.




  • JustEnoughDucks@feddit.nltoTechnology@lemmy.worldWhat the hell Proton!
    link
    fedilink
    English
    arrow-up
    5
    arrow-down
    1
    ·
    1 month ago

    People can’t learn not to throw trash in the street, climate change that is backed by decades of science is a problem, or hell, they can’t even learn to effectively not click on super suspicious phishing links.

    How on earth are they going to learn about implementing encrypted DNS when most barely know the difference between a browser and a computer.




  • JustEnoughDucks@feddit.nltoSelfhosted@lemmy.worldNetwork Switch
    link
    fedilink
    English
    arrow-up
    5
    ·
    edit-2
    2 months ago

    I think it has to do with data differences between self hosters and data hoarders.

    Example: a self hosted with an RPI home assistant setup and a N100 server with some paperwork, photos, nextcloud, and a small jellyfin library.

    A few terabytes of storage and their goal is to replace services they paid for in an efficient manner. Large data transfers will happen extremely rarely and it would be limited in size, likely for backing up some important documents or family photos. Maybe they have a few hundred Mbit internet max.

    Vs

    A data hoarder with 500TB of raid array storage that indexes all media possible, has every retail game sold for multiple consoles, has taken 10k RAW photos, has multiple daily and weekly backups to different VPS storages, hosts a public website, has >gigabit internet, and is seeding 500 torrents at a given time.

    I would venture to guess that option 1 is the vast majority of cases in selfhosting, and 10Gb networking is much more expensive for limited benefit for them.

    Now on a data hoarding community, option 2 would be a reasonable assumption and could benefit greatly from 10Gb.

    Also 10Gb is great for companies, which are less likely to be posting on a self hosted community.



  • Do you have the Intel drivers installed on your machine? Are GuC and HuC working?

    sudo reboot
    sudo dmesg | grep i915
    sudo cat /sys/kernel/debug/dri/0/gt/uc/guc_info
    sudo cat /sys/kernel/debug/dri/0/gt/uc/huc_info
    

    On Debian I had to manually download the i915 full driver Zip, extract it, take out the Intel drivers, and put it in /usr/lib/firmware

    Then hardware acceleration worked on my Arc380.

    If you use QSV, your CPU iGPU will be the one that can use it, so make sure to set your render device in docker to the iGPU and not the RTX 2060





  • To be fair, if something is open by default or very easy to enable without informing about the risks, tons of people will have it exposed without thinking.

    It isn’t that “tons of people do it so it is normal and perfectly fine” but more “people don’t realize.” It also uses some nontrivial amount of resources to process and block those attempts, even if they never have a chance of getting in.

    There is yet a reason I can find to have it forwarded for home use. Need to ssh into a machine to fix it? VPN.

    There are plenty of secure web-based tools to manage your server without a VPN also.