There's a certian level of absurdity that is using 6x 1.5V lithium AA cells in a doorbell camera.
Esp as the cells are really ~3.7v native with an internal buck convertor (And yes the ones i have are ok with that many being used together)
GJ Gitlab, apparently I should update to 19.4.2 due to a venerability it's still saying that despite bing on 19.4.2
I must prefer forgejo thesedays its so much lighter but migrating that specific instance would be a headache.
Looks like link/sata controller issues rather than flash errors.
ata-Patriot_P210_2048GB_*Removed* DEGRADED 3 146 0 too many errors
Read/write/checksum errors.
Yes ofcourse it would be a 2TB drive throwing errors, and not one of the smaller ones.
Ah yes typical large energy company
On an email correspondence about an new tariff, call us if you have any questions, provides a number which goes to customer service who can't deal with that tariff yet.
Well done 🙃
Well that's surprising the bazzite deck Nvidia image seemed to "just work"
I thought gamescope didn't really like Nvidia.
So apparently my Bess showing 100-200w of extra house load is a known issue :-/
Presumably there's a bug with the way it's calculated because it's definitely not a measured value they only have one CT and that's the grid input to the house.
Apparently a ddr4 sodimm will run alongside a ddr4 dimm
At least on my weird Frankenstein pc in the living room.
I was planning 2 sodimms but the CPU cooler got in the way
Was thinking odd that predbat didn't charge the batteries last night given how dull this morning was.
Erm yeh that's because I did some work on the server and forgot to start hass back up. Whoops
Supposedly brighter this afternoon so should make it to off peak (assuming it actually is)
HMm i think the X16 slot on my ml110 g9 is fucked.
GPU wont emuerate in it
HBA wont emuerate in it
X520 wont emerate in it.
Werid chinese B650 x4 southbridge card, oh yeh it's fine in that slot
I wonder why they don't design AI models as specialists which know when to say "I don't know" and then have a router that can try various models.
I know moe kind of does that bit even those are still carrying round tons of useless baggage as you still have the entire model.
Would it not be more compute efficient rather than having to have the entire massive model loaded on all the compute nodes.
Surely then the bits that are hardly used would be running with less replicas and using less
Hah so the IDS/IPS on unifi gateways seems to only do something if it's a directly connected (local) IPv4 network.
For a subnet thats via another router (even if it comes in on one of the "Protected" Interfaces = nope
Ipv6 traffic also nope?! (my wifi2 vlan is directly connected)
Lets see who spots it...
How many cockapoos? (The answer is 2 they're both brothers, littermates infact)
Oh wow the patch to the alphaess integration in predbat I had the slop machine do got merged
That said it did fix the charge hold not working within the limits of what's possible with the available vendor APIs and the project uses llm contributions elsewhere
In the case of particularly insidious bugs this bug spawner doesn't render on a battle mage gpu under Linux luckily I knew I'd run into this issue before so switched pcs to one with a Radeon GPU
The game does render correctly in windows but yeh you end up with loads of the enemies and wondering where they came from :-/
AI = actual idiot?
Ah yes the chicken disposal service.
Forgetting you put the machine you are migrating to on a smart plug temporarily and then telling the voice assistant to turn off everything.
Opps
When one server is being the switch for the other one, because the local switch in that corner of my house is out of 10G ports
Also i'm swapping the R730 that's currently standing on it's side for a tower server thats slightly location appropriate, and possibly slightly less power hungry being a single CPU instead of 2.