general-chat

[dense locust]

5060 best 720p gpu to ever exist

🔥

[worn night]

except when you have the experts of newegg discord at hand

[wary orchid]

i know that, but troubleshooting and stuff, i dont really got the time

[acoustic palm]

yup then the whole process is easy

[wary orchid]

doing something wrong scares me too

[worn night]

this place is truely the best server of tech ive been in

intel is close second and amd third

[worn night]

ltt and ztt are weird and corrupted, respectively

[acoustic palm]

unless you do something wrong, or you get very unlucky, theres no troubleshooting to be done

[fresh timber]

Quok

Look

[hoary zodiac]

We serve the consumer

[acoustic palm]

theres like 1000+ tutorials online

[acoustic palm]

damn

[hoary zodiac]

comrade

[worn night]

thank you brother

[fresh timber]

9070xt runs this game so well

Just don’t turn path tracing on

[deep parcel]

what game?

[lone mango]

Looks like indiana

[misty sparrow]

to be fair path tracing tanks any GPU right now without enhancements

and usually to no visual benefit

[gleaming sedge]

I've made like the biggest crime with a lain li o11 dynamic mini snow

[proud jackal]

show

[gleaming sedge]

One sec

[rigid sinew]

I did

[worn wave]

(If that is Indiana Jones) the game in general is quite well optimized. Bit vram heavy but otherwise runs quite well

[fresh timber]

It runs well until pathtracing

[worn wave]

Well yeah lol

[fresh timber]

I’m installing a recommended Linux distro for ai rn

[acoustic palm]

@worn wave do yknow stuff about iems

[fresh timber]

Good job

[gleaming sedge]

You sure?

[worn wave]

Nope

[acoustic palm]

do you know anyone who does

[worn wave]

Nope

[fresh timber]

Mhh

Which one

[worn wave]

Well, optimum tech made a video about them but that's the closest connection I have lol

https://youtu.be/CtoxlafzeyI?si=paZjdQ8OkKX5MxTF

https://youtu.be/QgnF2uIDU6o?si=y7mqurTYpk9dBXnX

[acoustic palm]

yeah i was looking for the like cheap good experience

just need some advice on stuff

[fresh timber]

Quok

Buy 5090

Easy

[acoustic palm]

i want sound quality + mic quality

[gleaming sedge]

Ngl i like it

[proud jackal]

what the frick

[lone mango]

[gleaming sedge]

Hey all of the panels do go back on when not plugged in

[lone mango]

https://tenor.com/view/geralt-fortnite-griddy-looped-dancing-gif-18178584610833807159

[fresh timber]

I’m on Linux

[misty sparrow]

oh god no it's june

[grand kernel]

got my daily stable 110% PL profile done

time to get a daily stable 70% PL profile and compare benchmarks

[lavish stag]

Building my sister a pc using my old 1650, what psu wattage should i go for

[fresh timber]

I got arch arch installed

I use arch btw

[grand kernel]

so at 70% PL the 9070xt is hovering 2530MHz at 210W

[lone mango]

[lone mango]

Thats crazy

[grand kernel]

70% PL can run a steeper undervolt

-140mV vs -105mV

[lone mango]

yeah

[fresh timber]

i cant get LACT to see both gpus

rn

that sucks

oh

it does

[grand kernel]

the 110% PL profile hovers around 3020MHz at 334W

now to benchmark the performance difference

2560x1600, FSR 3 balanced, all settings maxed, path tracing, no frame gen

110% PL: 57.57 avg fps, 50.26 min, 65.71 max
70% PL: 50.62 avg fps, 44.17 min, 59.85 max

not bad

[fresh timber]

anybody know how to have a ai model scale between gpus

[grand kernel]

88% of the performance for 63% of the power draw

psycho RT, path tracing off, FSR 3 quality, no frame gen:

110% PL: 83.70 avg fps, 73.68 min, 100.34 max
70% PL: 72.66 avg fps, 63.50 min, 90.19 max

this might be the way to go over path tracing for this card

[sinful blade]

Someone does here

@fair onyx

[fresh timber]

he doesnt know amd 😦 but its wroth a shot

[fair onyx]

Me!

[fresh timber]

i got it to work on 1 gpu, reading tells me i need to get bigger model

[fair onyx]

Oh, you're using ollama

[fresh timber]

HI SYATN

u dont know it

[fair onyx]

I know ollama somewhat

[fresh timber]

irrc

[fair onyx]

AMD also has a lot better support now than it did before

[fresh timber]

im running dual gpus rn

[fair onyx]

what are you trying to do?

[fair onyx]

I do too haha

[fresh timber]

trying to see if i can get my 9070xt and 7700xt to work

i got 28gb of vram combined

[fair onyx]

That should be possible. I assume you are on linux?

[fresh timber]

yuh

[fair onyx]

And you're in ollama, so you're tunning what, .gguf format?

[fresh timber]

i installed Cachyos arch

[fair onyx]

I mean, on paper, it should just work ™️

[fresh timber]

oh good

i had to distro hop a few times until i found a distro that worked with the 9070xt kek

one worked and i got happy then i figured out it was running on 7700xt onloy

[fair onyx]

Yeah, gguf and lcpp (what ollama is based on) are extremely flexible. To the point where you can often times mix and match different brandes of GPU's even (AMD/NVIDIA/Intel)

[fresh timber]

uh im ngl im just downloading random models

[fair onyx]

and in fact, you can even pair together different PC's to work in unison as well. Like Thunder bolt between Mac mini's and stuff

[fresh timber]

im downloading gwen:32g rn my dads internet sucks rn

[fair onyx]

makes sense. All text based tho?

[fresh timber]

I SAW THAT

[fair onyx]

Oh nice, Qwen 32b is really sollid. What bit depth are you downloading? Q4 should be good on 28GB I think

[fresh timber]

yeah text based for now

when i learn more ima try image based

[fair onyx]

test is the most useful for sure

[fresh timber]

dads internet sucks so bad my messages are comming in late

lol

[fair onyx]

Let me know if you ever need help. I am actually a professional in that scene lmao

[fresh timber]

pog

[fair onyx]

its where I made my money (training stuff for companies)

[fresh timber]

ollama discord got me to the os im using rn

[fair onyx]

Oh nice

[fresh timber]

im.. a fan of this os

its nice

[fair onyx]

thats good to hear. I have never had the balls to swap to linux lmao

too man bad experiences with linux over the years 😭

[fresh timber]

lmfao i only swapped for GPU support

ROCM doesnt support my gpu on windows yet

[fair onyx]

Ahhhh yeah

9070XT is too new I think

[fresh timber]

yeah

it has good vram

im weird id make a cluster of AMD gpus

[fair onyx]

I have 40GB VRAM together as of now

My dream goal is to buy 2 of the intel dual B580 GPU's with 48GB each and pair them for 96GB VRAM. Then I could run some real fun stuff lmao

[fresh timber]

I swapped to my phone smh

Internet so bad en installing a bigger model

[fair onyx]

soooo, how big is the file you are downloading?

[fresh timber]

18gb

[fair onyx]

ok yeah cool, that should be good

is it Q4?

[fresh timber]

yuh

Uhhh

[fair onyx]

Q4 iq imatrix would be the best bet for you

Do you understand how quantization works for models?

[fresh timber]

No

I sent the link it’s just not sending

Linux sure does know how to squeeze every drop of bandwidth

[fair onyx]

So models are usually FP16/FP32 accuracy, whixh is insanely high accuracy, but uses a ton of memory

What we do is quantize the models (lower their accuracy) to use less memory. You can lower it more and more, but eventually, it will start to get dumber. There are special quantization types that make them less dumb

So Q4 will chop all of the values off from 16 bits to 4 bits. But something like IQ Imatrix, will measure every bit and give the more sensitive ones more space, and the less sensitive ones less space. That allows the model to be more capable at the same size

Qwen 32b at FP16 is like 75GB, for example

[lone mango]

@upper rain

[fresh timber]

Wow

[fair onyx]

Q8 is half the accuracy (still like 99.98%), but only uses ~34GB

[fresh timber]

It’s getting close to bed time so I can always install the big boys over night

[fair onyx]

Q4 is 1/4th the accuracy (about 98.6%), and is about 18GB

[fresh timber]

Wow

[fair onyx]

and there are other sizes of course

[fresh timber]

This stuff eats vram

It’s weird for me to think I have vram

[fair onyx]

oh yeah. Wait until you look at models like R1 lmao, or god forbid behemoth

[fresh timber]

😭

[fair onyx]

R1 uses ~1.3TB of VRAM, and behemoth uses ~5TB

[fresh timber]

This is making me wanna make a system with a mid cpu but good pcie

I would be weird and put toghter a few rx6800

[fresh timber]

Holy

[fair onyx]

But as somebody who has run some bigger models, ~32B models are the peak right now. The bigger ones are much slower to get updates, and less people run them, so improvement is really slow on them

[fresh timber]

Going from 12-28gb vram means I can finally do big boys things

If this goes well I’ll buy a better adapter to plug in my 2nd gpu

[fair onyx]

Q3 32B is really solid in general (especially for complex reasoning/code)
Gemma 2 27b is also insanely solid (for question and answering)
GLM4 32b is great for creative stuff

You can find models that can do alll sorrtttsss of stuff

[fresh timber]

What’s the cheapest Nvidia gpu you can think of with 16gb

[fair onyx]

a modded 3060ti lol

[fresh timber]

Lolll

[fair onyx]

mine was $300

[fresh timber]

Wowww

You ever consider a modded 2080ti

[fair onyx]

its a 3060ti FTW3 ultra modded from 8GB VRAM to 16GB VRAM

[fresh timber]

Ooo a nice one too

[fair onyx]

20 series is a little too old to be good for modern AI

doesn't support popular tools

[sinful blade]

Full Model (DeepSeek-R1, 671B parameters)
GPU: Multi-GPU setup, minimum 16x NVIDIA A100 80GB GPUs
VRAM: Approximately 1,500+ GB without quantization
CPU: High-performance server-grade processors (AMD EPYC or Intel Xeon)
RAM: Minimum 512GB DDR5
Storage: Fast NVMe storage, 1TB+ for model weights and data
Power Supply: Enterprise-grade redundant PSUs, 5kW+ capacity
Cooling: Data center-grade cooling solution
Networking: High-speed interconnect (100+ Gbps)

DeepSeek-R1-Distill-Llama-70B
GPU: Multiple high-end GPUs, minimum 4x NVIDIA A100 40GB or equivalent
VRAM: 181GB for FP16, ~90GB with 4-bit quantization
CPU: Server-grade multi-core processors
RAM: Minimum 256GB
Storage: Fast NVMe SSD, 200GB+ for model weights

DeepSeek-R1-Distill-Qwen-32B
GPU: Multiple GPUs, minimum 2x NVIDIA A100 80GB or equivalent
VRAM: 65GB for FP16, ~32GB with 4-bit quantization
CPU: High-end server-grade processors
RAM: Minimum 128GB
Storage: NVMe SSD, 100GB+ for model weights

DeepSeek-R1-Distill-Qwen-14B
GPU: High-end GPU, minimum NVIDIA A100 40GB or multiple RTX 4090
VRAM: 28GB for FP16, ~14GB with 4-bit quantization
CPU: High-performance multi-core (16+ cores)
RAM: Minimum 64GB
Storage: SSD, 50GB+ for model weights

DeepSeek-R1-Distill-Qwen/Llama-7B
GPU: Consumer-grade GPU, minimum NVIDIA RTX 3090/4090 (24GB VRAM)
VRAM: 14GB for FP16, ~7GB with 4-bit quantization
CPU: Modern multi-core (12+ cores)
RAM: Minimum 32GB
Storage: SSD, 30GB+ for model weights

DeepSeek-R1-Distill-Qwen-1.5B
GPU: Entry-level GPU, minimum NVIDIA RTX 3060 (12GB VRAM)
VRAM: 3.9GB for FP16, ~2GB with 4-bit quantization
CPU: Modern multi-core (8+ cores)
RAM: Minimum 16GB
Storage: SSD, 10GB+ for model weights

@fresh timber

[fair onyx]

the main backend I use (EXL2/EXL3) has no support for anything below 3000 series, or AMD's RX 7000 series

[fresh timber]

Darn

[fresh timber]

DAMNNIT

I had a rx6600 I would have added

[fair onyx]

You CAN use older GPU's with ollama tho (not older AMD ones tho as far as I know, i think that one is also onl 7k+ series)

[fresh timber]

Oh shoot I forgot 7600xt exist

[misty sparrow]

Why is the AI responding like an introvert with no social skills

[fair onyx]

yup, this is spot on (tho nobody should run any of these models now days cause they are severely out-dated)

[fresh timber]

16gb gpu that got bad reviews for gaming performance

But it has vram

[fair onyx]

Yeah, thats what matters for AI uses

[fresh timber]

Ehh they cost the same as a rx6800 but 6800 misses features

[fair onyx]

I have my 3090 24GB and my modded 3060ti 16GB for 40GB VRAM total

[fresh timber]

Are u exited for that Intel stuff

[fair onyx]

EXTREMELY

[fresh timber]

48gb vram irrc for a good price

TWO GPUS ON ONE CARD

[fair onyx]

I just need them to come out, be good, and intel to improve support for SYCL in important AI frameworks (which they are making way faster progress with than AMD)

[fresh timber]

I heard they suck at ai rn in ollama server

[fair onyx]

gguf/ollama will be the easiest, but something higher-end like EXL2/3 (what I use) will be harder to have intel support for

[fair onyx]

I am not surprised haha

Intel is very very new to this, but let me tell you, they are making WAY faster progress than AMD

[fresh timber]

What I’m seeing is if your like me AMD is just fine, if your a bit more serious go Nvidia

[fresh timber]

For sure, this is their 2nd gen of desktop gpus too

[fair onyx]

Like, it took AMD longer than the entire existence of Intel ARC GPU's to add proper ROCm support for theri 7000 series cards

And yet here Intel is already adding support for their insanely new cards lp;

[sinful blade]

I just want a gpu stronger than mine

[fair onyx]

I am very happy with my 3090. It runs everything I need very well

I have no need for stronger as of now

[fresh timber]

I should try stable diffusion next

[worn wave]

Wait didn't you have a 5700xt?

[sinful blade]

Wife has it, as it’s the bare minimum to play oblivion

[worn wave]

5700xt >> 1660S

[fair onyx]

For sure. Thats my specialty, and I have a TON of experience with image gen

[sinful blade]

She crashes 10+ times a day but she has fun

[worn wave]

Ah

[fresh timber]

Does it scale between gpus

[fair onyx]

AMD moment haha

[fair onyx]

it does NOT, but image gen is typically a lot less VRAM intense than text gen

[fresh timber]

I crashed 3 times trying to test the 9070xt

Wanna know what’s interesting

[sinful blade]

I just want something midrange for myself and her

[worn wave]

You can get basic image gen on 8gb

[fresh timber]

I crashed less when I was running at pcie 1x than 16x

[fair onyx]

for example, SD1.5 can run on like 3GB VRAM, SDXL on 6GB VRAM, SD3 Medium on 8GB VRAM, and SD3 Large on 24GB VRAM (but they can all be quantized to be smaller) @fresh timber

there are also other models than just stable diffusion. It really depends on what you would wanna generate

[fresh timber]

Before the rocm days my 3060mobile beat my 7700xt

[fair onyx]

I do a ton of furry generations and stuff, so I models specialized for that. But There are models for everything, honestly

and my god, if you know what you are doing, you can generate some absurdly high quality stuff. I don't sell anything I gen, but I could easily

[misty sparrow]

Honestly some of the stuff I've seen looks stupidly convincing

[fair onyx]

but like this is a standard human generation ( trained the character for a friend) on a furry model, and the quality is still this high with no edits or anything

with some edits, I could make it way higher quality of course

[fresh timber]

Ai is fun

[worn wave]

Sytan is Flux still really good for realistic image gen? I remember when it first came out it was kinda smoking the competition in that aspect

[fresh timber]

@fair onyx do you like those ryzen ai max systems. Like 64gb vram if you configure it that way

[misty sparrow]

You can't fix small details like the girl being too close to the stove and the architecture being unrealistic but still impressive

[fair onyx]

Furry warning lol. This one model can do over 400 different art styles, and all of these are unedited simple outputs. The quality has gotten unreal 😅

[fresh timber]

:0

[misty sparrow]

oh god not furry art

[fresh timber]

96gb

[fair onyx]

Silence, non girlfriend haver

[fresh timber]

Lmfao

Don’t you have a boyfriend silly

[misty sparrow]

honestly I think it's a good thing I don't have a GF or anybody

[fair onyx]

Yeah, these little things are neat, but pretty weak for some things

[misty sparrow]

no joke

[fair onyx]

Yeah, I have a boyfriend lol. I am what people call, Gay AF lol

[fresh timber]

Did you see what this one. It’s desktop level

[grand kernel]

man I'm disappointed

[fresh timber]

Let’s bully him for not having a lover

[fair onyx]

the GPU is, for gaming, but for AI, they are pretty weak. I believe the top 395+ AI MAX is like... half the AI performance of a 2060?

[grand kernel]

-140mV is stable if I set it in game, but I can't actually launch the game without going back to -110mV

[fair onyx]

for some things its great, for others, not so much

[grand kernel]

I wish the radeon driver would allow a higher negative frequency offset

[fresh timber]

Oh.. so u know why it’s not quick in ai.. it has ai in its name

[fresh timber]

.. run 3d mark

[fair onyx]

it depends on what type of Ai you run

[grand kernel]

it's unstable when it's trying to boost up to the highest parts of the curve

[fresh timber]

I did -140 and I crashed

[grand kernel]

3dmark isn't a stability test lol

[fresh timber]

Steel nomad is rough

[grand kernel]

I can pass steel nomad on these settings that can't even make it to the main menu in cyberpunk...

[fresh timber]

wtf

For me I can play Fortnite fully raytraxed but can’t run steel nomad

[grand kernel]

steel nomad passes at -125, cyberpunk crashes within 5 seconds at -120

lol

[fair onyx]

I currently have models for:
Math
Code
Complex reasoning
Storry Writing
Questions and Aswering
Roleplay/Dungeon use
Medical diagnosis
Image analyzation
Captioning
Silly Stuff
And more

[fresh timber]

Does vram speed make a difference Mr sytan

[fair onyx]

I have models that can solve linear algebra, calculus, trig, differential equations and more like its nothing

[misty sparrow]

what is "silly stuff"

[fresh timber]

His ai boyfriend

[sinful blade]

Ask your models if they can come up with a business name for me @fair onyx

[fair onyx]

jokes, memes, stupid things lol

[fresh timber]

Train a model on you and your boyfriends messages

[misty sparrow]

wrong person, ask me and i'll make your ass go broke

[sinful blade]

I need one for a pc building business to put on business cards

[fair onyx]

No need to even train one. You can just feed it as examples to replicate lmao

[grand kernel]

let's see how much performance I lost going back to -110mV on 70% PL...

[fresh timber]

lol

[fair onyx]

I would never do that tho, I love my boyfriend a lot, and no AI could replace him lmao

[fresh timber]

Imagine a ai that acts like your boyfriend

[fair onyx]

Artificial Intelligence is no substitute for his Natural Stupidity

[fresh timber]

Ai is artificial stupidity

[sinful blade]

Didn’t some Instagram model do that

[fair onyx]

Silly stuff.

[fresh timber]

Idk

Damn 5070ti is slow

[fresh timber]

Oh you have a ui

[misty sparrow]

I would just paste the entire script of a Breaking Bad episode and ask it to add brainrot

[fair onyx]

Oh yeah, you will wanna look into Silly Tavern, trust me. Its like THE software

[fair onyx]

oh, it can do that easily lol

[fresh timber]

Is verbose a good way to test speed

[fair onyx]

verbose usually just means it prints all of the additional (usually hidden) information for debugging

[grand kernel]

went from 50.62 avg fps to 48.70 😦

[fresh timber]

O

[fresh timber]

Test in ai

[fresh timber]

Offtopic, wtf do the ai computer units do in these gpus

[misty sparrow]

This one is so far removed from the original prompt it's pretty unrecognizable

I guess that's the point

[fresh timber]

i haven o girlfriend

[fair onyx]

Thats the best one I have unforunately, I used to have a model that did it wayyy better before lmao

[fresh timber]

what dumb stuff should iadd to my pc

[fair onyx]

just a sec

[fresh timber]

pcie to something

i just figured out how ima put the 7700xt in my main rig when its time

[misty sparrow]

that software that looks like a dating sim but helps you do your taxes better

[fresh timber]

ima just run my riser outside my case

[fair onyx]

@misty sparrowThere you go lol

I have much better models now than the last time I used this lol

[fresh timber]

imagine if they gave the 9070xt 24gb vram or even 20

imagine'

AI

[fair onyx]

like the 7900XT/X? lol

[fresh timber]

yeah

7900xtx has 24, 7900xt has 20

9070xt is a xtx in performance essentially, and in raytracing its beyond

[fresh timber]

i need a model like that

[fair onyx]

death. lmao

[fair onyx]

this is a card, which can make models behave in different ways

[fresh timber]

oo wow card?

[fair onyx]

Yeah, so in Silly Tavern or other UI's, you can add info to make models behave in different ways

[fresh timber]

o wow

[fresh timber]

whats up with people spamming 3060s

[fair onyx]

thats how you can get a single model to do sooo many different things

[fresh timber]

wow

[fair onyx]

These are all different ones I have made (that I can share lol)

[fresh timber]

wow

oo the model is 85% finished

@fair onyx got any other model suggestions i can install before i go sleep foro the night ill sleep in like a hour or two

[fair onyx]

@fresh timber

These are all the same model with the card directing it on how to respond

You can do sooooo much with local models lol

[fresh timber]

SKIBIDIII

[fair onyx]

You can make these models do insane things lol

gonna send something in DM's

[fresh timber]

i like

if i get smart enough id setup a little bot to do ai stuff in my frined groups server

[fair onyx]

I have one, just a sec

its my current fave model

[fresh timber]

time to see what happens

[fair onyx]

https://huggingface.co/mradermacher/QwQ-Snowdrop-GGUF/resolve/main/QwQ-Snowdrop.IQ4_XS.gguf

that one

Snowdrop is what I have been sending pictures of in here

snowdrop is a thinking model made for roleplay/creativity

[fresh timber]

i get to learn how to make ollama use them

pOG

[fair onyx]

Snowdrop is an incredible writer if you tell it what to do

sending a pic example in DM's

[fresh timber]

SYATN

IT WORKS

ITS SCALING BETWEEN B OTH GPUS

[misty sparrow]

so you could get an AI to do anything?

there's AIs used as therapists right now good lord

[fair onyx]

Yes

[fair onyx]

I have some made just for that. They work better than most IRL therapists I have used

[fair onyx]

hell, I have medical diagnosis models that have been able to properly diagnose issues my mother and family (and myself) have had for years that doctors have overlooked or ignored

[misty sparrow]

I feel like those would work better if they weren't text input and you could talk to it and have it respond back using voice

[fresh timber]

LOOK

[fair onyx]

There are ways to do that locally

[fresh timber]

you think i can get away with bigger

[misty sparrow]

even giving it voice feedback?

but with a realistic voice, not that Microsoft Sam stuff

[fair onyx]

Yes

[fresh timber]

quick math im using 24gb of vram

[fair onyx]

yes

just a sec

[fresh timber]

swyatan hwo do i get a UI

[fair onyx]

You will wanna use silly tavern. Its a lot to learn, but its worth it

[fresh timber]

oh

[fair onyx]

https://github.com/SillyTavern/SillyTavern

[fresh timber]

offtopic but ites cheaper for me to buy another 1x to 16x riser than it is to get a 1x to 1x riser

so guess whos buying another 1x riser

seriously why is it like that

[fair onyx]

darn it, I am trying to find a text to speech model, and I can't find it

[fresh timber]

@fair onyx when im finally home ima add a rx6600 for another 8gb of vram and ima ee what happens

[fair onyx]

it won't work, I don't think

RX6000 series doesn't have AI drivers AFAIK

[fresh timber]

oh shoot

i forgot the rx6600 doesnt have offical rocm

flip

i forgot

side note, i bought another riser

im daisy chaining them

[fair onyx]

@misty sparrowThis is a demo of a new text to speech model that can voice multiple people at the same time, with tonal inflection, pauses, breaths, laughing, and more

[S1] Dia is an open weights text to dialogue model.
[S2] You get full control over scripts and voices.
[S1] Wow. Amazing. (laughs)
[S2] Try it now on Git hub or Hugging Face.

[paper verge]

Yo

why is that soo good

wtf

[fair onyx]

[S1] Hey. how are you doing?
[S2] Pretty good. Pretty good. What about you?
[S1] I'm great. So happy to be speaking to you.
[S2] Me too. This is some cool stuff. Huh?
[S1] Yeah. I have been reading more about speech generation.
[S2] Yeah.
[S1] And it really seems like context is important.
[S2] Definitely.

[paper verge]

it sounds like your textbook voice analogs

lol

but its very good somehow

[fair onyx]

[S1] Hey there (coughs).
[S2] Why did you just cough? (sniffs)
[S1] Why did you just sniff? (clears throat)
[S2] Why did you just clear your throat? (laughs)
[S1] Why did you just laugh?
[S2] Nicely done.

This one is the insane one

[fresh timber]

i wanna make

[misty sparrow]

Sounds pretty convincing..why the hell are they able to cough now?

[fresh timber]

imagine robots walking around with speech generation

[fair onyx]

Oh, and look at this

[misty sparrow]

Could just be personal preference but I usually don't find listening to the female voices comforting

[fair onyx]

you can voice clone as well

This is the sample audio to clone

[S1] Open weights text to dialogue model.
[S2] You get full control over scripts and voices.
[S1] I'm biased, but I think we clearly won.
[S2] Hard to disagree. (laughs)
[S1] Thanks for listening to this demo.
[S2] Try it now on Git hub and Hugging Face.
[S1] If you liked our model, please give us a star and share to your friends.
[S2] This was Nari Labs.

And this is its clone

so it cloned both of their voices off of 3 seconds of audio

[misty sparrow]

wow

[fresh timber]

how good is deepseek

[misty sparrow]

i've seen it happen with AI videos being used for comedy like those Charlie videos

voice matches but the content in the videos is clearly AI-generated, but that's the humor somehow

[fresh timber]

damn its popular

[fair onyx]

deepseek R1 is massively out of our leagues (the 1.2TB model)

[fresh timber]

damn

im just on ollama looking for random big models to install

[fair onyx]

your hardware is nowhere near powerful enough for "big" models 😅

unless you mean popular?

[fresh timber]

big to be is 32b

[fair onyx]

If you have any questions about specific models, I have lots of experience. I have been doing text gen for over 5 years now

[fresh timber]

what does this mean

[fair onyx]

Just made this

[S1] Dang, is so cool (laughs). I need to show it to "I have no gf" and "Vextroboomin"
[S2] I know, right? It's amazing (Laughs). How could I not want to share this?
[S1] Share it where, the NewEgg discord server?
[S2] You know it. Thanks Sytan for prompting us to say this.

[fresh timber]

hi

[fair onyx]

Oh, what model are you using?

[fair onyx]

@misty sparrow

[fresh timber]

qwen32:b that one uses enough vram that it scales between my gpus

[misty sparrow]

Creepy

just creepy

[fair onyx]

Ok, so you're not using a system prompt (one of the ways you can give models personality), so its going to default to cold and unfeeling as a model (AKA, I have no emotions and stuff)

[fresh timber]

ill learn that stuff at a later date

oftopic but my gpus running 3.3ghz

[fair onyx]

Core speed matters a lot less than VRAM/inter connect speed for AI

[fresh timber]

i guess linux has a better driver

[fair onyx]

for my cards, I can limit my 3090 to 60% power and 3060ti to 50% power, and I still get 100% the performance of if it was using full power on both

[fresh timber]

u think i can get faster if i went from gen3 x1 to gen 4 x1

[fresh timber]

oh man my cards are using so much just having the model loaded

[fair onyx]

probably a bit? I use Gen 3 x1 on my system

[fresh timber]

mhh

ill try when i go home

[fair onyx]

its all I have, but the pcie speed does hold it back a bit

[fresh timber]

my main gpu will be going to 5.0 x16

2nd can go to 4.0x1 until i figure out how i can mout int into my case and have it connected to the buttom 16x physical slot

current plan is to daisy chain my risers to putside of my case

[fresh timber]

is this normal, but the 2nd card is using like 10w doing nothing and my main card is using 100w

[fair onyx]

The main card should not be using that much power. The secondary card makes sense though, that's about what a standard idle power draw is

[fresh timber]

odd

[fresh timber]

u ever got a model that can access the internet before

[fair onyx]

Any mall can access the internet if you set it up properly. Silly tavern has some extensions that allow you to do that via Google API

They can search terms, go through web pages, compile information, and then answer questions on it. It's not something I've ever done, because I haven't really had a big need for it

[fresh timber]

o wow

[fair onyx]

I've never said something like that up before, so I'm not too sure how in-depth it is, but I do know some people who have

[fresh timber]

i could see that being useful if ur asking about current events or about things thats constantly changing

[fair onyx]

Yeah, you could totally do that. I have considered doing it before, I just haven't gotten around to it

[fresh timber]

id figure it out if im making a simple chatbot for my friendgroup

[fair onyx]

I know it requires you to get a Google API access key. I think you can do like a hundred searches a day for free or something like that, and then the model would query the Google search engine and compile the information it gets back

[fresh timber]

that sounds like its fun

[fair onyx]

You can also set up agents. Lolms that have visibility capabilities, to where they can take screenshots of your desktop, and navigate around it. There are even some agentic models where you can type in a prompt and be like "Open up my settings to the display" And it will look around the screen, find a settings icon, click on it, and navigate through the menus to the display settings

That's also stuff I haven't messed with before, but I have some friends who do

[fresh timber]

i imagine my friendgroup asking the ai dumb questions

[fair onyx]

Yeah, when you first get into it, you'll do that all the time

And then you'll find out the amazing things you can do, and then you'll use it for that

@fresh timber did you see the snowdrop writing example I sent in DM's?

[fresh timber]

I did

[fresh timber]

my gpu problem is unrelated to ai, its just not downcllocking at all

😭

[fair onyx]

That sounds like it's stuck in performance mode

[fresh timber]

yup

it has no reason to be running max rn

its weird i give it a 24w pl but it uses 47

@grand kernel u get more control in linux btw, you can set lower powerlimits

it wont let me go below 46 it eems

ima go back to windows, i need to download some games overnight

[fair onyx]

I think I'm gonna have to leave you here for now. The boyfriend and I are gonna hang out@fresh timber

[fresh timber]

hanging out with out me, wowwwwww jk have fun

[fair onyx]

Best of luck with the AI. I should be able to help you with other stuff later on

[fresh timber]

what time is it for you

[fair onyx]

Midnight lol

[fresh timber]

oh hanging out at night

ok bye bye

[fair onyx]

We're long distance lmao

It's the only time he's free 😅

[fresh timber]

oh

[grand kernel]

I don't want to set lower power limit, I want to set lower frequency offset

[fresh timber]

ima reboot it might let me

[grand kernel]

i want to set something like -750MHz to stabilize an extremely aggressive undervolt at low power limit

but the driver limit is -500MHz

missing out on some performance at low power limit

[fresh timber]

powerstates on iwndows are broken

after fixing it im using ALOT less than i am on windows doing the same thiing

[lone mango]

Is that 12w with two monitors

[misty sparrow]

https://cdn.discordapp.com/attachments/849750633967910942/1376993589200158780/IMG_9906.gif

[fresh timber]

Yes

It can’t do that on windows

[lone mango]

Well higher vram clocks mean more power consumption

Assuming rdna 4 behaves that way and you didnt oc vram yet

[fresh timber]

Even at the same settings it can’t

[lone mango]

[fresh timber]

On Linux I can control it more and force it into a lower power state

[lone mango]

For me I dont have much difference between the two

But that is good

12w is really good

What was it on windows

[fresh timber]

No

Win used 50

[lone mango]

[fresh timber]

Somehow it’s more power efficient than my 7700xt in my small games

Loll

[tranquil pebble]

Windows itself is inefficient so no surprise there

[wary fjord]

https://tenor.com/view/goodnight-sleep-bedtime-gn-good-nite-gif-7992865519168581883

[tranquil pebble]

7:35 am here. Timezones moment.

[dusk knot]

The very opposite here

Good morning lol

[onyx merlin]

so whats better 7000 or 8000 series ? lets say Performance wise what do you think will have the biggest boost performance wise over current 4000-5000 series ?

as in 4-5 years will be getting one of those series when I get new pc

[worn wave]

4-5 years from now none of this will be relevant

[onyx merlin]

what u mean ?

gpus always relevant

[remote geode]

thinking if i should switch to linux once i get a second 1tb drive

I think my B580 will still be relevant in a few years

especially if it ages like fine wine

[worn wave]

Prices will have changed so much and multiple new generations will have released, so it's not worth even considering right now

[icy ore]

Yup it literally changes day by day lol

[frigid stag]

Yea that’s why 5060ti still trailing the 2020 cards XD

“Much performance increase”

[royal crow]

Deadeth chateth againeth

[proud jackal]

[worn night]

more dead chat

guys if you want an endless stream of people who have 0% knowledge of what they're doing that need help with intel stuff, go to intel discord

[topaz crown]

Hello @worn night! It appears that you're asking for help. Please ask again in the appropriate category so you can get faster help from people specializing in that field. Your question may not be answered here.

[worn night]

nuh uh

[icy ore]

Welp

[royal crow]

Go away dyno no one likes you

[grand kernel]

@fresh timber higher negative frequency offsets might actually be quite important for unlocking maximum performance from this card

[tranquil pebble]

at least it isnt mee6

[royal crow]

Lol

[grand kernel]

currently testing undervolt stability in sunset motel in cyberpunk, I can only run -90mV with -200MHz offset but I can run -110mV with -300MHz offset

average core clock is around 3080MHz, so there's no actual performance loss from the -300MHz offset (3450MHz - 300MHz = 3150MHz)

but every time the game crashed at -110mV I see it was due to a brief spike to 3200MHz or higher

so ironing out the brief spikes seems to be key

[remote geode]

what are we overclocking

*undervolting

[dusk knot]

Hhhhhmmmmm

Hardly any activity today. What has the world come to!?

[pliant berry]

Im on vaca

[proud jackal]

https://www.tomshardware.com/pc-components/gpus/rtx-pro-6000-crowned-the-new-gaming-king-but-its-usd10-000-price-tag-makes-the-all-gold-dhabab-rtx-5090-seem-cheap

Guys, hear me kut

Out

[gloomy berry]

hello everyone

[lone mango]

2020

[nocturne flicker]

Larping the Anti-Warlordism

[worn night]

no

[dusk knot]

Buy it

[nocturne flicker]

Don't kut

[velvet summit]

Why is it so highly reviewed

[fresh timber]

Chat my friend has a 5600g and 4060, what should he upgrade too, idk what’s good budget rn besides 9060xt soon

[mental yoke]

Can someone please help me in finding a certain tpm chip for my motherboard?

[topaz crown]

Hello @mental yoke! It appears that you're asking for help. Please ask again in the appropriate category so you can get faster help from people specializing in that field. Your question may not be answered here.

[dense locust]

5060ti 16g

[fresh timber]

Is that gpu no cap good

[dense locust]

The 9070xt is in a league of its own

What do they want with this upgrade

[fresh timber]

Frames per second

They are a me, we just play marvel rivals all day

[fresh timber]

I’m ngl

It has less overhead than b580

9070xt can be with lower end but not a b580

[upper rain]

9950x3D and 5090

[upper rain]

it's the least bad nvidia offering, at least by MSRP

[fresh timber]

Dam

[upper rain]

really though probably like a used 6800 or something, they're pretty cheap on hwswap

[fresh timber]

So it really is good

[upper rain]

$300 easy 16GB card

[fresh timber]

Me convincing friends to buy used is a challange

[upper rain]

slower than a 4070 but not by a ton. easy to OC to that performance, and you get 16GB

[fresh timber]

lol I’ll have to get bro to get a better monitor

Both my friends they twins still use 1080p panels 😭😭

[lone mango]

Probably last year yeah

6700 xt is over that last time I checked

[upper rain]

6700 xt pricing has always been wacky, you can always get 6800's for $350 these days and $300 isn't unheard of, especially if you shop without ebay overhead

[fresh timber]

One has a 6700 that keeps crashing in video games

Or it’s his 13600kf I’m not sure

I ran linepack in it no problem so I’m lost

[upper rain]

i'd be inclined to blame the CPU that melts itself to death, but who knows

[fresh timber]

Yeah

I’m inclined to

Mines didn’t melt it self and it’s younger

[upper rain]

how is the 9070 xt treating you? day 2 right?

[fresh timber]

😭😭

[upper rain]

i assume you have your bios updated tho

limits the degradation a lot

[fresh timber]

Struggling to games that arnt limited

[lone mango]

Probably

[fresh timber]

Yeah mines updated when i learned about melt melt

[fresh timber]

Would the pc going straight black cause problems

Like your in a ue5 game and out of nowhere black

[lone mango]

Try different games to rule that out

But a black screen is probably amd

[fresh timber]

All bro plays is Roblox rivals and Fortnite man

[lone mango]

This is unfortunate

[fresh timber]

If it’s broken he can’t play his games

Only think I can think of is lend him a 7700xt

[lone mango]

Try updating adrenalin

If already up to date then try installing a previous version if you have the installer for an older version, or ddu

[fresh timber]

That been happened

Ddu reinstall

Ddu

Reinstall

It’s been atleast a year of doing that

Maybe 2 I didn’t count

Idk why bro won’t put that 12600kg he has siting around in

[royal crow]

wha

[lone mango]

I am curious to see if that will fix it

[fresh timber]

Same

I keep telling him to do it

[lone mango]

Well

I guess he will keep dealing with it then

We love stubbornness

[icy ore]

https://x.com/wario64/status/1928494864808697892?s=46

Wario64 (@Wario64)

MultiVersus is now dead

X · May 30, 2025, 4:53 PM

Rip

[royal crow]

@worthy meteor got the keeb and it is CREAMY

[tranquil pebble]

What killed it?

[worthy meteor]

keychron is fun

[icy ore]

Lots of things, monetization issues, server issues, just not polished

Prioritized profit over good gameplay

[lone mango]

Good

One less bad game

[frigid stag]

it wasnt a bad game

i played it like 2 years ago or so

when it came out. was fun but idk what happened after w it

[tranquil pebble]

yeah but they are usually sold out on half their stuff lol

That said I do have a K5 SE

[tranquil pebble]

smh, will companies ever get the message?

[icy ore]

Nope

[frigid stag]

it wasnt that at all

game was just unbalanced and had performance issues later on apparently

[icy ore]

unbalanced and performance issues = bad gameplay

it could be fun and it was at times but just one of those things

[fresh timber]

Runs fine on my 9070xt at 720p all low

[dusk knot]

https://youtu.be/o21CDqlCSps?si=ERN7uxJDfti9-dMo

[icy ore]

[fresh timber]

Rtx pro 6000 when

[tranquil pebble]

would it not be better for them to make a good game then milk it? That sounds like a more... sane approach

[fresh timber]

I think they want quick cash rather than a steady flow

[tranquil pebble]

lol good reason to not spend money on games right at launch

[dusk knot]

The 5090 is a mid range GPU now

[fresh timber]

5080 is low end

[tranquil pebble]

Take any nvidia GPU price, divide it by 2 and that is roughly what I'd pay for it

well divide by 3 in some cases

[frigid stag]

mid range?

[fresh timber]

I’m being told my gpu sucks and I should have gotten a 5070ti

[tranquil pebble]

nvidia 12 GB GPU going for $999 is insanity

[dusk knot]

Yeah

😭

[fresh timber]

Where

[dusk knot]

Nvidia lies again about the die performance

The 5090 is part of the failure batch lol

[fresh timber]

Lllll

Failure cars

Isn’t a 9070xt technical a failure batch too

9080xt rumor

[dusk knot]

Is it MAXED out from what AMD can provide?

[fresh timber]

Theory says no

[lone mango]

Unreal devs be like

[dusk knot]

Sad

[lone mango]

Very real future

[dusk knot]

If I get into AI, I might have to get the 5080 24gb

Should be EZ to sell my card

[fresh timber]

I found a video of Lisa Su talking to congress

[dusk knot]

I would get a response in less than an hour on FBM

[fresh timber]

Nah

[lone mango]

Yeah she was talking about ai stuff

[fresh timber]

I have 28gb vram for cheaper

[fresh timber]

TLDR: stop tarrif related stuff and we can win

[upper rain]

wdym?

[dusk knot]

To locally run AI while still having great gaming.....why not

[fresh timber]

Buy a 2nd card that’s cheaper like a 3060

U can use both depending on the workload

[lone mango]

[fresh timber]

I’m using a 7700xt and 9070xt toghter rn

[dusk knot]

Huh

[upper rain]

this is why i expect a hypothetic Navi 48 refresh would be on N4X. much of US supply can be made in Arizona and the other handful of N5 family fabs we have domestically

[dusk knot]

I can....pair the Vram?

🤔

[fresh timber]

Depending on what ai, yes

I’m combining vram right now

I’m using 24gb VRAM just running a model

[lone mango]

Dont most of them have multi gpu support

[dusk knot]

Never knew that

[fresh timber]

What @fair onyx told me, text gen does but not image

Image is a lot less heavy on the vram

[lone mango]

True

[upper rain]

what's the latency like i wonder? surely pretty awful

[lone mango]

I have an upscaler and gpu usage goes to 100%

My video upscaler also uses a lot of compute

[fresh timber]

Idk, it’s running better than before

[upper rain]

once models are trained it's really easy to accelerate them. i wonder why that software doesn't use the tensor accelerators on your GPU die

[lone mango]

Rn it upscales 1080p to 4k at 6fps while my 5600/6600 was doing like 3.3

[fresh timber]

The only reason it’s running better between gpus is because the combined vram makes up for more than just latency

And if you can fit the entire model in the vram it’s just a matter of coordination I would assume

[fresh timber]

Does AMD have tensors, I thought that was a Nvidia thing

[lone mango]

Thats just a name

Marketing basically

Both have ai accelerators

[north crescent]

RX9000 has matrix acceleration cores

"Tensor" cores

[lone mango]

Yeah tensor vs matrix cores

[fresh timber]

I have some

[lone mango]

Correct

[upper rain]

didn't know they were on radeon, but yeah they have accelerators, just not Nvidia's proprietary ones

they both accelerate low precision operations tho

so somewhat interchangeable

[dusk knot]

Hmm

Theo, you need a better gpu

[upper rain]

https://tenor.com/view/minecraft-villager-staring-phoenixsc-gif-25945930

[fresh timber]

Dr

Theo gpu is slow

What do you have JP

[dusk knot]

Something slower

[fresh timber]

5080

Right

[dusk knot]

A 5080. I'll be single forever

Jk. Lol. Great card

I just need better games

[upper rain]

just play minecraft tbh

kino

[dusk knot]

Waiting for them.

[fresh timber]

Theo has 5090

Right

[dusk knot]

I've been putting a lot of time into ark: ascended lol. I prefer it over Minecraft

[upper rain]

dinosaur rust, nice

[dusk knot]

As a solo player

[dusk knot]

Yeah

....

[upper rain]

i like building and automating so i like MC

[fresh timber]

Why 5080

9070xt better

[upper rain]

he was dropped as an infant

[fresh timber]

Ooo

[fresh timber]

I figured out how ima mount 7700xt

Ima mount it on a 1x riser and just screw it into the pcie

To support it ima use my sag bracket

If that doesn’t work ima just run a rizer to the outside of my case

[upper rain]

i forget, why are you limited to x1? surely you have a secondary x8 slot you can take from?

[fresh timber]

Cause.. asus in their unlimited wisdom has it in a spot that will clash into the psu shroud

[upper rain]

ahhh that's right

casemod time

dremel that sucker

lol

[fresh timber]

Case cost too much to do that

Due to my new motherboard for my am4 accepting 4 ram sticks my main pc will be upgraded to 64gb if I’m fine with the speed decrease

I have 4 8gb sticks it will take

[upper rain]

how much of a decrease?

[fresh timber]

4000 go 3200

[lone mango]

https://youtu.be/fGn-_qj76sk

[upper rain]

on the 13600kf right?

[fresh timber]

New vid to watch while eating

[fresh timber]

Yes

[upper rain]

probably not a huge difference, but if you aren't lacking for dram in your current/near future applications i don't see any reason to bother

[frigid stag]

I regret looking at tiktok, it’s the most braindead bs

[fresh timber]

So slow

[upper rain]

spoken as someone who has 64GB of ram i barely use

[frigid stag]

Tiktok comments are genuinely room temp iq 😭

[fresh timber]

Lmfao

[upper rain]

lol, and people were telling me that 5060/ti was going to be a big uploft over last gen

[lone mango]

Refrigerator

[frigid stag]

Real

[upper rain]

gort

[frigid stag]

I’m surprised how they get so many likes for that bs. Probably bots

One Mf saying people shouldn’t have kids cause men don’t go through child brith like women do 💀

[lone mango]

Well its not super small either

[fresh timber]

8gb 5060ti

[lone mango]

23-25% uplift or something

Still not enough for these games though

[frigid stag]

Bigger than the other 50 series cards except the 4090

5090*

[lone mango]

Yeah pretty much lol

In comparison to the rest its not so bad, but still not enough

[upper rain]

15.6% in case of 5060, 16.5% in case of 5060 ti per that graph

[lone mango]

Unless you value gaming and dont play these games

[frigid stag]

They convinced us you need frame gen for path tracing at good frame rates

https://cdn.discordapp.com/attachments/602974941486514197/1378694159548022855/IMG_0339.png?ex=683d887e&is=683c36fe&hm=03341a81c229d74c75fc16e13295cd373ccf3f840446ab3dd7706675a27b5b54&

Rtx pro does it 🤔

0 frame gen needed or upscaling

[lone mango]

A 22% difference

[upper rain]

18.1% for 5060 ti, 22% for 4060 per HUB's numbers (1080p bc vram)

[lone mango]

51.2/65.8 is 0.778

[upper rain]

awful for a low end card. % compound, low end gets a bigger gen over gen uplift becausw otherwise the delta between the low and high end explodes. which is what we see here. historically low end would get like 40-60%

[upper rain]

i was comparing 5060 ti vs 4060 ti, 5060 vs 4060

[fresh timber]

Imagine if they didn’t make the new cards have less cores than the card it’s ment to replace

We would get real gains

[lone mango]

21% uplift in f1 2024

[upper rain]

18.1% for 5060 ti, 22% for 4060 per HUB's numbers (1080p bc vram)

[lone mango]

Yeah but nvidia hampered everything low end

[upper rain]

this is with their average

[fresh timber]

Is 5060ti good

[upper rain]

so because it's awful it's somehow good?

[lone mango]

Sounds about right

[upper rain]

least bad

for nvidia at least

[worthy meteor]

less awful = still awful

[upper rain]

9060 xt is probs going to smoke it

[fresh timber]

Prob

AMD really is making alot of 16gb cards

[lone mango]

What I was trying to say was that its uplift in comparison to other 50 series cards

But yeah

Still doodoo

[fresh timber]

Remember when pcie 5.0 had no use besides nvmes, now we have gpus

[upper rain]

but that's still not really correct because the difference vs high end compared to previous gens is still abysmal. better than some of the Blackwell SKUs, but i'm pretty sure 5070 ti is actually more competitive there

[fresh timber]

Shouldn’t a 9070xt compare to a 5070 cause they both techncily the 70 class of these new generations

[lone mango]

Im just saying, they literally hampered the low end

Masterful scam, nvidia

[fresh timber]

Hi

[upper rain]

mhm i agree with that

[fresh timber]

Why is this cheaper than a 1x to 1x

Just buy these 😭😭

[upper rain]

that should be 1x to 1x electrically just in a 16x physical form factor

[lone mango]

[fresh timber]

Yeah I’m aware

It’s just.. a lot cheaper

I’m buying a bunch of them

I was in the market for a 1x to 1x but it’s cheaper to get that adapter

[upper rain]

fair enough lol

https://tenor.com/view/owo-love-cat-meow-heart-gif-4559932780455511834

[lone mango]

[upper rain]

made some delicious pasta btw

i love pasta i'm probably going to die if o keep only eating it tho

[fresh timber]

Do they make 4x to 16x risers mhhh

[upper rain]

NVME to PCIE adapter

get more PCIE lanes that way

4x

vs using a 1x riser at least

[lone mango]

Welcome back 6700 xt, looks like you lost some weight (vram)

[tranquil pebble]

Why does Intel use such a bizzare way of updating iGPU drivers? lol

[fresh timber]

Omg I could do a nvme adapter

[fresh timber]

You update them?

[lone mango]

fr

[fresh timber]

When I had a Igpu I never updated one

[upper rain]

based lol

[tranquil pebble]

yes

[upper rain]

i think now that intel has a few slightly usable iGPUs it is worth updating from time to time if you have one

are there even any BMG iGPUs out yet or is it all still baby alchemist stuff?

[tranquil pebble]

because the laptop has a dGPU which depends on the iGPU working properly

[fresh timber]

My future laptop might be a Intel one

[lone mango]

baby alc

[fresh timber]

I just want thunderbolt

[lone mango]

[upper rain]

hate that for you, get better soon

[upper rain]

they have good mobile CPUs ngl, just that AMD is so cheap nowadays

[tranquil pebble]

Raptor Lake H with 4060M

[upper rain]

strix point and stuff is so cheap

[fresh timber]

Yeah, I just want thunderbolt 4 on a cheapo laptop so I can sell my 2’d desktop and plugin a gpu

[upper rain]

honestly not that bad but still i hate gaming laptops

[fresh timber]

I will run 9070xt and 7700xt and see the difference

[lone mango]

I will forever regret getting an amd laptop until I eventually get one

one day

[upper rain]

does USB 4x4 not let you do that?

[tranquil pebble]

cant complain since I got it for less than $200

[fresh timber]

You need usb4.0 or thunderbolt 3-5

[upper rain]

should match some thunderbolts iirc

[upper rain]

yeah i think some of the AMD lappies have 4.0/4x4 but idk these days

[fresh timber]

Usb4/thunderbolt main advantage in my opinion is they can use pcie lanes

[upper rain]

oh yeah that's based actually

[fresh timber]

If I fail I will take apart the laptop and plugin a nvme to pcie adapter

[tranquil pebble]

found it has a second NVMe slot so I threw in a second SSD

a 2 TB one I had laying around

[upper rain]

i nearly did this with my laptop before realizing that my CPU is actually just worse than the pentiums and i3's i could get in full office desktops for the same price as all the stuff i'd need to mod my lappy

[fresh timber]

Lmfao that’s why I wanna get a nice cpu in mines

[upper rain]

2500u baybee, we love hitting 105C at idle and never clocking above 1.5ghz

[fresh timber]

I wanna go amd but they don’t have the io I want

[upper rain]

fair tbh

[fresh timber]

Aka thunderbolt

[upper rain]

AMD io is distributed strangely sometimes bc they totally have the lanes for USB 4 but they let intel dominate those use cases instead

[fresh timber]

FR

[upper rain]

but i guess they're starting to win in laptop anyway so maybe the market doesn't care too much overall

[fresh timber]

I just wanna plug a gpu in man and replace my 2nd so I can sell it

[tranquil pebble]

specs

[upper rain]

also i love olive oil the spanish and italians are my new favorite people for that

[frigid stag]

Dont x670e have thunderbolt ?

[fresh timber]

Usb4

[tranquil pebble]

the 48 GB was an upgrade I bought too

[upper rain]

6+4?

[lone mango]

yeah

[tranquil pebble]

6p 4e yeah

[upper rain]

good stuff

is that an actual raptorlake or one of the rebranded alderlakes?

bc 6+4 is totally a SKU adl can do

[lone mango]

I have 4+8 and crap battery life I have so much regret for not getting amd

[upper rain]

ah dear, that sucks

at least it's not meteor lake

(unless it is lmfao)

[tranquil pebble]

Raptor Lake H, unaffected by the Raptor Lake issues

[upper rain]

less affected. but by a staggering margin so meh

[lone mango]

Nah 1240p

[tranquil pebble]

I did ofc update the firmware to be safe

[upper rain]

comfort yourself by knowing you are not one of the poor souls stupid or ignorant enough to drop $1500 on a meteorlake

a meteorlake that gets shredded by (then) last last gen CPUs in a lot of ways 💀

[tranquil pebble]

Pentium 4 repeat fr

[fresh timber]

Damn I have a lot of pcie

[tranquil pebble]

When Pentium 4 launched, it was slower than the Pentium 3 it replaced lol

[upper rain]

i must take my leave now, university calls me (i haven't done any of my assignments due tonight 😇😇😇)

good chatting with u all

[fresh timber]

Bro didn’t do his homework

[fresh timber]

Ru a microcenter

[dusk knot]

I'm turning 30 next year

[fresh timber]

Pushing 30 unc

[upper rain]

im going to go ted kazynski my OS should not be advertising stuff to me

thats what i get for forgetting to boot to my linux partition

[lone mango]

Make linux the first boot option

[upper rain]

i usually do, but i've been OCing a bunch so i set windows back to first boot so that i don't have to change my boot order every time i have to restart bc of a crash

forgor to swap back

[lone mango]

Oh ok

[dusk knot]

1/3 there 😔

[lone mango]

Yeah I also use windows as the sacrificial testing environment

[fresh timber]

Happy born day

[dusk knot]

Not yet

In 26 days

[fresh timber]

Oh

[slate fractal]

whats a meteorlake

[lone mango]

Core ultra 100 series

[upper rain]

when a mommy chiplet and a daddy chiplet love each other VEERRRRRYY much...

[slate fractal]

ohh

[lone mango]

https://youtu.be/ERm1StY-4uY

When $400 graphics card has the same memory bus as steam deck

[fresh timber]

Hi

[lone mango]

[mental yoke]

Need a TPM Chip for my mobo

Bought one and it was the wrong one

💀

Dont want to use fTPM for some reasons I will not disclose

[royal crow]

you can get separate tpm chips for mobos?

[mental yoke]

yes?

[royal crow]

first ive heard of it

interesting

[mental yoke]

It should be common knowledge no?

it comes with your CPU

and you can get a sperate one

w11 requires TPM

[royal crow]

have i been living under a rock or sm

[mental yoke]

probably

[mental yoke]

A lot of gamers switch from Built in TPM to a pysical TPM

If you know why lol

[royal crow]

safer ig?

lol i had no idea

[mental yoke]

no

[royal crow]

?

[mental yoke]

dw about it

[royal crow]

https://tenor.com/view/orange-cat-cat-side-eye-cat-side-eye-orange-cat-staring-gif-4749659952322810298

[mental yoke]

you can figure it out yourself

[rancid juniper]

gato

[fresh timber]

Did you get your self hardware banned

[lone mango]

[upper rain]

lol how do you even do that with windows

[fresh timber]

They ban hardware ids

[upper rain]

yah but for what

what do you have to do for microsoft to ban you

[fresh timber]

Genuinely I wonder if people bought a used pc just for it to be banned

[fresh timber]

No anti cheats

[upper rain]

ohhhhh lol

yeah that makes far more sense

[fresh timber]

I’m bored

[lone mango]

Hardware id spoofing

[fresh timber]

Spoof it before you start hacking

Easy

[frigid stag]

I get like 1/3 of that

My pc isn’t optimal

[misty sparrow]

Start your life

do the things you haven't done yet

get money

[grand kernel]

posted my new build in #build-pics

[misty sparrow]

it's pretty good, even made the build all white and case looks very unique

never really seen a case like it

[mental yoke]

I can not confirm or deny

[mental yoke]

microsoft💔

[fresh timber]

@grand kernel what case is that, I like

[grand kernel]

jonsbo d31 mesh screen white

[tranquil pebble]

a machine that truly respected its user would spoof the ID on its own every boot

or have the option to at least

[mental yoke]

i think my mbp does