tech-chat-1

[rough salmon]

same goes with wifi

[deep musk]

Depends on the distance

[spring quail]

the cable would have to be ridiculously long to effect latency, etc

[deep musk]

Not how it works

[spring quail]

it works or it doesn't

[deep musk]

(Ethernet maxes out at 100 meters)

[stark kiln]

Its gonna be high up on the shelf with the thing connecting to my pcs ethernet port

[rough salmon]

yh i figured haha, i thought the longer its out the more latency

[stark kiln]

well im at like an apartment so distance aint a problem lol

cause rn its ass it barely goes 10mbps. I belive its due to the fact its in a horrible posititon near tons of electronics

[spring quail]

Setting up a mesh network fixed my issues, I also live in an apartment. Doesn't help when there's fifty other hotspots nearby.

[deep musk]

Yeah, since you’re in an apartment Ethernet will be more stable and more reliable than wifi

Because a lot of other neighbours are also broadcasting their wifi

[rough salmon]

i live at a shared place with everyone on same wifi (7 ppl) im getting my own extender in my room

[flat cliff]

No wet but you know how they say west the EST band and don't just do this on carpet. Assume I like didn't and did and now ram dead, like now it can't be used in PC and no boot because dead stick. Just testing it won't fry the slot? No water of bridged connectors or anything.

[stark kiln]

got it, ill get a cable soon, do yk what ethernet cables are good

[frigid blade]

cat6 is enough

[deep musk]

Get a true cat6 or cat6e

[random junco]

as long as its cat6+ and has good reviews its all good
my cables are all from aliexpress, no issue

[spring quail]

Yeah I have one node next to the modem and another in the kitchen, fixed pretty much everything.

[stark kiln]

how expensive are they?

[spring quail]

They're fairly cheap

[random junco]

my alix ones were like £1.10 (1.50 usd) per cable

amazon ones will cost miles more though

[deep musk]

Not expensive at all

[rough salmon]

sometimes you get a free ethernet cable from your broadband provider, mine does

[random junco]

theyre often bad

[stark kiln]

theyre barely 1 and a half arizona bottles, im gonna get that bad boy asap

[rough salmon]

the one they gave me was a cat6e

at 2m

[random junco]

hmm fairs

[random junco]

oh.

[rough salmon]

yhhh

[random junco]

actually that wouldnt matter for me because my switch is in the same room as me

[rough salmon]

wait no

its 1m

[random junco]

my cable for my laptop is 3m

[random junco]

holy shit 🤣

[rough salmon]

yeah i asked for 1m thats why

[random junco]

so you could have asked for any length... yet you chose 1m

[rough salmon]

my laptop was literally touching the router

[random junco]

tbh id still ask for more

[rough salmon]

buddy, dont ask

how much m would u take

[random junco]

5

[rough salmon]

5m is good for 6e

[deep musk]

Often a cat5 specked cable

[deep musk]

Ethernet max length is always 100m, always, no matter the generation

[rough salmon]

a cable that only does 100mbps haha

[stark flicker]

Anyone knows if there are "bed" charging stations? Or I'll just pick a random charging station

[rough salmon]

stupid question, is there such thing as a ethernet to fibre, to have it longer

[deep musk]

Just get a charging station?

[stark kiln]

Id kill for 100 mbps

[rough salmon]

i get 100mb

[stark kiln]

I get 5mbps at usual

[stark flicker]

There is but that's super expensive because you need to convert it

[stark kiln]

sometimes it feels cute and goes to kbps

[stark flicker]

Yeah I guess but I thought that maybe there was something optimized for bed ergo wise, idk

[rough salmon]

i get 1mbps at day time

[deep musk]

Just use a signal booster every 100 meters

[rough salmon]

mine goes to 800kbps sometimes

[stark kiln]

Ethernet time

[deep musk]

Not sure about that, would probably just cost more cause it’s “more” niche

[rough salmon]

whats ur latency

[stark kiln]

Depends, but its bad

[rough salmon]

im at 11ms

sometimes when i do cloud gaming it goes to 9ms

[stark flicker]

Yeah I don't really care about price but I'll just take a charging station then

Are usb port rated for a lots of watts limited to like 5W when plugged to something "dumb" that only needs 5W?

I need to charge sextoys lol

[tulip sparrow]

Yeah testing shouldn't be a problem

[deep musk]

Fuck, you just reminded me to do that too (I always forget)

[brave brook]

real top is cat6/6e and next step is fiber 🙂

[stark kiln]

just rawdog it

[stark flicker]

So can I plug them on any charger without frying everything?

[deep musk]

Theoretically, just make sure its using the right cable (they often get a usb a to proprietary connector)

[stark flicker]

Yeah it's a proprietary magnetic cable

[rough salmon]

bruh i want linux so bad i dont have a usb stick yet

[frigid blade]

So much oversharing

[deep musk]

All my toys max out at 5 watts so you should be fine (from various brands)

[rough salmon]

son

[deep musk]

Not a dude

[rough salmon]

its a meme

[stark flicker]

But like if the charger is 40W or idk it will limits itself to 5?

[deep musk]

Don’t care, still not a dude

[deep musk]

The toy will very probably limit itself to 5watts

Since it has a battery management system

(Bms)

[nocturne oxide]

Should I build a pc now or wait for ddr6?

I can buy a mac mini while waiting

[rough salmon]

wait for ddr6, then resell that ddr6 once ai starts chomping on it all

oh wait, thats a good idea ima do that actually icl

[stark flicker]

OK yeah thanks

[deep musk]

Happy charging

[stark flicker]

[deep musk]

[brave brook]

you can't (they increased prices and removed stuff). Maybe once M5 mini launches, but won't cost $599 i suspect

[glad gulch]

Yeap

[rough salmon]

apparently steam machine and steam frame will be following the steam controller reservation system (am i safe since i was vac banned from black ops 2, 5 years ago?

[brave brook]

also DDR6 is most likely a 2028-2030 thing, not sooner than that

[deep musk]

Fellow toy user?

[glad gulch]

Who said you needed a USB stick

[glad gulch]

Nah, just tech things

[deep musk]

Sounds like a skill issue

[glad gulch]

make a partition on your hard drive and then just use that to install Linux

It's the cursed way but It Just Works™

[stark flicker]

Thanks
Those things are so expensive it's scawwy lol

[rough salmon]

im gd anyway, im eligeble for the reservation

for controller

[deep musk]

Ohh okii (sometimes toys are funky and not super standard

[nocturne oxide]

Yeah... I'll just get a pc when I can

[glad gulch]

Most of the time you don't want to shove too much voltage into devices

Amperage, they'll draw what they need

[rough salmon]

ya

just cuz they have 4 slots dont mean you can fill all 4 plugs with gaming pcs

[glad gulch]

It's why you can hook up a 2 amp charger to a Kindle but it still takes hours to charge because of the 800 mA draw limit

[glad gulch]

I mean you can, but you shouldn't

[rough salmon]

kaboom

[glad gulch]

Not quite

Depends on the total wattage to the wall

[rough salmon]

what would happen, eletricity would trip?

like if the extension lead was like 15amps

[glad gulch]

And really it depends on the weakest link in your circuit

If you're drawing 1000 watts through the ghetto $5 extension cord that cord is gonna be the first thing to fail

[deep musk]

(Depends on the max wattage of each unit and the standard voltage used in the contry)

[glad gulch]

Yup!

[deep musk]

110 volt max 1800watt peak but 1400watt sustained and for 220 volt it’s 2500 watts peak iirc

(Forgor 220’s max sustained watts)

[glad gulch]

Yesterday I managed to pop the breaker with a vacuum plugged in

Which answered my question, our dormitory IS in fact shared-wiring

[deep musk]

Oof not fun

[deep musk]

Please tell me you’re in a contry with 110 cause if it’s 220, there’s a lot of of people on one single curcuit

I think I’m hearing a kibby crying for help qwq am in an orchard so it la possible there’s one

Am gonna try to listen more closely and search for it

[rough cove]

Hello, can someone tell me what that subdial bottom config is? I’ve been driving myself crazy trying to figure it out

[brave brook]

my 2 room apartment is wired up with 3x 25A @ 250V 😄

[deep musk]

Daaaaaaaaaaammmnnnnnn

[glad gulch]

110, yes

[sour remnant]

Is anyone here knowledgeable about audio? I have a query that I feel I’m too stupid to solve myself but on the offhand is seems simple enough to fix 😅

[glad gulch]

More like 117, I poked my multimeter in and for some reason it was at exactly 117 AC volts

[glad gulch]

Which is?

[rough cove]

I sorta am depending on the question

[sour remnant]

#audio-tech

[glad gulch]

Not 115 or 120 as you might think, exactly 117

[azure topaz]

thats normal

[deep musk]

Ohh oki

[brave brook]

to be fair, 3x25A is the normal, 3x20A is the "low", and you can go above 3x25A if you have exceptionally high power usage (3x63A is the typical high energy wiring)

[deep musk]

Here it’s 120v 15amps

[rough cove]

Nvm figured it out (I’m kinda stupid sometimes)

[brave brook]

i mean wiring up to the apartment input. Sockets are obviously wired for 'only' 250V 16A

(or 3 phase 400V for induction cookers, but you don't have to use 3 phase for that, it just limits how many and at what power you can use at same time)

[deep musk]

Ohhh okiii i thought it was 220 25 in the wall socket Ks

[warm hinge]

What is a good 4K monitor

[brave brook]

it's 250V 16A. Still more than double of yours tho 😄

[deep musk]

Yes qwq

[warm hinge]

Even though I can get a 2K one but I want to use it to play my Xbox games on Windows especially Halo and Forza with the best quality

[empty quartz]

In canada, its 240v, most outlets are 120v. Our 240v is weird though

[warm hinge]

What do you guys think what is the best 4K monitor especially with HDR?

[deep musk]

It was just the village’s whiny kitty, forgor she was allowed to come in the orchard (gave her loads of pats, didn’t take a pic because her owners don’t want us to post pics online)

[robust trail]

Excellent

[deep musk]

Nicey

[warm hinge]

Any recommendation for a 4K monitor?

I just mentioned that right now

[cold skiff]

Loads of recommendations

[stark flicker]

4k oled 240Hz 32"

[cold skiff]

You need to think about your budget, the size you want, the panel type you want and head to Google

Oh and refresh rate how could I forget

And if you want oled you need to know if you want matte or shiny

[warm hinge]

Alright sounds like a good option

[stark flicker]

It's the best but not the cheapest

[warm hinge]

I see

[warm hinge]

I'm looking for one that's on a budget at least $200 to $500 and probably between 240hz, 60hz, or 180hz

If it's OLED I'm fine with either matte or shiny

[stark flicker]

You don't have the budget for 4k 240Hz nor oled

What's your gpu? @warm hinge

[warm hinge]

Mines is an NVIDIA GeForce RTX 4070 Super

[stark flicker]

Yeah well you can play in 4k with that but you won't achieve 240fps, nor will you have loads of fps in the big AAA games

[tall sandal]

Is this mobo good?

[warm hinge]

So I should try getting a 2K monitor since my GPU wouldn’t handle 4K very well

[stark flicker]

It's not 2k but 1440p, and no do what you want

I have a 4070S with a 4k screen, it's doable, I hate less than 4k

But with your budget you can get an oled 1440p and oled is the single most important thing in life and that's before s*x and drinking water

[warm hinge]

Alright then investing in 1440p would be better for me

[formal sonnet]

Nah, don't go with 1080p (2K) in 2026. Do what @stark flicker suggested - look for a 1440p display that's either OLED or 120Hz / 144Hz, or both, if your budget allows.

1440p is love; is life. ❤️ Both gaming and productivity are amazing on 1440p displays between 27" and 32" in size.

[warm hinge]

I would look into those displays that is a better recommendation

[stark flicker]

You too are horribly annoyed by people calling 1440p "2k"? @formal sonnet

[formal sonnet]

No, I'm annoyed at companies whose marketing departments are completely WRONG about how to market their products.

[random junco]

i have a 32" 1440p and i love it

[stark flicker]

I have a 1440p miniled 180Hz and everyday I'm telling myself "it's horrible I want 4k oled like my other computers" 🙁

But not everyone has the budget for 4k and oled 🙁

[formal sonnet]

TL;DR: 1440p is not 2K, cannot be 2K, and will never be 2K. It's more like 2.5K, but "#K" isn't really an official term that makes sense in a world where non-16:9 aspect ratios exist, since Ultrawide displays make the "#K" system useless as a descriptor.

[formal sonnet]

I just purchased a new monitor... you'll be unhappy to know it's not OLED, but it checks all the other boxes: 39.7" 2500R curved color calibrated IPS display, 120Hz, 5120x2160 (WUHD), DisplayPort/HDMI and Thunderbolt connectivity, and it supports appearing as 2x displays over Displayport as far as Windows is concerned.

[stark flicker]

Lol is that the Dell at 2k€?

I wanted it at some point for work

[formal sonnet]

Close; it's Lenovo's equivalent: the P40WD-40. I wanted the Dell, but they haven't been on sale nor available for a while on Dell.ca so... 🤷 They're all supposed to use the same IPS LCD panels (Dell's, Lenovo's, LG's, Samsung's, etc.) so it really boiled down to the control board features for me.
https://www.lenovo.com/ca/en/p/accessories-and-software/monitors/professional/64b4gar1us

[stark flicker]

I didn't know Lenovo made one

It really is great for work, the "black" ips panel does help with contrast a lot but absolutely destroys pixel response time... If it's the same for the lenovo

I'll check the prices, I wonder if it dropped a bit

But I was waiting for the equivalent in oled, also I found the 2k€ price tag insane

[formal sonnet]

Same thing I found - it'd be nice to have an OLED version of these monitors, but they seem to start at $2000 CAD which is way too much.

[stark flicker]

Yeah 1600€ for the lenovo and dell... When I see my LG 77G5 for 2200€ my brain doesn't accept it lol

How much did you pay for it?

The LG 45GX950A is at the same price lol in EU

[formal sonnet]

Lenovo claims the P40WD-40 is "worth" $3049 CAD (ha! yeah right - makeup whatever you want, Lenovo; it's worth closer to $2500 CAD at most) and "on sale" for $1469 CAD. I paid $1208 CAD before taxes and environmental fee. After tax + fee it was $1355 CAD.

[stark flicker]

Oh wow that's super cheap

Yeah for 840€ I would have bought it

[formal sonnet]

Yeah, like, it's the most I've ever spent on a monitor, and is double what my 5060 Ti cost (which is now a bit anemic for this display but whatever), but on the other hand... once you get on the business monitor train, it's REALLY hard to get off. It's kind of the same deal with getting on the OLED train.

[stark flicker]

LG 39GX950B

I think this one would be end game (until we have 5 or 8k lol)

And yeah if you work with it I can understand, some features are soooo neat to have

But... And that's dumb... I work 95% of the time with a black screen inside a terminal.... I want oled even for work

[formal sonnet]

Oh sick, that's the LG monitor that supports either 5K2K @ 165Hz or WFHD @ 330Hz natively. AKA productivity mode, or gaming mode. 😄

[stark flicker]

Yeah that's neat though wfhd... Erf, I don't know

Maybe it will be fine, most of my games on that computer are not "big" games, the beautiful ones are for the TV computer

[formal sonnet]

Yeah, I'd sooner the monitor do 3440x1440 @ say... 240Hz instead of WFHD @ 330Hz.

[stark flicker]

Yeah 🙁

[formal sonnet]

Forgot to reply to this last night, but yes, that's what I've seen everywhere. If it were only marked up a couple hundred, I'd be so tempted to have someone smuggle one into North America for me. But not at nearly double the MSRP; that's just extortion. 5070 Ti cards are around $1200 CAD on average; the pre-overclocked models are around $1400 CAD. Guarantee the special Frieren edition card would be something like $1700 CAD or perhaps even close to $2000 CAD if they sold it here.

[slim mauve]

Hey people

With the current memory market

Is 160€ for 1TB NVME decent?

There a deal from 180 and idk if I should consider it

[limpid dune]

at the current memory market, it's a good time to go outside and touch grass

[limpid dune]

I just picked up something like that and they included the wrong one in the box... I got the 100hz SDR version and they sent me the 240hz HDR one. Peak

[formal sonnet]

Depends entirely on the exact make and model of SSD and your use-case. Some aren't worth any amount of money for certain use-cases, or worth anything at all due to massive failure rates reported by other customers. Also, PCPartPicker is a great way to compare prices for certain PC components. https://nl.pcpartpicker.com/products/internal-hard-drive/#A=960000000000,1500000000000&sort=price&f=122030,122042,122060,122080,122110

[slim mauve]

:/

[slim mauve]

It's a LEXAR one

I've heard good about them

[random junco]

pfft

[formal sonnet]

Lexar is just a brand... they make many different models.

[slim mauve]

NM790 1T

[limpid dune]

Lexar isn't too bad, but yeah, that price is kinda on the edge, I'd only buy it if you really really immediately have a use for it

[slim mauve]

Oke

You recommend waiting for prices to drop (aka is that gonna happen soon enough?)

I'm planning on building a pc beginning of summer

[formal sonnet]

Reviews look good, and apparently it's fast despite being DRAM-less, but also... if this is intended to be a boot drive, I recommend getting a drive with DRAM cache.
https://nl.pcpartpicker.com/product/B8QKHx/lexar-nm790-1-tb-m2-2280-pcie-40-x4-nvme-solid-state-drive-lnm790x001t-rnnng

[slim mauve]

Like July or smth

[formal sonnet]

https://tenor.com/view/robin-hood-fortune-teller-future-prediction-gif-25704815

[slim mauve]

What's that?

Dram

[random junco]

cache

[formal sonnet]

Please read with a grain of salt, as all SSD manufacturers would of course prefer you purchase their more expensive SSDs. But also, tada!
https://phisonblog.com/dram-or-not-the-difference-between-dram-and-dram-less-ssds-and-why-it-matters/

[tidal geyser]

all i want is a panther lake

[frigid blade]

52c Panther 😩

[tidal geyser]

what

thats nova lake

[limpid dune]

I like how they went for 16 P-cores, 32 E-cores... and four more LP-E cores just to save a little power. On a HEDT chip. for some reason

[tidal geyser]

well

have you seen lpe cores?

they idle at nothing

and take up almost no space

[limpid dune]

oh yeah, but... this is a >300W processor, we don't care about heat. And it would be really nice if these cores had an L3

[tidal geyser]

at peak yea

but it aint a server cpu

most people are still going to have this cpu at idle for most of its life

also no need for l3

my lpe cores manage pretty decent gaming :3

[nova barn]

how many people rember this video

https://youtu.be/Gnhv3WFnmH8?si=LYFYoduYzuKuFVYF

[torpid sinew]

regarding the failed "LTT steamdeck vest", maybe try some version of working trousers instead? I have a pair with pockets large enough to house a entire iFixit Mako kit
It should also fit a GPD Win Max (2019), so techncially you already have something that dont suck

[empty quartz]

Just figured out a rock i found might be fossilized coral

[nova barn]

lol at how bad this is it is 10 feet or more from a 65 inch tv when what recomadn is 8 feet or 6 feet

'

[stray portal]

Basically doesnt matter

[formal sonnet]

🤷

[nova barn]

what about my qquestion

[tulip sparrow]

There are entire sites that have done the math on this already LOL

Example https://www.rtings.com/tv/reviews/by-size/size-to-distance-relationship

[nova barn]

i had people told me they lie and 10 feet for 65 inch is good enough when i been have headake

[tulip sparrow]

https://calculator.ff.de/viewing-distance-and-field-of-view/

[nova barn]

what about i try tell them that and they say it bies and what we have is good enough dont want to move the couhc or get a bigger scren

[formal sonnet]

We cannot assist with people in your life who refuse to listen or perform their own research to make educated decisions. It's like leading a horse to water - only the horse can decide if it actually wants to drink the water provided.

[nova barn]

i just find it funny

[tidal geyser]

https://tenor.com/view/war-never-changes-war-fallout-soldier-run-gif-16498934

[tidal geyser]

i would suggest you taking a eye examination

my head hurts if i do not have my glasses on and looking at something further away

[nova barn]

what do you mean

[tidal geyser]

eye strain

[shy arch]

Actually wise words

[limpid ivy]

Got that to, not fun but tolerable :3

[nova barn]

i am told it becuase i do game too much not becuase of tv

[true stirrup]

If you haven't been to an eye doctor in a while go get checked like sim mentioned
You might not even know that you might have sight issues
When I got my first glasses I was completely shocked

[nova barn]

i have been just some month ago

have glasiice but still get it

[frigid blade]

Sorry I just woke up when I sent that

lol

No time to think straight

[tidal geyser]

your eyesight can become worse over time i can already notice i need new ones as i am starting to have more and more headaches

[nova barn]

but it was just this year

[haughty compass]

Found this 1080p 60in sharp TV for $62 at the thrift store after discounts n stuff (50% off cus it's Sunday and I have a coupon). Is it worth picking up?

[grizzled frost]

that thing looks fucked

[limber nacelle]

i mean.... if it works, and it's all you can afford... and you need a TV.... yes?

[haughty compass]

The static is how sharp does it when no channel

[haughty compass]

Idk I just like big tv

But idk if worth

[limber nacelle]

i mean it's a 1080p 60" tv, the PPI on it should be absolutely fucked

so if you're gonna be sitting anywhere close to it well

prolly not great

[haughty compass]

Truw

Funny enough I believe my parents have this exact tv

Used to be in living room, now its in their room

[mellow sonnet]

majority of tv is still streamed in hd or fhd so that tv actually makes more sense resolution wise as opposed to a 4k where those who use streaming would benefit more from it

[serene pumice]

what? my 60" 1080p Plasma TV is fine

[mellow sonnet]

the ppi is fine for a tv

[limber nacelle]

i mean..... again it just depends on how far away you sit

if you can sit far enough away i'm sure it's gonna be fine

i use a 1080p projector to watch stuff at night and it's fine enough for me 🤷

[nova barn]

How close I was planning on sit up a 1080p projector with a 100 inch screen

[left dove]

yeah my bedtime tv is a 24" 1080p display at the bedside

1080p at 100" will look like ass

[nova barn]

It what I can afford

4k projector even used are 1k

[bold olive]

22ppi is wild

[left dove]

so dont do a 100" screen

because the lumens required to make that palpable is not something available in 1080p projectors

[nova barn]

It is in eBay used one

[left dove]

gross, ew.

[nova barn]

People in the home theater group would disagree

[left dove]

mmhm

sure
and im ghandi

[nova barn]

You much think money grow on tree

[left dove]

I mean it does if you get a job

[limber nacelle]

i mean... it's fine enough for me for now cause i only watch stuff before i go to bed

[nova barn]

Hey I have one blame low wage

[limber nacelle]

i think my main grip with the one i have is that LG cheaped tf out on it

to my understanding, they used plastic for the lens which heats up and changes the shape of the lens which makes it go out of focus after you have it on for a while, so you have to keep readjusting the focus

which is annoying af

[clear canyon]

Try to find a better job, what kind of job you have currently?

[limber nacelle]

im thinking i might as well go for something reasonably high end for my next one, like a vidda c2 pro after i sell this one off

[nova barn]

It a job at a small store and they need me and told me to Ignore any one who disagree with that and it pay double minimum wage in my area

The job is not the problem I have lot of money in bank it just that family won't let me spend it

[clear canyon]

They need you? They also said to ignore anyone that disagrees? That does not sound normal

[clear canyon]

If ur an adult, they cant control what and when you spend your money

[limber nacelle]

not necessarily true

if you live under their household, then it's hard to kinda argue about that sorta thing 🤷

[nova barn]

I most mean I love the Job and too stressed out to look for a different job and the job market is terrible now

[limber nacelle]

and it's also not always realistic to "just move out"

[clear canyon]

They cant legally do it

[limber nacelle]

it's not about legal, it could just be.... pressure

[clear canyon]

Yeah ik, im saying they cant control it like if you was a minor

[limber nacelle]

as in "don't waste your money like that", etc

[clear canyon]

Yeah

[limber nacelle]

even if you could legally speaking spend your money however you like, it sucks to draw the ire of your parents while you live under their household

[hexed tapir]

I always knew you were old but I never knew you were ancient

[nova barn]

I am confused I was told by other people that sit too far doint cause eyes strain only too close

[deep musk]

who's that fucking dumbass?

[limber nacelle]

i feel like this is kinda context dependant

[deep musk]

the tittle was "how to choose the best general gpu in 2025"

[limber nacelle]

i think nobody would disagree that it's better to get a 16gb variant but depending on price and need..... 8gb might make more sense for some people?

[limber nacelle]

aight bet easy, https://www.amazon.ca/PNY-NVIDIA-Blackwell-Workstation-GDDR7/dp/B0FKD5GF8W?source=ps-sl-shoppingads-lpcontext&psc=1&smid=AECIPMEQNVKO2

the best gpu :)

[rose stream]

For cost and depending on what settings most people play at, it might make sense.

[deep musk]

actually, yeah, with such a shit title might as well

[limber nacelle]

yes hence very context dependant

[limber nacelle]

should be "how to choose the best gpu for yourself in 2025"

cause best is always context dependant :)

[rose stream]

Though personally, I do sometimes wish we had 10 gig models, even if there will be a few compromises at the architectural level for them.

[limber nacelle]

i think 12gb would be perfectly fine for a 60ti class gpu

16gb for 70, 20gb for 80

but we're not allowed to have nice things in 2026 😔

[tidal geyser]

i just use igpu :3

[deep musk]

just watched the "how much vram should you get" and was saying "unless you play at crazy resolutions like 4k you should get around 6-8-10 gb of vram"

[tidal geyser]

xd even at 1080p 10GB can struggle

[deep musk]

I gpu with 2tb of ram, you can allocate a lot of ram to that little ipgu

[limber nacelle]

as much as you can reasonably afford :)

[tidal geyser]

its unifed

anything not used by the cpu can be given to it

[deep musk]

yes, but he didn't mention that

[tidal geyser]

aka i can have a 31GB igpu

well not can but have

[deep musk]

oh? mac user?

[limber nacelle]

i mean tbh for my needs ill take all the vram i can get, provided the gpu has dedicated RT cores outputs display

[tidal geyser]

what

[mellow sonnet]

sounds about right

[deep musk]

you said the memory was unified

[mellow sonnet]

6gb can be fine

[tidal geyser]

yes thats how a igpu works

[limber nacelle]

sure, if you play games from way back

[tidal geyser]

its shared between the cpu and igpu

[limber nacelle]

or only comp games 🤷

[mellow sonnet]

unified usually means the ram is on the soc

[tidal geyser]

windows is just a bit stinky

[tidal geyser]

so what lunarlake

anyways no not what it means

its just that its shared

[deep musk]

not really, you have to set a max amount allocated, typical igpu doesn't use unified ram

[tidal geyser]

what

[mellow sonnet]

green smartass vs green smartass

[tidal geyser]

its a dumb windows limitation

a igpu does not have dedicated vram

its just ram force allocated for the igpu

[deep musk]

an igpu can't use more ram than what is allowed in the bios, x86 cpu doesn't support dynamic ram allocation between cpu and igpu

[tidal geyser]

what

[vital marlin]

What’s the best tool to scan individual files for virus/malware?

[deep musk]

yes

[mellow sonnet]

windows defender

[tidal geyser]

tf you mean

[mellow sonnet]

trust the white role

[tidal geyser]

virustotal

[deep musk]

https://www.reddit.com/r/hardware/comments/14tay2f/can_mseries_unified_memory_be_replicated_on_x86/ I was confuzed x86 can use ram like this but not on common consumer hardware

[rose stream]

And then there's someone like me with 20 gigs of vram and I hardly use that much of it anyway. Lol.

[tidal geyser]

can x64 do it?

[deep musk]

read the comments

[mellow sonnet]

i got 16gb and i rarely use it over 12gb

[tidal geyser]

did you just forget about consoles?

[deep musk]

64 bit and x86 are two different things, x86 is the architecture and 64 bit is (very simplified explaination) how high the cpu can count

[limber nacelle]

7900xt?

i have the same amount of vram :>

[rose stream]

Yep.

[deep musk]

you're not using a console as your pc

[tidal geyser]

x64 is the architecture

[deep musk]

no

[tidal geyser]

it is

[mellow sonnet]

x86 is the architecture

[tidal geyser]

for 32 bit

[mellow sonnet]

no even 64bit is x86

just called x86-64

[rose stream]

Thing is, the core itself is more than plenty, while the vram amount is probably overkill at this point.

[deep musk]

no

https://en.wikipedia.org/wiki/X86

[mellow sonnet]

we got an expert in chat

[rose stream]

Instruction set architecture, specifically.

[tidal geyser]

what does that have to do with x64 not having unified ram on desktop

[limber nacelle]

i just want more for substance painter because that shit eats up vram like nothing else

[deep musk]

https://en.wikipedia.org/wiki/X86

[deep musk]

64 bit has nothing to do with this

[limber nacelle]

tbh my gpu is probably the slowest out of everyone here in chat atm but it doesn't really matter

i just want more vram rahhhhh

[deep musk]

I have a 6750xt

[limber nacelle]

still faster

[mellow sonnet]

i own the sleeper 4080 super

no amd gpu in my house

[rose stream]

I still need to get into 3d at some point, especially after someone mentioned to me to 3d print a design I came up with that... Admittedly, I really would like to some days.

[tidal geyser]

eh fine

but what does that have to do with the original point

[limber nacelle]

i'd really recommend it, especially sculpting

[mellow sonnet]

i have hands on a 960 2gb which is a cool unit

[rose stream]

Which one do you have again?

[tidal geyser]

you said x86 cant have unified memory yet consoles are a thing

[limber nacelle]

4000 ada SFF

[rose stream]

Aaahhh, makes sense then.

[limber nacelle]

it's a relatively slow card cause it's 75w 😔

but it's fast enough for me

[deep musk]

damn, I also have a mobile 4060, desktop 7600, 1080, 980 fe, a770 engineering sample, a bunch of ati hd 2400 and 3400, quadro 4000, k2000, quadro 4800, etc I have a bunch of slower cards 😅

[rose stream]

x86 is just an isa though. You can do unified with literally any isa if you want it enough.

[tidal geyser]

.

[deep musk]

I forgot about consoles, but your pc doesn't have it

[tidal geyser]

why

you said x86 does not support it

[mellow sonnet]

the soc has to be designed in a way

[tidal geyser]

in what way

[mellow sonnet]

a way

[deep musk]

because it's a super niche thing for pcs, maybe some very expensive and rare laptops do

[mellow sonnet]

my way by frank sinatra

[rose stream]

Again though, x86 the ISA itself is not limited to it.

[tidal geyser]

and like i said windows is just a bit more annoying about it yet it still does it

[rose stream]

In that case, it's a chip design issue, not an instruction set one.

[deep musk]

yeah, I confuzed myself with something else

[tidal geyser]

its a software issue

[limber nacelle]

damn even my shitty 2080tis beat it 😔

[rose stream]

That too.

[limber nacelle]

absolutely clobbers it

[deep musk]

it's using swap to have "more ram" for the igpu

very much not the same thing

[tidal geyser]

what

[mellow sonnet]

2080ti > 1080ti

[deep musk]

(linux subreddit but it's the same on windows) https://www.reddit.com/r/linuxquestions/comments/19c1ggx/a_swap_file_is_a_type_of_swap_memory_for_computers/#:~:text=Well a swap file is,cat /proc/swaps

[tidal geyser]

so the igpu that does not have vram but uses ram as its vram is not the same when swap is being used aka stuff is put into ram

you are thinking of the wrong kinda swap

[deep musk]

up votes on a post doesn't matter, it's the gpd damn answers that matters

[tidal geyser]

when your gpu fills up it spills over to ram

[tidal geyser]

thats about ram swap

[mellow sonnet]

just dont use igpu

imagine the frames with your brain

[deep musk]

I'm too tired to continue to try to explain this

[tidal geyser]

there is no explaining you are you posting random reddit posts

tho i also confused x64 for amd64

[mellow sonnet]

green roles arguing

https://tenor.com/view/gem-finding-gem-gem-alert-jasswa-looking-for-gem-gif-4573895384736430461

[tidal geyser]

amd64 is the actual architecture

[deep musk]

no, I'm trying to explain

[mellow sonnet]

x86-64 is intels 64bit

[tidal geyser]

explain what

[tidal geyser]

amd made 64 bit and intel licenses it

[mellow sonnet]

intel sells amd x86 and amd sells intel a 64bit version of it

[deep musk]

that your pc doesn't have unified memory

[tidal geyser]

go down the page

[mellow sonnet]

it means that amd cant live without intel and intel cant without amd

[deep musk]

did you like read the litteral first line??

[tidal geyser]

ok lets go to the most simple level, what is unified memory

[empty quartz]

:3

[plain cypress]

tech chat 1 showing they have less braincells than an orange cat

[deep musk]

memory that can be dynamically shared between the cpu and gpu

[deep musk]

yes, really, and I'm loosing even more rn

[mellow sonnet]

whats under the hood

[deep musk]

nicey

[empty quartz]

Hey :( i just got here.

[plain cypress]

can't lose what you never had

[empty quartz]

M1 max, 32gb ram, 2tb ssd

[deep musk]

true, as a puppygirl I can't even argue against that

[plain cypress]

https://klipy.com/gifs/morning-good-99--k01KRA4VA3MRE16V4E0TYFQKK0F

[mellow sonnet]

does mine beat it

[empty quartz]

In cpu yes. Gpu no

[mellow sonnet]

truth?

[tidal geyser]

what makes this a hardware problem tho

[empty quartz]

Have you beat in gpu, screen, and cooling @mellow sonnet

[deep musk]

your cpu and gpu memory controller

[plain cypress]

in single core yes, in multi core no...probably

[empty quartz]

No

[tidal geyser]

they are the same on a igpu

[mellow sonnet]

well fuck

[deep musk]

not really

[tidal geyser]

in what way

[empty quartz]

I have the 24 core, should’ve clarified. But it still beats the m5

[plain cypress]

oh damn m1 max was only 10 cores
how the times have changed

[mellow sonnet]

https://tenor.com/view/epic-spongebob-epic-spongebob-speedrun-any-percent-geile-klöten-when-imposter-is-delicious-gif-19750086

30% better in cpu though

[empty quartz]

Remember when mac gaming was a good joke.

Nowadays macs can actually game decently

[mellow sonnet]

now macs are literally gaming

[deep musk]

your cpu and gpu's controllers don't work the same way, at all, I don't know the exact details since I'm not an engineer

[mellow sonnet]

back then macs were just shiny office laptops for double the price

the airs shipped with the worst mobile intels

[empty quartz]

Exception being imacs/imac pro, and the mac pro

[tidal geyser]

but its the same one, you can just look at a dieshot

[mellow sonnet]

they should bring back the imac pro

[empty quartz]

Yeah

With proper cooling too

[mellow sonnet]

it would absolutely rip with m6

[empty quartz]

Im glad the imac pro didnt last long back in 2017

It used power like no tomorrow

[mellow sonnet]

mac pro died, imac pro revival

[deep musk]

even if it was the same one, it's not made for it, only very specific cpu have that, apple's cpus and very expensive server cpus are rumored to get those maybe one day

[tidal geyser]

so why do 500$ consoles have it?

[empty quartz]

Funny thing, the mac pro was still on m2 before it got canned

[mellow sonnet]

you dont say

[plain cypress]

what are we arguing about now @tidal geyser

[mellow sonnet]

m2 ultra

[tidal geyser]

unified memory

[deep musk]

for gaming performances

[plain cypress]

even the M2 version sold basically no units

[plain cypress]

have y'all discussed strix halo yet

or like the steam deck

[deep musk]

they keep saying their non apple pc has unified memory

[tidal geyser]

or like any igpu ever made

[empty quartz]

It doesnt

[plain cypress]

technically correct is the best kind of correct

[deep musk]

I know qwq, that's what I've trying to tell them

[plain cypress]

although I have no idea what pc specs they have

[tidal geyser]

igpu

[deep musk]

no normal consumer pc has unified memory

[empty quartz]

Unless they’re running a bc-250 they dont

[hexed tapir]

https://www.ign.com/articles/sony-announces-price-rises-for-ps5-ps5-pro-and-playstation-portal-blames-continued-pressures-in-the-global-economic-landscape wdym the PS5 pro is $900

[tidal geyser]

the system ram is shared with the igpu, it does not have its own vram

[deep musk]

it's still not unified memory

https://tenor.com/view/riro-riro-ron-vtuber-anime-sad-gif-11862576451543336106

[plain cypress]

I would argue that desktop disaggregated designs do not embrace the idea of unified memory as we've come to know it

I would argue that, generally speaking, most mobile devices without a dGPU do have unified memory as we've come to know it

look at all these big iGPU chips AMD and Intel are churning out for laptops and handhelds

[tidal geyser]

its not a architectural problem

[deep musk]

you're not helping, it's not proper unified memory qwq

[tidal geyser]

its just windows not being good at being dynamic

[plain cypress]

okay what is the definition of unified memory

[empty quartz]

Only non apple consumer device i can think of with unified memory is the steam deck

[deep musk]

your cpu needs the physical memory controller

[tidal geyser]

no shit

[plain cypress]

cpus have memory controllers?

[mellow sonnet]

yeah

[empty quartz]

Memory that can be dynamically allocated with the gpu

[deep musk]

(some server soc are rumored to have some soon)

[deep musk]

this

[plain cypress]

is the issue that something like strix halo needs to have a preset iGPU memory limit assigned by BIOS when it boots

[tidal geyser]

show me what you mean

[deep musk]

not unified memory

[tidal geyser]

but when it runs out it just uses ram like how it already is

that shit is more so for compatibility

[empty quartz]

thats not unified memory

[tidal geyser]

shared memory between the cpu and igpu is not unified?

[deep musk]

just check =ed onl;ine, it was about epyc cpu and it turned out to be a false rumor

[plain cypress]

really nobody will have unified memory to the degree of apple since nobody else is vertically integrated
Game consoles come close but not perfect

[deep musk]

no qwq

[empty quartz]

Is it dynamic, and can change on the fly if you need more without having to restart?

[tidal geyser]

thats how its done now

and has been forever

[empty quartz]

You cant give your igpu 30gb of ram when you have 32

[deep musk]

and my iq levels are now negative

[tidal geyser]

so what happens when a igpu runs out of vram

well ram that is set off only for it

[deep musk]

the max is usually 4gb (for 128gb+ desktop ram setups)

[deep musk]

goes in the swap (ssd)

[tidal geyser]

there is not just one swap

also thats pagefile on windows

[plain cypress]

if we had a tier list of unified memory
S tier = apple
A tier = strix halo, PS5
B tier = steam deck cause fuck it why not
C tier = systems with a single pool of fast LPDDR5X memory
D tier = Dan Besser
F tier = traditional pc design

[lean heron]

@plain cypress it is

[deep musk]

doesn't make a difference

[lean heron]

some info is hard to get form just searching

passive aggressive name pmo

[tidal geyser]

it will first use ram then ssd

[deep musk]

yes, your computer has one pagefile, it can grow and shrink but it's used when your ram (or igpu "vram") overflows

[tidal geyser]

because why the hell would not not use available ram

[plain cypress]

asking questions into discord is the modern equivalent of ask jeeves
you're too stupid or lazy to google something yourself so you make other people do the googling for you so that you can feel some kind of validation when they regurgitate the top search result

[deep musk]

because the bios is not letting it do that

[tidal geyser]

can you just open up taskmanager and show your gpu tab

[plain cypress]

unfortunately, this is too much text to fit into a display name

[tidal geyser]

or even look at benchmarks of igpus spilling over

performance is about the same or only a tiny bit reduced

[deep musk]

have fun (I don't have swap, cause I'm too cool for that)

[empty quartz]

apple uses specialized memory chips, afaik really only apple actually uses them

[tidal geyser]

do you not even have zram?

[tidal geyser]

what is so special about them

[deep musk]

how they work

[empty quartz]

128 bit per chip

[tidal geyser]

and that is

[plain cypress]

the real special sauce is apple firmware and software

[empty quartz]

well the max uses 128 bit chips

[plain cypress]

the extra wide bus on the memory still goes out to lpddr5x

[tidal geyser]

?
Strix Halo tackles that with a 256-bit LPDDR5X-8000 setup combined with 32 MB of memory side cache. The latter is often referred to as Infinity Cache, or MALL (Memory Attached Last Level). I’ll refer to it as Infinity Cache for brevity.

[unreal bramble]

Who all here got the new steam controller?

[tidal geyser]

one person said soc but thats how lunar lake works aka what i got anyways

[empty quartz]

whats the bandwidth though

[plain cypress]

i'll have it on wednesday

[unreal bramble]

I get mine on Tuesday

[plain cypress]

i ordered at 10:01am and it is taking 9 days for delivery
fuck you gabe

[tidal geyser]

256 GB/s

[plain cypress]

pretty sure they packed and shipped orders in reverse chronological order timing

[unreal bramble]

I ordered mine at 11:58 am CST

[tidal geyser]

tho that does not do anything for unified memory

[empty quartz]

m1 max is 400GB/s

[tidal geyser]

ok but how is that relevant

[empty quartz]

its far faster

[tidal geyser]

ok?

[steel glade]

my buddy looking into upgrading his gpus. Hes looking between the 5070ti or the 9070 xt

[tidal geyser]

faster does not make it unified memory

[unreal bramble]

I’d get the 5070ti personally

[deep musk]

but having an igpu doesn't make it unified memory eihter

[empty quartz]

wider bus width per chip, and dynamic allocation are the 2 main things that seem to stand out for apples unified memory

[tidal geyser]

then what does

[plain cypress]

again, apple's full vertical integration is doing the heavy lifting

[tidal geyser]

dynamic allocation is software side

[deep musk]

having the correct hardware and software does

[plain cypress]

special microcode, firmware, software

only apple can excert such control over the user experience

[unreal bramble]

Apple is one of the best companies when it comes to hardware

[tidal geyser]

you have yet to explain what hardware is needed

[empty quartz]

@tidal geyser @deep musk its on dye memory, kinda like hbm in a way

[unreal bramble]

Even though it’s expensive lol

[deep musk]

the proper memory controller, I don;t have a part number to give you cause it's not how it works

[plain cypress]

its not on die what the fuck

since when is in package = on die

[deep musk]

I've told you this 3 times already

[deep musk]

kinda~ish

[unreal bramble]

This is such a heated debate

[deep musk]

but not really

[empty quartz]

thats what i meant, my brain is kinda burnt out from work

[plain cypress]

too little too late

[empty quartz]

@tidal geyser

doesnt go through the cpu at all

according to this diagram

[plain cypress]

neither does memory in my desktop

[tidal geyser]

thats how a cpu works

[unreal bramble]

Same lol

[tidal geyser]

it does not show the IMC either

the IMC is in the fabric

[deep musk]

the cache being accessible to both the cpu and gpu also helps

[grizzled frost]

what's with the three unlabeled blocks there?

[plain cypress]

3 stooges

[grizzled frost]

The 3 stooges are the secret to the apple performance

[deep musk]

probably apple's usually bs graph thingy were they show random stuff

[plain cypress]

unbeatable competive stooge adbantage

[empty quartz]

@tidal geyser seems that apples unified memory removes the separate pools entirely, which in turn removes the delay of transmitting information between the gpu and cpu

[tidal geyser]

nobody gave enough of a shit to actually click on the first link about unified memory

[deep musk]

(also the cache being seperated)

[deep musk]

so now you trust wikipedia?

[empty quartz]

not quite.

[tidal geyser]

well i truesed it before

[plain cypress]

since 4th grade

[empty quartz]

still separate pools for most systems @tidal geyser

[grizzled frost]

Is "Cache" as labeled here just the L3 cache?

[tidal geyser]

thats a software thing tho

[empty quartz]

its hardware

[tidal geyser]

how

[deep musk]

before I argue with you that wikipedia was true and then you simply trust it just like that?

[deep musk]

not what I meant

[tidal geyser]

what

[empty quartz]

apples design is quite different from most pc's

[deep musk]

no, if the physical electrical traces don't exist, it's not software

[tidal geyser]

tf do you mean

the igpu is connected to the IMC just like the cpu

[deep musk]

here\

[tidal geyser]

your cpu is x64

[deep musk]

yes but not the right way

[plain cypress]

How does Lego not have a rendering of their own box

[tidal geyser]

what is the right way

[deep musk]

x64 is not a fucking thing

[tidal geyser]

the wiki page you sent says it is?

[grizzled frost]

x86-64 is often truncated as x64

[tidal geyser]

x64 refers to the 64 bit extention of x86

[deep musk]

x86-64

[tidal geyser]

yes

[deep musk]

well that's almost the dumbest thing ever

[plain cypress]

x64 is what kids would call "slang" for x86_64 aka AMD64

[deep musk]

fuck me for using the right wording I suppose

[tidal geyser]

on windows if you even download anything you will never see x86_64 but will see a lot of x64 or even AMD64

[plain cypress]

as a linux user you should be ashamed
i've seen distro download pages labeled with x86 and x64 options

[deep musk]

and not using "slang" somewhere that using the correct terms is important

[plain cypress]

or even better, i386 options

[deep musk]

never seen 32 bit labelled as x86

[deep musk]

that's normal x64 isn't, amd64 is tolerable

[tidal geyser]

anyways again how is this relevant to unified memory

[grizzled frost]

Sorry that the industry moved to a shorter term to refer to the same thing

[deep musk]

idk, you brought that up

[deep musk]

if the prefix "x" is used for architecture, you can't simply use it for a variant of an architecture

[grizzled frost]

Take it up with the boring mfers in the industry man

Neither myself nor Sim coined the truncated term

[tidal geyser]

even in linux discords using x86 rather than x64 is halfway shunned

[plain cypress]

lego doesn't want you to touch grass lmao

[grizzled frost]

"No capes grass!" -edna mode

[deep musk]

ohhhh this is bringing me some goooood memories

[tidal geyser]

i do think its quite funny how intel got around having to do the dumbass shit amd has to on windows

just list the igpu as a 16GB gpu for games but do not lock it to only the igpu

but you can give the npu dedicated ram for whatever reason

[empty quartz]

@tidal geyser from the looks of things, it seems that even with an igpu, the hardware still has to split the memory pool in 2. im still researching though.

[tidal geyser]

why can you change it in software

[deep musk]

because you need to be able to do it in hardware, you need to be able to switch the lanes really fast

[tidal geyser]

what

switch lanes tf you on about

[deep musk]

the electrical lanes to direct the data at the right place

between the gpu/cpu

[deep musk]

I'm on the electrical engineering side of things

[tidal geyser]

thats the IMCs job

[deep musk]

yeah, that's what I'm talking about

[tidal geyser]

the ram data lanes go into the IMC and the IMCs job is well its the memory controller it handels all the subroutines and data managment

and from the IMC there goes lanes to igpu and cpu

when something gets loaded into the igpus ram it reserves space in ram and the IMC manages that from what i understand

the thing you are talking about with amd having a forced minimum of ram allocated to the igpu is due to some legacy windows stuff

[steel umbra]

so lego guides you to not plant booby traps on the ground? sad :(

[tidal geyser]

from what i have read the main reason igpus right now have to at least have some amount of ram set off is due to WDDM being a old shite and expecting there to be vram

jesus this was one hell of a thing for a marketing term

[mossy totem]

I bought 2 pairs of AR/XR glasses and I have never experienced such an enormous difference in quality. I got the RayNeo Air 4 pro's, and Virtue Luma Pros.

The Luma's are better in every way conceivable. I genuinely feel dumbfounded that there is only a $100 price difference. It's like comparing Google cardboard to a quest 3

The Luma Pro's are:

More comfortable, more stylish, have a better UI, better button layout, better audio quality, less sound leakage, a better connector design, a significantly better carrying case, better cable, bigger displays, clearer displays, brighter displays, less light leakage, better ambient awareness, less reflections, monumentally better edge clarity, external lighting, single tap shade adjustment from mostly see-through all the way to pitch black, they can play sound with the displays off, head tracking, a full 3D viewer companion app with content for free, support for more video formats, automatic 2D to 3D spatial conversion, spatial audio, IOD adjustment through software, and they have myopia adjustment for people who need reading glasses, per eye

The rayneos were the ones I got first and I didn't realize how terrible they were before the Luma's came in.

They have a weird display distance which feels like you're cross eyed when you use them. They also cut off the top and bottom of the displays and the edges are super blurry/fringed. Their UI is terrible too. It's like a one button cycle through OSD, and if you change some settings, the glasses fully restart without warning or confirmation of your selection, so if you wanna cycle through, you have to restart them multiple times. They are also really uncomfortable and hurt my nose and ears. Not to mention they stick out like a full 2 inches from your face

[robust trail]

Okay. Cool story?

[tidal geyser]

this is going to be a hard tech to sell unless they can fix the major issue that this has faced for so long

hello mark

[robust trail]

All AR HMDs are a pain in the as.

Birdbath optics are thiccc and they suck

[tidal geyser]

i think when they reach the point of like my glasses its a no brainer day to day

[tidal geyser]

btw me done linux things, so me now ask you about linux things for geforce

have you seen pixels / valve contractors "vram fix"

[robust trail]

I haven’t

[tidal geyser]

it just moves stuff that is in the background into ram instead of being stuck in vram taking up space

from what i understand windows already does this

[robust trail]

Yes, the driver does this natively from what I understand

[tidal geyser]

while on linux it does not

or well to the right degree

here is also the vram thing
https://pixelcluster.github.io/VRAM-Mgmt-fixed/

[robust trail]

Not sure, I’ll have to peek into it when I have time

[tidal geyser]

cachyos now offers a special vram management button for this

but it needs to be supported by the driver and pixel got it into amd and intel driver but not nouevo or however you spell it even

would be nice if one day it suddenly works with the more closed driver wink wink

eh i should go to sleep before it hits 4am and get a nice 3hrs of sleep

[robust trail]

lol

I let codex gut windows

seems to run with less trash

[mossy totem]

What issue specifically? I'm just curious

[mossy totem]

Just sharing my experience for the other half a dozen people interested in XR glasses lol

The rayneos suck horribly. I would never recommend them. The Luma's I could actually see myself using regularly, but I didn't buy either with plans to leep

[tidal geyser]

well you see im a bit more than just using linux i am a mod in the cachy discord and do a lot of other stuff

try to do this on windows

[mossy totem]

I have been meaning to swap to Linux. I will be doing it when I eventually upgrade my PC

Cachy OS does seem ideal. I run a lot of local AI

Does cachy support Wayland?

[tidal geyser]

too heavy so too much strain on nose, even the store page for one of the glasses only said its best to use em for 2-4hrs

[tidal geyser]

thats a DE thing

[tidal geyser]

cachyos would be best to run ai on due to optimized packages

[mossy totem]

Oh yeah, I'd never use them long term. I bought them for a very niche usage of which they are great for, but I am not keeping either pair

[tidal geyser]

and kernel and so on

[mossy totem]

Yeah, I have heard a lot of people saying that, especially for LLM offloading and streaming

[mossy totem]

DE? Sorry, I am very inexperienced with Linux and it's terminology. I do plan to swap, I just don't have the means at this moment

[tidal geyser]

Desktop Enviroment, aka kde, gnome, xfce4 and so on

kde and gnome are wayland out of the box and wayland will even deprecate x11 next year

you can also choose to just not install a DE for the lowest overhead

[mossy totem]

I'm sorry, I'm not 100% understanding what you are saying 😅

Does that mean I can't use Wayland? That's a pretty big requirement for me due to my monitor

Again, I am unfortunately very under educated in terms of Linux. I have tried swapping 3 times, but all 3 times I had horrible issues. That was many years ago though

[tidal geyser]

what distro

btw DE is the gui

[mossy totem]

I believe one was Mint. One was Ubuntu. And I don't remember the other. It's been many years

[mint umbra]

Think I will try bazzite next time I try Linux

[tidal geyser]

mint is x11, ubuntu im pretty sure is wayland and I don't remember I don't remember

[tidal geyser]

did you not use cachy for like 3 or 4 months now

or eh thats a bit long

[mint umbra]

Well after it got cooked I wiped the drive

So I'm just using windows rn

[tidal geyser]

sad

[mint umbra]

That was almost a month ago now

[mossy totem]

I have had two catastrophic Linux install issues that took out to a ton of my data alongside them, and then the third time I tried it locked my windows drive in a dual permissions BS loop. So I've been waiting to get a new PC so I don't dual boot, cause all its done is cause severe issues for me

[mint umbra]

Feel like Im not tech savvy enough to figure out how to fix issues I come across for catchy

Heard bazzite is more straightforward and harder to mess up

[mossy totem]

Ah ok, so is cachy non GUI normally?

[tidal geyser]

eh yea if its just gaming then bazzite

[mint umbra]

I do mostly gaming

[tidal geyser]

backup yo data

[tidal geyser]

no by default the first thing it selects is kde plasma

[mossy totem]

I do mostly AI. Running image generation trainings, inferencing LLM's that sort of thing. Cachy is what I have heard a lot about lately

[mint umbra]

Heard if you wait to long for updates and update on catchy it can cook ur shit lol

[tidal geyser]

no

i had a pc off for half a year and it updated just fine after resetting keyring

[mossy totem]

My issue right now that's preventing me from trying Linux is just that I don't have a drive for it

[fallow yacht]

@rigid fjord how's it in the uk rn?

[mossy totem]

If I had an easy way to install another high speed M.2 drive, I'd try Linux right now honestly

[tidal geyser]

its 3am for him

[tidal geyser]

got a free pcie slot?

[fallow yacht]

i know him he'll reply hours later lol

[tidal geyser]

true

i am so going to regert being awake this late

[mossy totem]

I do not unfortunately. Both of mine are in use by GPU's

And my second one for my second GPU is only 1x 3.0 😭

I have my main drive, game drive and AI drive. And that uses all 3 of my M.2's on my mobi

[tidal geyser]

if any of the drives has some spare space can just manually use that for the linux OS https://wiki.cachyos.org/installation/installation_on_root/

[south arrow]

How is sim?

[tidal geyser]

its 4am

[south arrow]

I send my regards

[tidal geyser]

i spend too much time problem solving

[mossy totem]

After the mass data destruction I have had before, I am never installing on a shared drive again 😅

I'd rather wait until I can do it properly than ruin a drive for the 4th time. I've already put off Linux for years cause of the issues I have had

[tidal geyser]

well its backed up

its annoying to redownload the ai model or games or whatever you would choose but no data is lost as it should all be backed up

remember your ssd could just die one day

[mossy totem]

That is true, I do know that. I've had it happen before and been able to recovery most data thankfully

I do technically have a 512GB SSD I could use, but it's Sata so that's not a great way to test it lol

[tidal geyser]

eh its not windows

anything SSD linux is going to be smooth

always funny when someone buys some 400$ pcie 5 10000GB read and write ssd but do not notice a differecne from their i dono pcie 3 ssd

[tidal geyser]

windows too tbh unless its complete and utter crap

[tidal geyser]

use ext4 for it, btrfs is really really really nice for small snapshots but ext4 is a lot faster, even if it does not really matter that much.

[mossy totem]

The reason SATA wouldn't really work is cause I wanna live stream weights, and a slow SSD isn't gonna do well compared to a 4GB/s NVMe haha

Jesus. I just realized how tiny 512GB is

Just one LLM I would use is like 100GB

[sinful cobalt]

https://tenor.com/view/drunken-goddess-reflux-femme-soule-famu-suru-pukey-goddess-shot-trick-pukeygoddess-gif-18083399484518002689

[ivory mountain]

i need this game

[spring latch]

wow I have discovered an odd system bug

on specifically this 2009 Sony TV I found that for some reason my B250M hates it and causes bugs typically associated with memory corruption

but wouldnt ya know it the TV was the issue and going back home to my ASUS VG32 fixed it

[waxen saddle]

I feel equal parts stupid and joyful because I hvae just now noticed that my monitor has USB Ports on it.
One B
one C
and the rest are gen 3 A

This is an increcible and joyful discovery

I bought this thing like 3-4 years ago

[fickle ivy]

Will this cause any major performance issues or is it fine?

[formal sonnet]

Computer is going to overheat and explode - turn it off now! (Just kidding - it's likely fine.)

[lean heron]

its fine

[fickle ivy]

A 280mm AIO might be slightly overkill for my CPU anyways

But it should be able to keep it pretty quiet

I hope the pump isn't loud

Honestly the cooler might use more power than the CPU 😭

[slim mauve]

Hey people is gigabyte a good and reliable GPU brand?

[lean heron]

It is

[hexed tapir]

bro you lost like 0.02% cooling capacity you need a new one

[bright surge]

id be more worried about a potential leak

[limber nacelle]

why would it leak? it's just the fins

[sleek sluice]

Are you playing with Coop?

[limber nacelle]

wait wot?

[sleek sluice]

Coopa.

[limber nacelle]

naur

[rigid fjord]

not to bad, but it's been raining on and off while being warm

[fallow yacht]

i meant a different climate

[shadow hedge]

why doesn't setting low or normal black level on my monitor change anything?

[kindred halo]

60MB/s on an external 2.5in HDD that reads "caution" on Crystaldisk is honestly pretty surprising

[tidal geyser]

hdmi black levels?

[sinful cobalt]

what cpu

[shadow hedge]

Yes, isn't it the setting for full range or limited rgb

[fickle ivy]

I5 12500t

[kindred halo]

Mate a single tower will keep that cooled

[tidal geyser]

its not

it just makes your image more black

[sinful cobalt]

lp cooler*

[shadow hedge]

It works on my tv but not monitor

I have acer ka242y p6

[tidal geyser]

acer

[kindred halo]

Lmao yeah

[sinful cobalt]

https://tenor.com/view/dumping-trash-daniel-labelle-frustrated-garbage-trash-gif-27186872

[tidal geyser]

the T class are low power versions of their desktop counterpart

[shadow hedge]

It's not that bad dude

[solid quail]

Gently blowing at the cpu* /j

[sinful cobalt]

thats what they all say

[molten sleet]

i love how inexpesive oled monitors have become so quickly

[kindred halo]

Why buy Acer 💔

Lenovo is in the 6-800 range for the LOQs

[limber nacelle]

nah my left buttcheek could probably cool it fr

[formal dragon]

god bless oculink

[tired fog]

do i get

• WD Red;
• Seagate Exos; or
• Toshiba N300
  for a NAS?

[cold skiff]

I dunno, do you?

[shadow hedge]

it was my brother's choice, i had a tight budget and i didn't look around much

[kindred halo]

Fair ig

[brave brook]

the one which is cheap and high capacity. Just be aware that some Reds are SMR, you don't want that for NAS

[tired fog]

thanks for the heads up

[brave brook]

i think WD Red is SMR, Red Plus and Red Pro are CMR

https://products.wdc.com/library/SpecSheet/ENG/2579-810238.pdf Red
https://products.wdc.com/library/SpecSheet/ENG/product-brief-wd-red-plus-hdd.pdf Red Plus
https://products.wdc.com/library/SpecSheet/ENG/product-brief-wd-red-pro-hdd.pdf Red Pro

SMR Red exists only in 2-6TB capacity

Red Plus is what Red was before they added the SMR models

[normal laurel]

gwha

[brave brook]

my Exos X16 16TB was fine from noise perspective too (not absolutely silent, but i can live with it), but there is a risk of fake new drives... that is, Chia miners faking SMART data to present used drives as unused

[kindred halo]

Hiiiii!!!!!!!

[spark path]

what are quite good sounding keyboard switches?

I don't want something that isn't completely silent but just doesn't make much noise that can be heard from more than 4 meters

[normal laurel]

Haiii :3

[brave brook]

if i remember correctly Topre was the best (rubber dome over spring) when it comes to quiet operation, but there is a whole #peripherals chat where people with more knowledge will have a say

[spark path]

oh thanks didn't know there was a chat for that

[brave brook]

https://hhkeyboard.us/blog/topre-switches 🤷‍♂️

[spark path]

does one exit with rapid trigger too? or is rapid trigger not a real thing and am fooled?

[kindred halo]

Hruuuu :3333

[stark flicker]

nah the most silent is choc v1 ambients

the "new features" such as rapid trigger mean magnetic switches which means you don't really have lots of choice with switches @spark path

[normal laurel]

Am silly

Wbu?

[kindred halo]

Very silly........

[wintry cliff]

random idea popped up in my head, would it be stupid to have an amd card and an nvidia card side by side on a x570 board? like for example if i have a rx6800xt and a 3060 for vram side by side

it should be dumb i think

[grizzled frost]

for what

[wintry cliff]

the 6800xt would be for 1440p gaming, the 3060 would be for anything non gaming, so ai inference and training, blender

that was my initial idea, but now that i say it outloud i realized the driver could hit each other lol

[brave brook]

well, you could join the 6800XT with 3060 through Vulkan driver support for LLM, but it is mostly pointless to do that.

driver will work fine, AMD IGP + NVIDIA GPU at same time works just fine. The issue is the memory bandwidth of 3060 for LLM

[wintry cliff]

can you elaborate more

[blazing tree]

Does a driver conflict cause a BSOD? For example, I'm using a disk from a friend whose PC configuration is different; I have AMD, he has Intel.

[brave brook]

token generation is determined by memory bandwidth of the GPU/CPU. 3060 8GB has only 240GB/s, RX6800XT has 512GB/s. So you would handicap the token generation to the speed 3060 8GB would be capable of.

You can always just try, but with MoE models like Qwen3.6-35B-A3B, there is a high chance it would be barely faster than using the 6800XT alone together with your normal system RAM.

With llama.cpp you can use one of the tools it comes with to figure out the ideal configuration to offload as little as needed to CPU, without a major performance hit (there is a hit, but still not 100t/s prompt processing instead of 2000t/s. You might drop to 600-800t/s, which is still reasonable).

[brave brook]

usually it installs what it needs, but what you describe often happened in older systems when you went from IDE compatibility mode to AHCI or RAID mode, or other way around

[brave brook]

for example on my 4080 Super i could run llama-fit-params.exe -c 200000 -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ3_XXS (i want that model with 200k context length), it will try few setups and then it will throw at me something like:

-c 200000 -ngl 41 -ot "blk\.29\.ffn_(gate|gate_up|down).*=CPU,blk\.30\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.31\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.32\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.33\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.34\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.35\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.36\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.37\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.38\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.39\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU,blk\.40\.ffn_(up|down|gate_up|gate)_(ch|)exps=CPU"

What that means that i could use those arguments to get 200k context with that model on a 16GB card, it would load all 41 layers to GPU, but then it would move some parts of those layers from layer 31 to 41 (30-40 because zero based numbering) to the CPU/RAM. It would move only those parts which are slowed down less by math being done on CPU & RAM and kept the parts which require GPU performance on the GPU, together with the context

[solid quail]

there's plenty of RAM like this for sale for around this price. Usually around 10€ per 4GB dimm

[wintry cliff]

i mean im on 32gb rn but im on a b450m which means i cant add more ram

[wintry cliff]

hmm i see, i mean i have a 6650xt rn which is more cooked but planning to upgrade to a 6800xt/3080 soon

havent decided on which one

[solid quail]

I'm considering on getting 4 4GB dimms for 40€ to pair with my 9600K for a secondary rig so I wouldn't have to take one of my spare 2x8GB kits

[brave brook]

Having said that, in this use case it is not making much difference, but in more limited situations it can make a difference of night & day. In the end, you can always just try and see if it is enough for your use case. Or maybe you could run a normal LLM on the main GPU and let's say ComfyUI for image generation on the secondary GPU

[solid quail]

when based on my testing, 9600K did not have noticeable benefit with faster RAM unless you overclocked the CPU as well

[wintry cliff]

im planning to basically rebuild the am4 setup to a x570/5800x3d and a 3080/6800xt in the future basically coming from a b450m/5600 6650xt

[solid quail]

if you have a decent b450 board, I'd just stay on that

[wintry cliff]

its a b450m asus prime k-ii

[solid quail]

if not, just wait till the RAM craze settles and upgrade to am5

[brave brook]

imo the real issue will be GPU sizes and PCIe slot layouts

[wintry cliff]

micro atx

[wintry cliff]

i mean so far the x570 boards i checked does allow pcie gen 4 x8 on 2 slots of the x16 slots

which would equate to pcie gen 3 x16 on each slots right, shouldnt it be enough

[solid quail]

I recently picked up a free b450m mortar max, 3400G and 16GB of hyperx fury RGB. Going to buy a 5900X and replace my main system with that

[brave brook]

i meant the layout of them vs GPU sizes. Having two x16 slots at x8 doesn't mean anything when they are spaced out in a way that first GPU blocks the slot 🙂

[wintry cliff]

ohhh

[brave brook]

i literally have two 1080 Ti's with aftermarket coolers which make them 4 slot GPUs and i can't put them on a X570-F, because the slots are 3 slots apart (1st and 4th)

if your GPU is only up to 3 slots - great. If it is more than that, RIP

plus even with 3 slots, be prepared for the GPU at top to have toasty times due limited airflow (and heat from bottom GPU)

[wintry cliff]

ah i see, so is it better to just fuck it and use one gpu

or use a blower style card

[brave brook]

i mean i have two 5070 Ti's in my LLM box and the top GPU is 15-20C toastier than the bottom one, but then i run it on X399 Taichi Threadripper board which has much more x16 slots and one of them is where i need it to be

[wintry cliff]

thats a lot of money tho lol

[brave brook]

yes, but not that much when you consider the price one would have to pay for the Claude sub ($100-$200/month).

but if you are fine with MoE models like Qwen3.6-35B-A3B or Gemma 4-26B-A4B, then your 16GB card will do fine too. Will be slower than a Claude or GPT model, and possibly bit 'dumber' than a Qwen.36-27B or Gemma 4-31B dense model, but still good enough for many tasks.

[wintry cliff]

I mean im only paying for a 20$ a month subscription but the fucking usage limits pissing me off so much im considering to switch to codex

[brave brook]

Windows or Linux ?

[wintry cliff]

Windows

[brave brook]

1. Go to https://github.com/ggml-org/llama.cpp/releases, for latest release (they build after every commit) pick Windows x64 (Vulkan), download, extract.
2. open command line in the folder you extracted the package, in command line run llama-fit-params.exe -c 200000 -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS (or the other quant you want from https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF). It will download the model to .cache/huggingface, it will take a while, it's 13-20GB depending on quant, you don't want to go under 3-bit.
3. take the command it generated and do llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS plus whatever the command said - in my case it is:

llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS -c 200000 -ngl 41 -ot "long list of filters i can't paste here"

4. after it is started, you can open http://localhost:8080 and chat with the model

[brave brook]

and to use it in Claude Code - you only need to create a batch file to set environment variables:

@echo off
echo "Claude Code - Local"
set ANTHROPIC_AUTH_TOKEN="llamacpp"
set ANTHROPIC_BASE_URL=http://localhost:8080
set CLAUDE_CODE_ATTRIBUTION_HEADER=0
set API_TIMEOUT_MS=3600000
set DISABLE_TELEMETRY=1
set DISABLE_ERROR_REPORTING=1
set CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY=1
set ANTHROPIC_DEFAULT_SONNET_MODEL="model_name"
set ANTHROPIC_DEFAULT_HAIKU_MODEL="model_name"
set ANTHROPIC_DEFAULT_OPUS_MODEL="model_name"
set CLAUDE_CODE_DISABLE_1M_CONTEXT=1
set CLAUDE_CODE_AUTO_COMPACT_WINDOW=200000
set CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=95
claude %*

to make easier to set model name, you can use --alias NAME (where NAME is what you want it to be called in API).

And that's it. You run the batch file, it will open claude code CLI against your local LLM server and local LLM model

you could do these in .claude/settings.json too (in "env": {} section), but i prefer environment set via batch file, or bash alias in WSL

[wintry cliff]

I only have a 6650xt right now which means a 8gb vram + 32gb ram haha

Might be cooked for a model that big

[brave brook]

Will work. Slower, but you can always test it. Forgot few more params to save a bit more VRAM, but I left my home right now, so I can post it only later. Stuff like disabling visual parts of the model, reserved space etc

[jade torrent]

flat

[brave brook]

Not saying it will be fast, but even 10-15t/s without usage limits means work is getting done.

[jade torrent]

oh god I never checked this for a long time

[vapid ibex]

@wintry cliff

[sinful cobalt]

time to put my 5060ti on the x16 slot and the 5090 on the electrically x4 slot /s

[deep musk]

Nah, x1 slot

[wintry cliff]

Okay that doesn't sound too bad actually

Ill try to run those when i get back home

[calm idol]

gotta limit to low reasoning for that though imo. otherwise you sit legit minutes there for a single answer/agent loop

[brave brook]

Yeah you wait for a slow answer. Which is still better than no answer for 5 hour because you ran out of the short limit, or no answer for 6 days because of weekly limit

[brave brook]

Also what do you want to use it for? Coding or OpenClaw/Hermes type of stuff?

For OpenClaw etc maybe even a Gemma 4 E4B in q5 quant might be enough

[wintry cliff]

Coding mostly

[broken lark]

Hey guys, is there a program or app that deletes bloatware and disables annoying stuff on windows?

[ocean axle]

yes

[grizzled frost]

Yes, but how much each thing does varies

It's best to just manually disable things, so you don't accidentally fuck your install with some weird third party software

[brave brook]

Then Qwen3.6-35B-A3B should be your starting point, with UD4 quant from unsloth, with 200k context.

[tidal geyser]

https://tenor.com/view/2-gif-7749019966943875056

[broken lark]

Hmm alright, and what about bloatware ? Or is all very agressive?

[tidal geyser]

you uninstall it?

[grizzled frost]

Base windows install doesn't even come with much "bloatware"

At least nothing that you can't uninstall in like less than 10 minutes

[broken lark]

Im resetting my pc, and want to disable most of the bs haha

[broken lark]

Alr

[ocean axle]

"copilot"

[tidal geyser]

you just unisntall it

[grizzled frost]

Pretty easy to opt out of it completely

[tidal geyser]

its just a web app

[ocean axle]

i prefer my solution to this problem

its the best way to avoid windows bullshit

[tidal geyser]

well i know the best way to do that

[grizzled frost]

"just install linux" ain't really a great solution

[brave brook]

In half a hour or so I can test you a worst case scenario - RX580 4GB+128GB DDR4-2133 😄

[grizzled frost]

at least not for the vast majority of users

[broken lark]

I’ll have a look then

[ocean axle]

i was going to say "dont use windows", but yes, im aware

[tidal geyser]

eh

its reached a point where i suspect most people can manage

[ocean axle]

for most things most can manage

[grizzled frost]

It'll be fine like 90% of the time, but that 10% is going to really suck for uneducated users

[ocean axle]

the problem is, the exceptions are still far too numerous

[tidal geyser]

people still think gaming on linux is faster than windows on most distros stock