#programming
1 messages · Page 445 of 1
it's not an obfuscation technique lol
they do
clangd-tidy do it
cant see it over the 432894721984137948713298 other warnings for my shitty garbagio
it's not esoteric at all
it is a technique i can use to obfuscate things
it's not intended for that
but it works. it stops people properly understanding it
yall sound like you've never written c or c++
does it count if it's exclusively poorly
by "it's an obfuscation technique" i mean "you can make it hard to read" not "they say if you want to make it hard to read, do this." i never said it was intended for that and said multiple times it wasn't.
if you deliberately try, you can make code hard to understand using this
fix available
trivial to fix, but you need to remember that int *a, b is not int *a, *b which doesn't always happen.
btw, with this: i tend to write code with 0 warnings, with //NOLINT(exact-reason-name)
anything can be made hard to read, multidef is the least problematic one
which is why we are tallking about it still
i jest
int *v is the easiest thing ever to remember and understand, i don't understand why it was ever an issue
Is it worth to learn asm, looks fun
But it's just feels like chinese to me
anything can be hard if you're lacking muscle memory/reps
-# no, really, that is real, you cant say it isnt
int * val; solved
I mean everything else will seem easier by comparison
also wrong
:dead:
int * val is most cursed
it is only thing every one agree is wrong
now it looks like a multiplication expression, and it might even be one
val kilmer inting it down mid
if you can't make everyone happy, make no one happy ;]
Okay
the only way to win is not to play
dont mind if i do
unique_pointer<int>v 
Calm down gng
std::reference_wrapper<int>v;
raw ptrs are non-owning btw
in modern cpp
C++ is easy
Pointer = point to that
Memory = safe
Code = good
waiting for std::optional<T&> ⏲️
c++ is actually one of the worst languages to exist
confucius say
man who put star next to type live in fantasy world, man who put star next to name see reality... but man who put star in middle have no friends at all.
going back under my bridge ;]
Star next to name for me

yeah, i kept typing it next to the name coz that's the right way.
I have my linter setup to correct it if I somehow miss it
just use auto
everything
just auto
I hate auto/var
this is a fantastic discussion
coz C++ has auto, why do we even have types? ;]
"Computer assume the type for me"
btw, i not use prettier at all
just use javascript for everything

🤮

Maybe if u like RCEs
if it's good for the windows shell it's good enough for me
idk how this corrector called
(not remember)
i dont use formatters
i just format everything myself
claude code is a react app which means react is the framework of the future
should remove all spaces: compiles faster?
i format with dd if=/dev/null of=/dev/sda
actually has anyone seen react lately, i think a wellness check is needed
ive not seen a react project in a while
I fell in love with cli lmao
gooood
cli is so beatiful
absolute true
my website is html and css, like god fuckin intended
at least 3 CLIs open at all times: building, running, and random stuff.
No php?
aint got time for dillydallying with your frameworks
i almost used astro a while back but ended up with nextjs again
why php for a static website
try vue
it is just modern modules system that most easier to understand and use
-# unless you need to use vuex or similar stuff
it's webscale!
the first 2 words were enough.
Astro is pretty close to doing a setup like that tbh it just makes style and script management easier
i do like wranglin raw js/html/css
That aura farming, when you open 3 cli and type some commands git etc to aura farming
nah i actually use php seriously, it's a good language
c++ has some nice features that i like
php package management may be worse the python
oh one thing
Php will be dead this year we swear
nope. that's a previous Linux user using windows.
Trust us
void *__restrict x;
php package management is actually really good, never had it break on me
Will looks something like this and you will 100000+ aura
Just because you sitting on your pc and typing git commands like it's your left lung
i just dont like pear
god damn, windows
who's using pear
I'm about switch to mx linux, this is my old pic
me last time i had to use php lmaooo
there's exactly 2 places i get php deps from, system repos and composer/packagist
neither one fails me
last php experience for me was mediawikii being dogshit so im biased recently
Write the whole website in CSS
I've got a cygwin window (for historical reasons), WSL, and 2 powershell prompts in an Anaconda environments. This is normal. it's not aura to try make windows useful.
mediawiki? yiikes
it is deprecated btw
restrict isnt though
not php fault though
it is...
Php ain't great but Javascript ain't either
php > js any day
windows supporting sudo and pretty much aliasing posix // core utils to the windows equiv is nice recently

Web dev community is given no new tech pretty much beyond frameworks
Idk what to make it useful, I have xmake, cmake, nodejs, lsp, tools things
Mostly I only uses git
i booted up my laptop to build something into .exe w/ pyinstaller for a friend and i forgor i was aon windows 90% of the time using terminal because it just worked
erm, restrict is only c now
one good thing that came from uutils is windows support 
but like, windows, why live in that hellhole
i dont
that's why i had to boot my laptop which is my only windows install
that i keep just in cases
format your friends pc to linux
if only
assimilate him
Plug in a rubberducky
we should write malware that installs linux
that's called the ubuntu live cd
year of the linux desktop one botnet at a time
ubuntu is not far from malware nowadays tbh
well they gotta eat yknow
code doesn't have as many cool bugs as it used to for malware to exploit.
Na its not that
needs stupid ppl instead of vulnerabilities
No fun malware anymore
i booted my windblows laptop and its time was stuck in february and then time service didnt want to sync because it had the wrong time so the micro$oft timeserver refused to allow the connection due to invalid cert
and the fucking service just straight up gave up and said "nah it's over" until i manually Set-Date
and then it installed updates for literally an hour
I have the same problem with my laptop that I boot like once a year
nothing out of the ordinary
http://www.twitter.com/danooct1
i hope you enjoy my great joke in the intro. thanks to Leurak for writing and contributing this one! it's SO COOL
Memz is a pretty fantastic computer virus/trojan that takes computer trickery to new levels. It's flashy, damaging, and hilarious all at once.
to submit your own file to be featured in a video, plea...
The last "fun" one
just realized my nerdsnipe sensor has being going off for the last 30 minutes. i should get back to doing things.
not a problem on linux, systemd-timesyncd doesnt care, neither does chrony, ntpd kinda did but if you -g it said ok whatever
they also dont use tls
sigh... i hate it when this happens...
why'd you install zoom, that's malware
so is chrome but it at least has some functionality as well
No librewolf?
it used to be
did they replace it with a completely different application
My shortcut for chrome is named "RAM Consumer 9000" and its only used for compatibility tests due to some issues i ran into with Alex
vivaldi is my consoomer of choice for my ram
btw, windows10 used to break my timesync (linux using hibernation), so i still have this shortcut
#!/bin/bash
sudo systemctl restart systemd-timesyncd
with latest (11.2025?) windows10 update that not happening
has all the consumption that comes from chromium
but instead of google being in charge
it's the guy from opera in 2006
perfect
Write ur own browser like a chad
vibe code browser
wonder if someone has attempted to do that lmfao
dont hear about it because it stops so far before being anything resembling a functioning browser
that they cant even hype it up
only issue is that you have to support extensions for js...
i am loyal to kde so i only use falkon
idk why there still no vim-movement-based browsers in normal languages
i only heard about one or two in py*hon
epiphany cringers stay away
cuz vim is stinky

is it better than librewolf, or you just like kde?
i was joking
i do like kde though
i have used falkon for funzies on occasion
it's not horrible
you know, with firefox lately it isnt sound like a joke
i vibe coded a browser! programming is dead!
look inside: chrome


they done electron'd again gawrsh dogget
i use zen browser, it looks a little nicer, based on FF and recently they said no AI in the browser so that's just a bonus
im sure there's a better analogue out there but it feels that way coming from zen

if i am setting up a new system im grabbing zen, im grabbing vivaldi, im deleting firedragon and floorp and librewolf and waterfox and all 329 other firefoxes
if i'm setting up a new system i'm grabbing curl and that's it
curl-impersonate is all you need
me having only librewolf on system :
Quite funny really
I'm so associated with NeuroSynthProject I basically am NeuroSynthProject to most people
dumb
fuck brave
Coming very soon
ayy didnt type it in time
D:
nah you did
I use brave, cus it has ads blocker
i appreciate you holding yourself to a standard

brave also contains a cryptocurrency and the adblocker has deals with ad companies
just install ublock origin like a normal person
adblockplus the browser
Dont worries I'll use firefox after I installed mx linux
ublock origin and sponsor block for any firefox based:
oh also they manipulate affiliate links
lul
whaat
Never tried firefox does it feels like
there was a whole thing about it a while back
yea grab the link reject your reality and subsitute their own
so like if you click on some sponsored video link it'll swap out that link for their link instead to get the money
i never heard about brave doing it, i only heard about brave's adblock being bad
it's also bad
feels like a browser, more stable than chrome ironically but google will make you think otherwise because they intentionally cripple their websites if you're not on chrome
and i heard they pooped their pants in 7th grade
zen + ublock origin + sponsorblock
are you talking about honey or brave?
i'm just waiting for ladybird to hit prod
brave's was about binance or something if you typed it in it would autocomplete a referral link
btw, ubuntu, really?
ladybird has some weird thing for me where it's not very fast for connections
like it's fine
but it has bad bandwidth throughput if that makes any sense
probably doesnt
average linux "dist" effort
ladybird is literally written from total scratch and catching up with 35 years of browser improvements, give it a moment 😛
kek

probably naive networking backend for simplicity
heard about ladybird for really the first time btw
yeah i assume it's just less optimizationslop
the latency is pretty scary bad tho
as well
i've followed it for a while, because i was interested in osdev, ladybird is a spinoff from a guys hobby SerenityOS project
they have some work 2 do on that end
they're prolly aiming for passing the test suite and then gonna go full on optimization
linux-inspired 
windows 95 inspired
euf, serenity looks awful for me
i like serenity, i prefer simpler times
half the time i use one of the win98/95/me themes for kde
but ladybird is in good hands, the guy is a former apple webkit tech
simpler is now
kde do best desktop ui i think
kde looks amateur as far as ui goes
not to say they are the best
zero balance
What DOES it support officially lol? And yea rocm afaik till 6.3.4 no?
gnome isn't that great either, it's too simplified, but i use it
xfce is kind of a happy medium, but some parts are too disconnected
i miss gtk3
official
does not mean good
honestly i think i'd have an easier time getting a couple mi50 to run rocm shit than i do the r9700
that being said 620 euro sounds like a ripoff i'd pay like 450-500 usd at most for a 32gb
at that point might as well yolo a mi100
i saw mi250x for mid 1000s the other day
not pci-e thou
you know, this little gap between edge and window, things like this make everything look much better
idk about widgets, never used them, but just basic de ui is good
this windows display and ability to bind them to super+q super+w is 500 points win
konsole being best terminal in the wild internet,
all icons and windows are good and i never had issue with them
0 lags or crashes after moving from debian to arch (was really rare no critical crash reports sometimes)
I wish..
620€ is litterally the cheapest i found for anything with 32gigs.
that 7 hunned dolla
only thing in kde is unclear linking to stuff like mtp devices automount
i still not know how to stop kde from mounting them
i do log out and perform jmtps in terminal instead 
Well ignoring Pascal and older with those 750 x4 @ 8gb a piece
maybe skill issue
Whoops, accidental ping
Oh? Whats that for a card?
this is meant for datacenter virtualization shit but it works normally
Aint the virtualizers like 4 cards in one?
For pcie passthrough stuff?
not in this case
What are they called normally?
it's a single gpu, 72 cu
Radeon pro v620?
Lmao..
Oo so vulkan should run nicely
will not work in pcs
🙁
it will
they say that because it's passive cooling and they dont want people buying it who are clueless
and try to return it
You need rebar and 4g decode + blower fan or shroud mod ig
tbf i dont need one, every time i flirt with the idea of getting a gpu for ai i know i will never actually use it
Whatsbthe radeon pro duo? Cuz the v620 is absolutely not doable for 450€ shipping lmao
i got this and a few other things from core4, they are legit
pro duo is 2gpu again
pro duo was contemporary with like
pascal
I personally dont need a big rig either, but my rx6700xt kinda struggles running anything useful.. and i would like to give my homeserver something useful to crunch with so i dont need to pay for mostly crappy privacy hating providers
Ah, aight. Essentially useless for me then, although ig 2x16gb ain't too bad, but the losses..
intel duo was certainly an era
you could get a v100 w/ 32gb
idk what kind of whacky prices they have for those over there tho
i have a bag of old duos somewhere
Meanwhile me just piling 3090s
yummy vram
A 32gig card could atleast run a 40-50B model nicely with plenty of context.. even a 24gig one would be nice for 25-40b dense.. and 120B moe wouldn't be too bad either on a 32gig card.. im torturing my 24gb ram laptop with its 6gb 2060 to run a 30B moe barely.. at like 7t/s
v100 is basically a slower 3090 (slightly) with more vram in 32gb form when it comes to inference
3090 slaps it in training tho due to ampere > volta fp perf and also having native int8/4
Ig a better investment than a 3090
Well training is exactly what I need
it'd still be better than a 3090 if your model cannot fit in 24gb though, which is my usecase and why i consider it basically a wash, cant stack 4 3090s with nvlink interconnect (although you can still stack them to an extent)
And i dont plan to train anything big, mainly userspace tools that i wanna make and homeserver stuff
i still think 2x v100 over nvlink ($100 each per GPU, probably another 200-300 bucks for the backplane for nvlink and being able to get it in a regular pc) is a dummy good bargain for what it is
I have the NeuroSynth, obviously
Though NeuroSynth as a model is actually kinda small
16gb ones that is
Yes i can make most userspace stuff with my 2060 especially with modern models if they feel like cooperating
Lemme see what they cost here lol
how's the 7900XTX pricing round you
that's got 24gb
rocm
but if you were already considering mi50 it's gonna probably strictly be better in like every way including current rocm use
Lmao..
Consider that a 6700xt costs like 300-400€.. ill check
Some of em are bids*
And most of these are bids
then you slap em on the backplane, connect to ur pcie slots via wiggly appendages
i just need them to put out a card that is not $10000 with 96 gb of vram
Does not look promising
32gb versions, 16gb is probably cheaper
what i was sayng is 2x 16gb v100 on nvlink
Yea, they dont seem to have em
that 167 one is probably it
GB10 
The nvlink costs more than some cards
that isnt a dgx spark*
it is most of that cost yea
the math for me is 2x $100 v100 gpu, + the nvlink backplane $200-300, total $400-$500 for 2x gpu worth of compute and 32gb vram
At that pricepoint im back to just one pcie 32gb card though.. but ig yea more compute..
this price range is pain really
if you can get up to like
$1000-$1500
it can get more fun
170€+170€+270€+100€ coolers + pcie adapter cards ??€.. seems bad
just move to the US
Yeah i dont have much to play with..
theyll send you back for free just fight a guy from the govt or something
Thats a whole different can of worms lol
Pretty sure my heart and stomach wouldnt survive, its already iffy here
Theoretically yes, rn no, and also not really for what i wanna do.. its quite annoying..
Ig ill have to wait until 24-32gb vram falls to like 500€
well you're not supposed to eat the worms
Lmao
what is your specific requirement for 32gb again
Running bigger models at decent speeds (25-40b) @ Q4-Q5
Selfhosting local ai essentially
gotcha yea i mean that would be about what you'd want
youwhat about a mac mini with big ram lol
Yea, rn im thanking qwen for their 3.5 series, absolute godsend that 35b a3b model, but my laptop doesnt run that with 24gig ram, (well actually.. maybe itll with linux im gonna try that soon, under Linux im short on ram..) and Performance is bad for their Qwen3 30b a3b.. 6-8t/s .. not terrible but good luck getting anything done with it
ig m4 pro starts $1399 damn apple fck off with that jump
Apple is way out of the price budget lol
Also dont they struggle big time with Long context?
I was looking at the TiinyAi Startup but 1600€..
Sure for what, 80gb vram?
Kinda neat
Better than a DGX i think
for inference specifically, idk the tok/s but it's probably better than large ram offload even with what it is
do not buy for anything involving training tho
really not bad for inference tho
cant train cuz it doesnt have tensor cores
12.8 tk/s still on 27b gemma2, in the 30s on 8b llama 3.1, llama q6 and gemma q4kmhttps://www.reddit.com/r/LocalLLaMA/comments/1eqfok2/overclocked_m40_24gb_vs_p40_benchmark_results/
inference only really opens up the options
Hmm 280€, thats quite ok..
Though at 50000 token context.. let me see..
3 minutes of context compute
Not bad, espescially with propper caches
Is there anything with 24gig and more compute?
For under ~450€?
And i assume i could bios mod the p40
I do have external flashers and stuff
3090 if you know where to find
everything seems to be overpriced where he is unfortunately
Should I revive this 10 years old laptop with linux, doesn't have hdd/ssd yet
looks much older than 2016
I did that with my laptop i got in 2007, put Linux on and it runs like a dream. Using it to do basic python stuff and my gaming laptop for heavier stuff
I only have a 500gb had, seems to be ok, Linux doesnt use up too much like windows does so that can save money
Omg what are these ssd prices
You have unlocked new role
My school pc 48 dollars is cheaper than ssd and ram bro
Lmao
hmm. Can I get your guys advice on something i'm making? idk if talking about my A.I streamer that plays dnd is against the rules or not XD but im assuming I can just talk about specifics in the coding in it. I'm trying to get her to basically roll a dice which I can get her to do but she doesnt really produce story results with it or even really combat results and I was wondering is it ok for me to ask for advice in here? lol im making this all in python
Windows minimum is basically 2TB for Win11
windows minimum is 128gb for c: + however much everything else needs
oh my god
this bug
c statement ordering
explode
// broken
#define ip_r(T) (*(T*)((uintptr_t)(owned.ip + 1) + (ip_off += sizeof(T))))
// fixed
#define ip_r(T) ({ T _v = *(T*)((uintptr_t)(owned.ip + 1) + ip_off); ip_off += sizeof(T); _v; })
// fixed but not as painful to look at
#define ip_r(T) (*(T*)((uintptr_t)(owned.ip + 1) + (ip_off += sizeof(T)) - sizeof(T)))
this was such a headache
Well for most people 2TB is an effective minimum
it was for patching this program
the worst part is i've already done this fix before in another codebase
phew
welcome to hell itll only get worse
I am currently just using whispers rn im gonna switch to RVC later
I think whispers is a good tts and just amy
or anything else vts capatible
if ur doing a vtuber model with it
That's all WSL is worth
leave local trade in orgrimmar
happy happy
soon ™
i don;t know if this is a massive coincidence and i'm just reading into it because im coping or something but, like a month ago i bought one of the "nvme ssd" sold on temu for absurdly low prices, like way too good to be true, obviously fake item with like a 4tb claimed ssd being sold for like $80/4tb even with prices for everything else, and they were designed to look like samsung 980/990 series drives even using their model numbers on the fake labels on the drives. idk if anyone cared at the time but i posted what i got for my $40 purchase of a "2tb ssd" which was a mostly inert demo board with blank pads
the furthest pad is populated but i was never able to get the controller to read/write to it despite it trying to
the controller was functional and reported back "2tb" but all writes/reads were nearly instant and all disappeared never to be seen again if attempting to store anything beyond in the small cache
i sent that to temu for a refund
which i got
but every time i saw them i got all pissed off for liek the last few weeks
so i've been on a crusade
contacting them over and over asking why they're ok with listing fake garbage on their site
and it's been esclated a few times
wait we can buy ram from temu?
and im looking today and it loooooooooooks like a lot of the stores and listings i reported are all gone
my cursor has started lagging
why are you building llvm from source
so i might've actually maybe possibly helped precipitate the removal of a ton of fake garbage listed there
and 21.1.8 at that
i found out that musl stdenv made a faster
i just bricked my pc 🙏
yeah,
usually about anything aliexpress can have, up to a certain pricing level, you can find on temu
what even is temu I was being told it was scammy
it's just a dropshipping marketplace

with a super gamified coupon/event system which is scammy imo but it's technically above board it's just obtuse about what you get for what you spend and the rules to claim anything are silly basically
what about the llvm libc

im getting bored of waiting

oh it just went up
not bored anymore
2300
wow
i also need to build lld
llvm has some gigantic object files so checks out i guess
huh. Maybe I might try upgrading my pc like this im lowkey getting desperate
im getting closer to wanting to start using LoRA
it used to be like 50% of this "top sold" in this category were all listings of the same fake shit "ssd"
and now i only see what looks like 3 left
this is how you end up with 2gb mlir binaries 
the rest are more or less legitimate
ive bought several of these drives lol
2500
personally can vouch for orico and fanxiang, ediloca appears to be fanxiang and also fikwot possibly topsel
seemingly either the same companies or jsut very similar operations, using YMTC flash
2600
dunno longevity yet cuz theyre new things in general but the fanxiang 880E i have is faster than my 990 lol
and also is the orico 7400e
2700
I'm an idiot.. lol no wonder I've been struggling so hard I have like almost no debugging tools in place im making my life way harder then it needs too be
TTS models are small enough that just grab any
running in the dark is indeed not ideal 
No.. lol no its not
3000
Don't explode
will know is ediloca is any good soonish

love bunnies lol
I wish I could have a pet bunny im just really bad at being gentle 🙁 my cats also hate me
IT OOM'D

i restarted it
the memory keeps going up
16gb
17gb
18gb
ye screw this
clang: 1400μs
gcc: 1290μs
gcc musl: 1180μs
throughput
surprisingly similar thing just change a few letters around ya know
forgot this was running lmao
I thiiiink i got it this time
Ask AI to solve your lifetimes troubles rather than re-reading the rust book's chapter on lifetimes
i just add more lifetimes until it compiles
Such beautiful syntax
its extra fun when you get a random github issue linked directly from the compile error
As Funyun Ring said, shits expensive here.. 700€ minimum
oh yeah i forgot to mention, linking clang takes like 50gb or so
or 40 i don't remember
chat
do i buy the broken 3090ti that sometimes gives red vga light on the mobo, but sometimes works?
it could be the PSU from the seller being bad, but it could also be completely fucked
that just sounds like nerd gambling to me 
its 450

if it was like half that I'd go for it maybe but that seems like a lot for such a gamble 
i can probably ask to go lower
rn im waiting for them to reply on whether its random, or when the card uses a lot of power
if its jsut due to bad psu that would be great
oh
@sage crag i need to remind you to talk about vm
how much do yall think i can get for the ddr3 ram
a single stick of what?
2gb 4gb upto anyones guess
oh right
uh
i was about to go eat
all i know its a laptop mobo as old as me
specs are geforce 8400m gs with core 2 duo t7500
mmmm very poweful
i found it via the serial number
its a 1GB stick of DDR2
holy hell
these use 1.8V, more than any modern ram
i think ddr2 is completely worthless
my 8gb ddr3 could buy me a porsche 
i was hoping atleast it could fund me this week's quota of snacks
i feel like you'd eat it on shipping selling it
sell it to the recycler
<- keeps purchasing 16gb ddr4 ecc sticks every couple days
get those 2ct of gold
when i sent my 4x16gb 3600 in for RMA, it was worth 750, now its already up to 970
i could by a (shitty) used car with this ram
at this point sanding the cpu die down to the shiny rainbow logic and throwing that under my microscope could get me more likes than i would get in selling this
i'm so glad i bought that 4tb nvme in 2024
im so glad i bought 20TB of storage when it only increased a litle bit
me irl
im seething i didnt buy 2tb nvme when i purchased my laptop and now am stuck with 512gb
oof
heres hoping it goes down in atleast an year
oh yeah, that reminds me
i keep buying 1-2 tb gen4 nvme as i see them pop up on a discount at all
I have a funny old corpo laptop I wanna put nixos on
i can buy 256GB m.2s for a 20 in cashies
so i got one of those lying around now
its smoll
i bought SSDs pretty much every year during black friday, good deals 
with the slight issue that the battery doesn't charge anymore
thats when i got my 24tb hdd
black friyay
i wish i had gotten a gen5 one before prices rose
i didnt have a reason to get one when i finally upgraded to b850 which supported the gen5 nvme since i had the 4tb 990 still and the 980 was hanging in there
shouldve grabbed one on the millions of sales
oh ye. for the pc with the 3090 for 1K trhat i was gonna sell the gpu of to superbox. guy started ghosting me
im getting my money back luckily
that's good
well, the getting money back part
nah its fine lol
we always had a rule in electronics here to buy electronics in person
ebay buyer protection my beloved
always served me well
i still cant find those fake drives on temu
maybe that will be my great contribution to the world
my legacy
buying in person is a bit annoying when i have to go to spain for it
tfw had to go to spain
thats like 60 bucks in transport for me the human
shipping costs skyrocketing

a package to spain is like 10 bucks
i got money back guarantee anyways, so ill just do shipping
i was also gonna buy some manga of a different uy, but they also started ghosting
vinted might not be that great of a marketplace
true
they're 2kg each, 10 books
guess it's pretty much useless now that the newer gpus just support this directly
it's similar purpose to the radeon v340 etc
it jsut looked pretty interesting at first glance
classic fans
chonky fans
i can already hear them
WEEEEEEEEE
screaming
how much current do delta's 40mm fans draw again
i love the way these end up configured based on whatever circimstances surrounded them getting installed
wasn't it like 5A or sth
did I post the old tesla's I have here idr
there's a guy on here wanting to exchange his 4070ti super for a 5090 
why is there a traingle thing on it

pull air from above it seems like
sorry, im everywhere
vedal levels of wealth

they're old teslas and I kinda don't know what to do with them
is vedal a millionaire already?
I'd like to use them for video encoding but ugh
he's jensen huang
idk if they can actually help with encoding
is anyone a mini pc fan expert
i emailed firebat asking what kind of fan they intended to have mounted in the 2nd chamber of its case
and
they emailed me this
the dell otiplex has got to be the most annoying thing ever.
i search 5090 and get dell optiplex 5090s for sale
same with 3090
that looks like a fan
damn
gimme a model number at least
37mm mini PC fan JST 1.25 3 screw
5v dc 7.2mm thick 46 mm total diameter
searching this
is
not very reliable
lmao
try "router fan"?
i can get that far but it's rough
looks like it could be https://can.grandado.com/products/t124010sl-2-diameter-37mm-30x30x23mm-dc12v-0-10a-3-lines-cooling-fan-for-soft-router-cooling-aluminum-sheet-graphics-card?variant=UHJvZHVjdFZhcmlhbnQ6NDM2MDAxMzk5
nah that's just wiring
not quite actually
this one is closer https://www.newegg.com/p/0SS-01XX-012G4
it has the overextended legs
yeah it has to be specific 147degree/67degree spacing
cheapest 5090 is 3250
but there is a "5090 PCB incomplete" for 500 locally 
but what the hell does that even mean?
probably not got the core or the memory
@olive sable
hi
it's a firebat but not this model, i dont think this fan is it but i assume it'd be similar in form even tho the diagram is just the bare fan
i see
right now
okay
both problem solve
code so far if interested in various abominations
as much as having a 5090 would be cool I just can't see a world where its cheaper to buy one than to buy a good gaming gpu and just rent a 5090 for ai
what the frick
rrrr_op
raar
raar is dutch for weird
the instruction implementations are based on some of my previous work
i see
dutch is english for made up
like for the difference in price between my gpu and a 5090 I could rent 4k hours on a 5090

speaking of abominations i should start implementing bash scripting for the web
they have read it, but aren't replying
the difference between the previous work and the current work is:
- the previous work dispatches using a LUT with computed goto
the new work pre-translates the binary and then does computed goto without a LUT
- the previous work runs one task (though it technically allows for multitasking, the runtime isnt part of the vm)
the new work is multitasking by design
aside from that, the old one was a library and had design limitations, this one does not
haha very funny nvidia
theyre similarly optimised, but the older one is much easier to optimise since its so simple
the broken 5090. guy says he got them from amazon in pallets. no sign of life
i think i shared benchmarks here before, but here's a newer one
i think the last time i posted it, it was 1300 microseconds
nice improvement but mostly splitting hairs
since 1024 * 256 tasks is too many
wrr
its not entirely complete yet either
turning off jump, loop, and, function alignment makes a tangible improvement to performance
down another 20 microseconds
i was googling something unrelated and this came up
@tender river 


I feel like the nix store would crush the ntfs file system
based on nothing but vibes
nt filesystem
so this is a case of smh my head

NT stands for New Technology I think
ye
filesystem
Windows “New Technology” (NT released in like 1990s, right?)
fart stinky

fs

they also deleted the lisitng and their asccount is 0 days old 
i wonder what would happen if i go after it in person
lemme at em
cuz then id be paying in cash
you'd be getting your cash stolen perhaps
ye i think so
i told them im only paying throught he built in platform, and its too far to go after it in person
should cover all my bases
never pay off platform
You ever think about how fucked up it is that the best display we all own is on our phone in a form factor that's too small for us to ever truly appreciate it

It probably still doesn't clear the multi-thousand nit HDR performance of your phone
it does
Wtf
my phone cant blind me, the projector can
Well, your phone is a lot smaller so it is proportionally a lot harder to blind you
a projector is apparently usually between 50-250 nits
but that's brightness on the projected imag i think? which is the fair comparison
Yeah, case in point my OLED monitor maxes out at 250 nits in SDR content and that is enough to be uncomfortably bright, but the same brightness on my phone isn't
my phone is 800 nits, but its 5 years old now and oled degrades, so we're probably a bit lower.
my projector got somewhere to 900nits
My phone is 2200 nits max in HDR where my monitor is 1000 nits max
mine doesnt lol
and the pixel density is crazy
And I will never appreciate what that does for specular detail because the screen is too small for them to truly matter anyway on my phone
And yeah the pixel density is mad on phones
for pixel density i have another projector 
the lightcrafter 4500 has a really close focus
my pixel 10 (pro xl version spec here idk about the other ones) is oled, 486 ppi, 120 hz, meets hdr10+, 2200 nits in high brightness and peak 3300 nits
im still rocking the smasnug a52
Pixel 10 Pro here, roughly the same specs
i figure probably pro would be sameish but size diff
compared to non pro
who would haev their own version of the screen ig
I usually keep my phone's that long, I only upgraded because my last one had developed a screen fault where there was a giant white line down the image
my 4k 32 inch monitor is 140 ppi
And honestly upgrading on contract was cheaper than replacing the screen anyway
my phone has the fun thing where the back plastic is not atached
i think my monitor does beat it in rsponse time
4k 27 inch here so around 163ppi
They should be roughly on par in that respect, response time is down to the nanoseconds on any kind of OLED technology
im gonna buy a new phone eventually. just not in the budget rn. and it stillw orks so who cares
well and refresh coupled, it's a 240hz
Either way my phone still dwarves it at 495ppi
yeah
ofc it's about view distance too
i cant find a damn number for response time on pixel 10
Yeah I have a shallow desk which is why I went for 27 inch 4k
just that it has "very good" response time
As for my pixel 10, I can literally put my eyes as close to it as they can go without losing focus and not see pixels (ironically)
i feel like 1080p on a phone is already enouhg. 1440p on some of them is really pushing what you can actually see with human eyes
Whereas with my monitor if I put my face absurdly close I can still pixel peep even at 4K, not that it bothers me
The pixel 10 pro is 2856 x 1280
It's fairly close, yeah
Obviously there's no one-to-one with the weird aspect ratio of mobile phones these days
roughly same amount of pixels ye
Xl is 1344 x 2992
1440x2560 = 3686400
2856 x 1280 = 3655680
So the pixel density is probably lower given the size increase right?
so 30K pixels less
Not that it matters for a screen this small at a density this high
probably, it's 6.8 inch
4021248 pixels
3840x2160 is 8294400 pixels
so it's winnar in overall pixel count
for a screen 5x the size lol
i sit far back (i think)
lol
I like the brick like nature of these coolers
@nocturne olive does it have to be 24GB?
i can get a 20gb 7900xt for 500
Has to have CUDA
the 1K pc seems to be a dud
whaaat after all the good things i said about rocm
Well I don't have the tools for fixing
My system kinda needs to run consistently
I guess gotta wait for another 3090 or PC with a 3090 to show up again
How sad
i like when things include gloves
lol
"be careful dont break our nice thing"
10 points to gryffindor
"using this tool produces debris which might be sharp or otherwise bad for your skin. now that we gave you gloves it'll be your own fault if you don't use them."
i got gloves with a set of files. i've used them one time and that was because the aluminum filings left a dark mark on my hands.
i also got some with a tool to replace the gas piston of my chair. those were more because they supplied the smallest possible tools with the least possible surface finishing and your fingers would hurt without something between them and the tool.
but i had better tools with the same bit.
someone is selling a 3090 for 600 without heatsink 
the last included ones i had were included with a selving unit for like a garage
and it was definitely hella sharp
in the interlocking bits lol
maybe i jsut sell my own 3090 to superbox and get a 4070ti super
Well it is entirely up to you
It's not like 16 gigs of VRAM isn't enough to do simple AI stuff
24 gigs is really only needed once you start getting into serious models
i can get a 4070 ti super for 400
Well with that you'd be making 100€ profit from selling the 3090
i wouldnt call it profit, since i bought this 3090 for like 600 a year and a bit ago
Well you'd still end up with more money than you had before bying the 4070 Ti super
That's a more accurate description
Either way, it's up to you, for you the biggest changes would be
Gain 100€ and some extra gaming performance, lose a few GB of VRAM
The 4070Ti super is 20% faster
Well I was already gonna pay something like that for the 3090 that I was hoping to get earlier so like nothing really changes
without dlss the performance is pretty close in games tho
RasPi users of #programming, can i use a raspberry pi 3B as a wif <> ethernet bridge to connect non wifi enabled devices to wifi (i have no ethernet, i wouldn't need this if i did)
i think so ye
ye you can
limited to 100Mbps
Raspberry Pi 3B WiFi isn't the best in my experience anyway
so in practice it's gonna be way slower than 100 Mbps 







