Well?

I don't know anything about graphics cards, but the girl is cute.

>>716241835 (OP)
No thanks, I'm still using Intel HD graphics 4000
Cute girl though.

There's nothing worth upgrading for. I'm still using a 6700 XT I bought for the re4 remake, and an 8700k i bought in 2019. Every game runs fine.

It's on my mind but currently I don't play games that warrants all of my VRAM on my 4060 Ti. That and I haven't saved up yet, so I'm considering holding onto my current GPU until there is a new revision with 16GB VRAM at an affordable price.

>>716242629
if you got the 4060 ti i definitely dont think you should rush for an upgrade

>>716241835 (OP)
boughted the 6600 xt, not getting another gaypew until it explodes

>>716241835 (OP)
I might bite for a 9080 XT or 9090 XT. AMD-chan is cute.

NOOOOOOOOOOO!
THEY CENSORED HER ENOURMOS FUTA COCK!

Is that right side caption supposed to be censoring her gigantic, veiny, lubed up futa cock?

>>716242685
Missed this thread entirely; I have the 8GB version. I accepted my fate and live with it since gameplay is more important for me than graphics so this will do for now.

>>716241835 (OP)
Buy 9070, flash to be 9070 XT. That's the best value.

>>716242102
The stagnation consoles had on the xbone/PS4 to the XsX/PS5 finally came to PCs
there is no reason to upgrade whatsoever
games look fine on medium settings and high settings don't look that much better since you HAVE to use TAA or upscaling on the newest engines to the point the art itself in these games is made thinking on the end user using one of those
If you have anything but the lowest 30 series Nvidea card or the lowest 6000 AMD card, you are fine, just set everything to 1080p and medium and you will never feel the need to update until they force RT everything on every new game

>>716242629
Nvidea was kinda right on VRAM, you don't really need over 10 gigs... but only because every body buys Nvidea and all of their affordable cards are sub 10GB VRAM so all devs are aiming to keep Vram usage below 8GB by aggressively swapping data back and forward from the Vram and SSD
The few games that don't do this are the ones where pic related happens

>>716241835 (OP)
>bought a 1080 when it was new
>died 2 years into its life
>gpus since have had crazy prices for various reasons and I have been stuck with an expensive 1650 from 10 or 12 years ago.
Is there any decent semi-modern GPU for a 200-300 price?
I dont really play a lot of high end modern games anymore but its nice for the rare blue moon that I do, that I am at least able to run it on ultra low.

>>716243083
Kek

>>716241835 (OP)
After using the 9060XT and ditching it a few weeks later, I'm fine waiting till gpu's are actually an upgrade and worth the wait

>>716245606
Then you get aggressive texture pop, what use are more fps when half the frames look shit?

I don't see a reason to upgrade from my 1070

>>716241835 (OP)
>Polaris
No, I'm sticking with my RX 560 2GB

>>716241835 (OP)
I already bought a 7700XT and even that feels kind of unnecessary

No I snatched up a 4080 super before they went extinct
Ill consider AMD when it breaks

>>716241835 (OP)
Would

>>716241835 (OP)
>Polaris
We will never get something as good for that price. The closest is probably Intel.

>>716241835 (OP)
But I've already got the best bang for buck 7000 series card AMD-chan! It's too soon to upgrade!

>>716252286
Arguably, the RX6600 already exists.
And it runs basically everything you could want at 1080p, even modern AAA goyslop like milds, just not at ultra settings
Don't let le fine wine AMD drivers tint your glasses too rose, back in the day polaris cards performed on AAA slop just about how the RX 6000 series does today on current AAA slop.

>>716242102
>I'm still using a 6700 XT
Ive got meself new ultrawide oled monitor (shit is so cash) and it starts to fall off unfortunately. I am upgrading to 9070xt next week.
But yeah, great still great card for anything up to 1440.

>>716253518
>ultrawide
imagine falling for the meme
should have just bought a bigger monitor

>>716241835 (OP)
The 5070Ti is about the same price as the 9070XT. Unless you use Linux, there is no reason to buy it over the 5070Ti. The 9070XT is a fake MSRP and AMD didn't make enough to keep their 10% marketshare. It fell to 8% and the 9070 XT is no where to be found on the Steam survey.

>>716256701
>5070Ti is about the same price as the 9070XT
what the fuck??

>>716257621
It is because AMD just did not produce enough cards. TSMC's N4P process isn't running at full capacity, so they could secure additional short-notice capacity by paying 25% more than the long-term contract rate, but they’re choosing not to. AMD DGAF about discrete GPU market share. It is happening with Intel GPUs too. They are selling above MSRP too just because Intel is choosing not to produce enough.

>>716256701
Incidentally I just recently upgraded from a 1070 to a 5070ti and what a difference

>>716258806
I made a similarly huge jump from a 960 to a 7800XT.

>stopped using Amada Kokoro for marketing just before AMD actually got good again

RIP to a real one

But I already have a 9070 XT!

>>716241835 (OP)
I'm not retarded to buy an AMD GPU, only indians do. They have less than 8% of the market share and that 8% is mainly stupid indians and retarded linux users

>>716262092
Buying a AMD GPU cucks you out of most PY and Cuda libraries
wanna mod unity games with 3D Vision or bepinex? Cant do that, closest to it is using ZLUDAa to emulate cuda

HDR video? AMD has no alternative
Super resolution ? VSR absolutely sucks compared to nvidia alternative

Do yourself a favor and never buy an amd gpu

>>716241835 (OP)
>ever buying gAYYYmd

I'm still using my 1071 TI, surely she's of legal age by now

>>716263057
>conveniently leaves out all the affected games are dogshit

>>716264341
also the fact it's just a goytracing problem
that's dogshit tech that needs to die asap

>>716241835 (OP)
I went from RX 590 to RTX 3090 but the big boy uses too much electricity and I can't find any laptops with 24GB Vram

I'd get one with a 4080 or 5070 if it was just for games, damn those memory hungry AI models.

>>716264573
3090 Ti**

>>716241835 (OP)
i've got a 3080 and i'm sticking with it for another 5 years

>>716264573
just shell out for that $3k or w/e nvidia box
or spin up on-demand gpu cloud and host models there, it's not like it'll cost that much since it's like $1.60/hr. for a 2xA6000 setup on lambda

>>716263736
She's a classic, a vintage waifu at this point. Sure, she's not pulling triple digits in CyberSlop 2079 Ultra RT Psycho mode, but she’s still cruising through 1080p like it’s 2018. As long as she doesn’t start artifacting or pulling a thermal tantrum, let her keep living her best life.

>>716266127
>just shell out for that $3k or w/e nvidia box
Damn that's expensive. Is it even out yet?

>>716267393
i think it was supposed to come out in july so maybe? issue with the thing is it's unified memory so it's gonna be slower than just using a gpu. only benefit is it's 128gb (how much of this is actually available as vram is unclear), so in terms of price it's not awful since you'd be paying the same or more for the equivalent GPUs.

i don't really think it's intended for people wanting to run a local model, though (as no company ever wants this to happen). probably intended for dev work or toolchain development before deploy.

>>716268450
Mac Studio M3 Ultra is also an option and has 512GB of Unified Memory but that's like $15K

>>716268586
yeah, if you have a bigger budget that probably isn't an awful option although it's gonna be slow as hell compared to GPUs. sadly the optimal config is probably still just picking up 4x3090Ti (destroyed by mining) for ~3-4K and slapping them in some server. if you're somewhat privacy minded and don't plan to use it for hundreds and thousands of hours (because you are a quickshot) i think renting GPUs isn't the worst value proposition. otherwise paying for big model APIs is probably the best and highest quality option.

that said i just run nemomix 12b q4 on my 4070ti lmfao

>>716268847
Gonna be honest with you mein neger I'm mostly only using Stable Diffusion, training RVC models and the video gen one that I can't remember the name of. I want a laptop so that my family will stop nagging me about the high power bills (I don't pay for it so it's only fair). But I'd like to have the ability to use local TTS (multiple languages) and local music gen in the future when those two are available, so I don't know if going from 24GB to 12GB would be the right move, taking futureproofing into account.

And yes I know local TTS already exists but the only decent ones are English only and I'm interested in more languages. Not to mention the ones available right now are not expressive enough.
God I wish ElevenLabs models would get fucking leaked already. Or Gemini's.

>>716241835 (OP)
All anime and video game girls want to be with East Asian Jomon men only.

>>716241835 (OP)
sex

>>716263057
nobody talks about it because thats a one-guy problem not a general problem

i buyded (sic) a 6800xt a couple years ago. it'll be at least 5 years before i updoot it.

>>716264341
>I DIDNT WANT TO PLAY THAT GAME ANYWAY!