Thread 718833770

Worth the wait? Or should I bite the bullet and grab one now while prices are "low"?

>80 Super +8gb ram
>ti70 Super +8gb ram
>70 Super + 6gb ram

>>718833770 (OP)
only real oily niggas get high end cards like the 5080 or the 5090

You a real oily nigga, right OP? You roll with the hood?

Is this Nvidia's new FUD to prevent people from switching to AMD?

They just released 8GB 5060 and blacklisted people from reviewing it, as they're well aware just how terrible it is.

I'll just wait until next generation of AMD, and will buy that.
Ray traced hair isnt worth this much money

>>718833770 (OP)
Don't buy njudea
AMD isn't great either
Intel? Lol

nvidia needs to do something because the 50 series is not selling that well after the initial wave of buyers. Even in my backwater shithole I can buy a 5090 of any manufacture by now. The rest of the models are all basically forming a pyramid of shite in a warehouse.

50 series exposed AMD for having good drivers now

>>718833770 (OP)
>is the shittiest generation worth it

>>718833770 (OP)
you should wait because bad devs will make it mandatory to have 18gb ram or it crashes
eternal bloating will always make less vram obsolete, they already made indiana jones and monsterhunter 12gb doa

Super cards are usually priced the same or less than the cards they replace, so if that is true then these might be worth it.

I'm gonna be going from a 5080 to a 5090 soon because i need more vram for ai generation...

>>718835812
coomer...

>>718833770 (OP)
I still use a GTX 1060 6GB paired with a Ryzen 5 1600.

>>718835887
Pretty much. 16gb of vram was enough for static images, but now that i've been doing wan2.2 video stuff, 16gb gets filled really fast at relatively low res/16 fps. While i could commission artists to draw that stuff, some of what i generate is just too degenerate that i would like others to know about.

>>718834718
>Is this Nvidia's new FUD to prevent people from switching to AMD?
yeah bro the SUPER cards are a completely new brand FUD concept, this never happened before

>>718836291
Those Super models were "leaked" as early as 5060 just released and it they were supposed to launch in like 6 months, as far as I remember those "rumors".

>>718834718
>Is this Nvidia's new FUD to prevent people from switching to AMD?
Nobody is switching to AMD in the grand scheme of things.
The RTX5000 series has been dogshit outside of two mediocre cards (5070ti, 5090) and yet it has outsold the AMD RX9000 cards massively. The The RTX5070 (shit card) and AMD RX9070 are direct competitor cards, and the 5070 is outnumbering it 14:1 on Steam.

>>718833770 (OP)
worth waiting until the next generation
the shitty 5090 power connector has a real risk of causing a fire idk about the other cards you should have just gotten the 4090 desu

>>718836560
The 4090 has the same connector.

>>718833770 (OP)
Reminder they make 48 gig versions and you're buying overpriced cut down ones.

>>718836663
And draws about 150 Watts less.

>>718833770 (OP)

>just got a 5080

FUCK guess I’m selling and upgrading

>>718833770 (OP)
it's impossible to tell exactly what's the right move but I suspect whatever you do it will be a bad investment

>>718836085
>i generate is just too degenerate that i would like others to know about.
What do you gen?

>Ti Super
why dd they do this ? Ti and Super used to mean different things . also correct me if im wrong but Tis used to come out after the initial batch nowadays they are part of the initial lineup

i have a 3090 and i see no card to upgrade to that will give a meaningful jump without spending a shit load of money so i just stay here

>>718836370
nigger he is talking about how supers have been around for at least 4 generations.
there was ALWAYS gonna be a 50 super.

16GB minimum

>>718836541
>The RTX5000 series has been dogshit outside of two mediocre cards
In what way?
I'm in the market for a GPU right now and I can buy a 4070 for like $60 less than a 5070.

>>718838145
try not to be racist next time buddy

oh the games we play. ...

>>718833770 (OP)
You'll feel like a real idiot when 60 series launches and comes with some new exclusive DLSS5 feature.

>>718838285
outdated

>>718833770 (OP)
>Nvidia

>>718834718
Probably not, this is typical Nvidia, release garbage cards, then release the good cards. Repeat.

>>718838267
try to be less of a retarded nigger instead.
nigger isn't a race, its a subhuman choice.

>>718833770 (OP)
the 5080 super leaks made me more interested in this series with 24gb vram but it depends on the price
I can see them going for like $1600 msrp which will probably mean $2100+ in reality
might as well just get a 5090

>>718833770 (OP)
I'm going to tell you a secret the next gen of consoles will have 32gb of ram and 28gb will be available for vram.
So do not buy anything that has less vram

>>718838997
Super cards are meant to be a refresh and substitute for the existing cards, not a higher tier.

>>718839071
Pffft yeah sure

>>718838050
fellow 3090 chad here. If 5080 super is going to have 24gbs of vram then I feel like it's worth it for what I presume will be 1k dollary-doos. DLSS 2 hasn't been great in modern titles and I definitely want multi frame gen, or even just regular frame gen. FSR 3 frame gen has been an okay band-aid for the 3090's lack of ability to use nvidia's own version, but I'd like to utilize proper frame gen for once.

>>718839126
we'll see

>>718839330
See you next year in March

>>718839545
You can use dlss4 with a 3090.

>>718837936
Stuff like this:
https://files.catbox.moe/u10nn8.mp4

>>718839765
what model did you use for the video?

>>718839734
PS5 struggled with tons of games, PS6 won't be any different mate

>>718839826
https://github.com/Zuntan03/EasyWan22

Using this basically at default. More vram you have the better.
Here's that generation but upscaled/frame interpelated.
https://files.catbox.moe/e6hwp4.mp4
I prefer lower fps however though, feels less like ai.

Someone give me a good image upscaling model for real life pictures using pytorch.

>>718839893
cool, thanks anon
I only have 16gb but I can try with runpod or something like that

>>718839976
I'm running that at 16gb of vram 5080 (32gb system ram). It's why i want more vram so i can do longer/higher base res.

>>718839126
There is absolutely no way in hell these cards are going to be anything approaching normally priced, not a beginning of a chance

>>718833770 (OP)
Since every new gen now gives you software advantage locked to it, it's not wise to buy super editions anymore. You either buy yourself new gpu on gen release or wait till next one.

>>718839126
>are meant to be a refresh and substitute for the existing cards
They never are, always priced above original cards.

>>718833770 (OP)
>Nvidia

>>718839854
Not my point

>>718840026
don't know if this will help but you don't need more vram, it stitches the videos together that it generates automatically, haven't tried it myself
https://www.youtube.com/watch?v=ImJ32AlnM3A

I laugh at these threads because you fucking idiots buy these retarded 5070tis and 5080s and 5090s that don't fucking work on Windows or Linux. I mean literally don't fucking work. Black screens, corruption, HDR issues, pytorch errors, I already ran ddu and did all the steps but it leads back to the NVIDIA .DLL driver or NV PKG error when you look at crash tools.

>>718840508
Yeah the choice is miss out on the next software feature or deal with gimped vram.
Thanks nvidia

>>718840727
Nigga the Nvidia software doesn't fucking work at all the hardware is joke unless you're running FreeBSD or some CUDA based operation that doesn't depend on Windows or Linux.

>>718840669
skill issue

>>718840669
Still not buying AyyyMD

>>718840787
>>718840790
>>718840771

>>718840790
>>718840787
I just sold my 5080 just because it literally doesn't fucking work. You guys and youtubers told me it had better drivers than AMD but its fucking lie in 2025.

In the end 4090 sistahs won bigly

Will never need to upgrade until the late 2020s as long as you stick to 1440p

I'd rather use RADV than Windows.

>>718840841
okay, cool

>>718833770 (OP)
>tfw i got my 4090 in early 2023 for MSRP

>>718839758
What the 3090 got from that weird update is hardly DLSS 4. It's just a more stable version of DLSS 2. Not only that, but from last I checked, you have to manually inject it from their app into games. On top of that, not every game is supported. Why they even bothered to say that the 3000/2000 series are getting DLSS 4 support is beyond me. I imagine the 4000 series sees better improvement from the update, but the 3090 certainly didn't.

>>718838145
>nigger he is talking about how supers have been around for at least 4 generations.
3000 series had no Supers

>>718840669
Works on my machine

>>718840954
I haven't seen malwarebytes in years
had no idea it was still around

>>718841020
The upscaling works, is slower using the transformer model but it works, the framegen doesn't.

>>718840841
real poorfag energy coming from this guy over here

>>718841020
>dlss5
>dlss6
>dlss7
>pathtracing
>walkwaytracing
>hairtracing
im tired boss

yes. get gpus with more vram so nigger devs can brutforce even more of their lazy shit

>>718841228
>devs
it's gonna be ai telling them what to do
that's what's being forced at me where I'm working anyway, it's 'ai first' bullshit
retarded product managers
I don't give a fuck though at this point, it's gonna all crash and burn

>>718841192
WE MUST GO FURTHER BEYOND AND BREAK THE LIMITS OF REALITY

>>718841020
Dunno what you're talking about. The most important part of the DLSS technology suite is the upscaler, and the DLSS4 upscaler works even on RTX2000 cards.
Sure, stuff like DLSS4RR runs poorly on older RTX cards but that was always a very niche thing not nearly as relevant as upscaling.

>>718842468
DLSS 4 upscaling also runs poorly on older RTX GPUs when compared to CNN model

>>718841046
Why would it need supers when its already super.

>>718834718
8bg ram is useless nowdays. 12 and up, and 12 is like entry level, with 16 being much preferable but 24 being the good spot.

>>718842782
Eh, runs about 10% worse which you can just offset by using a lower preset. CNN model Quality looks like T-model Performance.
It doesn't run nearly as badly on older cards as RR does.

>>718844008
>runs about 10% worse
That's not how it works.

On my now former 2070S (I switched to 9070XT in May, but I still experienced DLSS4):
At 4K, DLSS2/3 in Performance mode takes about 2.7 ms
4K DLSS4 in Performance mode takes 3 ms more.

That's over twice as expensive.

When going from CNN to Transformer, the framerate would go from
144 fps (6.94 ms) to 100 fps (9.94)
120 fps (8.33 ms) to 88 fps (11.33 ms)
60 fps (16.67 ms) to 50 fps (19.67 ms)

>>718833770 (OP)
Do you know if this version will cause house fires too? Genuine question becasue it's my main reason to not buy a new PC right now.

>>718846167
What an insane way to live your life.

>>718840540
Maybe it's retailers reducing prices of the original ones so they aren't sitting on shelves alongside the super

>>718846290
The thing is melting cables and the board is designed to not care at all until it burns.

>>718833770 (OP)
>>718833770 (OP)
>>718833770 (OP)
>>718833770 (OP)
>8gb of ram
No

>>718846167
12vhpwr is faulty standard to begin with.
thats why i refuse to buy any nvidia gpu and stick with 9070xt triple 8 pin.

>>718846690
OP probably meant 8GB on top of whatever the existing cards have so a 5080 super/5070ti super would have 24GB in total.

>>718846716
Yeah apparently a 9070xt with that connector burnt down recently

>>718844867
4xxx/5xxx have a lesser performance hit than 2xxx/3xxx cards with transformer btw

>>718839765
an iceberg died for this

>>718833770 (OP)
physx?

>>718846986
omg NOBODY CARES

>>718846924
Yeah, because they support new instructions.

FSR4 also runs on non-RDNA4 GPUs, when emulating FP8 through FP16, with performance hit, which is actually comparable to DLSS4 on pre-RTX 40 series performance hit

>>718847120
i care, your statement is factually incorrect

>>718833770 (OP)
Super cards aren't going to be available at msrp, you'll be waiting long enough you might as well wait for the 60 series.
There's still a pretty strong demand for cards with enough vram to be useful for ai, come back in 4-6years and you'll be able to get as much vram as you want for cheap.
If you're on Turing or Ampere (other than maybe the 3090 or 3080ti) there's no reason to not upgrade
>but prices might get even better
Things are just as likely to reverse and go back up.
>>718834843
>>718834882
You can't afford any of these anyways
>>718835148
>50 series is not selling that well after the initial wave of buyers
delusion

arent the 40 series cards better?

>>718847219
>because they support new instructions.
*because they don't support new instructions

>>718846881
Yup it affects Sapphire & Asrock. Doesn't affect Powercolor.

>>718839545
>If 5080 super is going to have 24gbs of vram then I feel like it's worth it for what I presume will be 1k dollary-doos.
yea but what are the odds regular people can get it at that price? they will be scalped to hell and back so ill probably have to wait anyway

>>718833770 (OP)
AI bubble is going to pop and only retards by niggerman 3 and god of s o i we wuz Kang in Egypt edition so who cares. 16GB is plenty for the next 5+ years and not even a 5090 can run modern games at 4k 90-120fps.

>>718847315
Are you retarded?
In what ways has there been a regression with blackwell?

Remember people only buy AMD (lol intel) out of spite because they to "stick it to the man" by supporting factually inferior products and holding back videogame development to "own" people on the internet. The damage AMD has done to the industry and how many years it has set back adoption of technologies that would make games greater than ever is devastating.

>>718846954
I hope i live long enough where this becomes interactive real time.

>>718847509
RADV

>>718847485
idk man i have a 4090 and it seems better than almost all of the 50 series cards
even im confused as to why the new cards arent better

>>718847315
AMD has better drivers on Windows & Linux than Nvidia now. The equivalent Nvidia card is always worse than AMD.

>>718844867
>When going from CNN to Transformer, the framerate would go from
>144 fps (6.94 ms) to 100 fps (9.94)
>120 fps (8.33 ms) to 88 fps (11.33 ms)
>60 fps (16.67 ms) to 50 fps (19.67 ms)
That's ray reconstruction you fucking retard, transformer dlss is basically free even on turing and ampere

>>718847634
Shut the fuck up, retard.

I'm talking about DLSS Upscaling, you illiterate fuck, using my own fucking data

>>718847665
Well your data is wrong, transformer dlss isn't that heavy on older cards.

>>718847571
RADV doesn't automagically fix hardware that is fundamentally broken.

>>718847718
NTA but it absolute is heavier than CNN.

>>718847608
>5070 vs 9070
Yeah okay try a fair comparison next time.

>>718847718
Consider jumping off a bridge, retard.

>>718847315
More like there's barely any performance improvements so people aren't gonna upgrade from 40 series
inb4 multi frame gen

>>718847718
>>718847787
Preset K is Transformer model btw, retard

>>718847120
retarded shill, kys

>>718847509
Didn't nvidia drivers shat themselves earlier this year

>>718847871
they still do

>>718833770 (OP)
For what? What games require it?

>>718847787
>>718847845
this ran amazingly on my 3070

>>718833770 (OP)
>5000 Super series
you can't force 16x anisotropic filtering with that POS, so a lot of games are a blurry mess forever, also no physx

>>718847742
>>718847787
A 30%/40% regression is what you'd see with ray reconstruction, so it's either that or you changed another settings somewhere or something. Just the dlss you'd see a nonexistent or small difference, less than 10%
>>718847783
It is a fair comparison the cards are equivalents in raster. What he's posting is just bullshit.

>>718833770 (OP)
if i learned anything from the 4000 series Supers, it is that the Super version of the 4070 Ti was not worth the wait, I should've just bought a regular 4070 Ti and enjoyed it all that time

>>718848047
>so it's either that or you changed another settings somewhere or something.
How fucking dumb are you?

>>718847608
This guy is a mentally ill /pcbg/ spammer btw

>>718848047
>It is a fair comparison the cards are equivalents in raster.
Except they are not. DLSS is superior to native so you should be accounting for at least +30% performance for nivida. Turning off dlsss for comparisons is just showing AMD bias.

>>718840669
DLSS 4 makes it all worth it mate, try playing any UE5 without DLSS and you know the pain

>>718848076
There's nothing to enjoy about the 50 series at the moment so you might as well see if the super cards are worth a shit.

>>718847485
>In what ways has there been a regression with blackwell?
1. no physx support
2. no texture filtering override support

>>718848076
didn't regular 4070 ti have just 12GB?

>>718848246
>useless features
Do you bitch your cpu no longer supports 3DNow! ?

>>718848269
>12gb vs 16gb
irrelevant

>>718848076
>Frog
>Retarded
Every time.

>>718848331
>I don't neeeeed compatibility and preservation
fuck off faggot

>>718848103
How fucking dumb are you to claim a 40% regression with transformer dlss? Literally show me anyone else on the planet claiming something similar with proof.
If I cared more I would go grab the 2070 from my garage and show you what a retard you are. 40% is the cost for ray reconstruction on Turing. It's either that or you changed a setting or something.
>>718848269
It did, but you'd never run into issues with that outside of native 4k/4k dlss quality + framegen.

>>718848246
>NOOOOO EVERY GAME MUST HAVE 60 YEARS OF PERFECT SUPPORT AND COMPATIBILITY
kys zoomer

>>718848505
>Literally show me anyone else on the planet claiming something similar with proof.
Probably because I was the only person on earth to care about this shit? My master thesis was literally about upscaling.
>If I cared more I would go grab the 2070 from my garage
Do it, faggot.
>ray reconstruction on Turing. It's either that or you changed a setting or something.
In a game that doesn't even support ray tracing, great job, retard.

>>718833770 (OP)
More VRAM just means genAI faggots will want them even more.
Nvidia will no doubt raise the MSRP and once again you will have no chance to get the cards at that price anyway.

>>718837242
And the only 5090 to get if you care about its longevity is another 500-1000 bucks, since it's the only one with per-pin power monitoring.
Very cool Nvidia, disregarding years of power standards to cut costs by a few cents, and forcing the consumer to actively monitor their hardware at all times, just to make sure it doesn't turn their house into an inferno at any given moment of the day. Fuck the previously safe cables, just put all that power into one to save $0.47 per GPU that you won't even sell 1 million of, where if the angle of the cable is off by a decimal point of a degree and not perfectly flush it turns into a blaze.
Neat!

>>718848571
neck yourself shill

Bizarre that none of the board partners breaks ranks and just makes a 5090 with 4 8 pin connections. You could charge 500 more for a model like that.

>>718848635
>Probably because I was the only person on earth to care about this shit? My master thesis was literally about upscaling.
People have tested it, the gap you're describing literally doesn't exist.

>>718848571
>>718848331
get a load of this bootlicking faggot

>>718848749
That's one way to make sure nV won't call you their "partner" anymore.

>>718836076
>Ryzen 5 1600.
I have that CPU and man it's showing its age. I'm looking for a 5700X now during the next BF

>>718848905
Your picture literally shows that Transformer is at least 1.08 ms more expensive than CNN on 3060.
144 fps would go down to 124 fps.

And there's been 2 years difference between Turing and Ampere.

>>718848990
Its fine, just ask XFX.

>>718849228
>It's fine, just ask EVGA RiP
Fuck nV and their total control fetish.

>>718848905
>>718849185
And your retarded picture is missing the information on the resolution.

https://www.techpowerup.com/review/nvidia-dlss-4-transformers-image-quality/
>With native TAA at 1080p and "Ultra" RT preset the game runs at 25 FPS (base performance).
Here, you fucking retard. I was talking about 4K, not poverty resolution like 1080p.

>>718849185
>>718849470
I'm not giving you another (you) or actual deep dive on why you're retarded, but do know that you're really really retarded.
You shouldn't talk about things you don't understand

>>718849470
Nobody uses - or cares - about 4k.

>>718849783
Feel free to figure out on your own what the performance impact of DLSS is in each major resolution, compared to native INPUT resolution, like native 1080p to 4K DLSS Performance (1080p input), retard.

>>718849798
Literally the standard output for this generation.

>>718849783
>>718849990
Also just mentioning the obvious fact, which might not be too obvious to you.

Upscaling gets more demanding the higher the output resolution is.

ITT seething AMPOORS.

>>718839734
But next-gen consoles will be coming 2027

>>718847247
nobody cares
nobody is buying the 5070 with 12gb and the ti and 5080 are priced absurdly higher than it with only 16gb, so people aint buying it too.
stores are stuck with unsold merchandise so they wont try this shit with super.

>>718833770 (OP)
don't supers have lower cuda cores than TI? you're kind of cucked either way on vram or cores.

>>718850869
>nobody is buying the 5070 with 12gb
Best selling card this gen
>stuck with unsold merchandise
Market share went up with blackwell, that's at minimum as good as any anecdotal evidence you have.

I am considering a 5070 Super. I've survived with a 2070 for so long that I think I could survive with a 5070 Super just as long.

>>718833770 (OP)
I'm only getting a card if it has 48GB of VRAM

>>718841192
so fucking tired you have no idea

no 72gb vram, no buy

>>718851215
This

My 3060 runs every game I want just fine. I don't need RTX bullshit, I need vram for running my own AI.

>>718851215
So why didn't you buy RTX6000 Ada?

>>718834718
Only thing thats is preventing people to switch to amd is amd

>>718851378
there's literally nothing wrong about 9070XT

>>718851320
>I need vram for running my own AI.
There's a card for you, it's called the 4090/5090
>>718851378
AMD isn't incompetent or anything, they just don't care about desktop users.

>>718851154
I'm using a 1050ti and thinking about the 5080 super. Benchmark says it's 1,000% faster but my 1050ti only takes 75 watts of power and the 5080 is like 450 watts or something so I'm sitting here wondering how the 5080 is going to effect my power bill on days I come home from work and just browse the internet and don't do anything fancy.

>>718851461
If you slap a waterblock on it then yes.

how much worse than a 4090 will a 5080 super be

>>718851461
It doesn't even support the upcoming DX12 extensions (unlike blackwell) so it is already obsolete.

still waiting and saving money for this
i hope it gets released come christmas

Sign me up for fat RDNA5 when it releases.

I've actually tested the frame difference cost by different DLSS presets because 2 anons ITT were arguing about it and I get curious.

I've done this test on Cyberpunk 2077 with RTX 3050 on 1080P and my results are(no ray tracing options were enabled):

Latest preset:63 FPS average on builtin benchmark
Preset E:68 FPS average on builtin benchmark

Looks like the anon who said that transformer model has a high cost is right. It's very possible that on a higher resolution and a higher framerate the transformer model could cost a lot of FPS.

nope, still not upgrading my 1080 ti

>>718851509
4090/5090 don't have enough VRAM you have to get into the enterprise end of the market instead, and then you're paying $30,000 for your AI card instead of $3000

I for one, can't wait for the chinks to obliterate Nvidia's monopoly.

What advice do you have for a guy who wants to build a new reliable PC with a decent graphics card who doesn't care about having the latest and very best/most expensive possible option?
All I ever see in these threads is "should I get the current BEST card or wait for the newest card to immediately buy that newest/BEST card"
I just want something decent that won't catch fire and is actually available to buy without fighting 1,000,000 aisloppers to get my hands on it.

>>718851534
Anon, 5080 is a full GB203 die already, 5080S will only have more VRAM and maybe slightly higher TBP.
So don't expect more than 5% more performance with the refresh.

>>718851514
It might still use a little more, but when it's downclocked just browsing, it will probably be negligible

>>718851610
UDNA will be the next AyyMD GPU architecture.

>>718851683
Chinese are currently soldering more vram onto high end nvidia cards because doing so is 10x cheaper than buying the enterprise stuff.

There is no good reason for the prices nvidia is charging besides absolute greed.

>>718851736
Thanks, I was always curious about that.

>>718851792
Well that then. AMD and sensible naming shall never meet.

>>718851719
buy a used 1080 ti

>>718851868
makes more sense than Nvidia's architecture names like Epstein, Stalin, Darwin

>>718852005
They're running out of 19th century computer geeks to name them after lmao.

>>718851868
It's because they are merging their compute and their gaming architectures. They split CGN Polaris, Vega into RDNA and CDNA but they fucked up and they are reversing back to single architecture. I hope it's not trash my 7900XT is getting old.

>>718852181
>my 7900XT is getting old
FSR4 is getting backported to older architectures and they just fucked up and accidentally released FSR4 source code,

>>718851879
used is no longer an option because [used] used to mean a dude who plays videogames wanted an upgrade and is selling you his old card. today [used] means the card has been red hot day and night for a year because of some crypto ai jeet milking it for all its worth

>>718852345
It's too heavy to be worth using, that's why they didn't do it themselves
Also I think it was only an unfinished "lite" version that was leaked, not full fsr4

>>718833770 (OP)
>higher tdps

>>718852396
cryptonerds underclock them

>>718852345
I doubt it will be any good.
From what I gathered they tried using INT8 instead of FP8 but it didn't work out.
I want to upgrade to 4K OLED in the future so I might change the GPU as well.

>>718835308
I thought the shittiest generation was the 4000s.

I dabble in stablediffusion every now and again. I never end up posting the pictures anywhere. Is it viable to sell these pictures somehow or make money off of them? I feel like since everyone has access to this kind of stuff it's value is lessened.

>>718852759
what's so shit about it?

>>718840508
But what if you didn't buy the new gen yet ? Then you'll get new features by buying even a non super card from the new gen

>>718840845
If I could get a 4090 for less than 1200 USD I would just bite the bullet. But they're expensive even used which is such bullshit

>>718852434
They will be selling RDNA 3 based apus for years so they will put the effort it. For once radeon isn't going turbo retard and announcing this long before it is ready.

>>718852849
spicy connectors

>>718853380
not exclusive to 40 series

>>718853738
But it started with them, hence why they were shitty "gonna burn your house down"

>>718852396
silicone degrades with temperature fluctuation, the gamer shutting off his pc day and night does more damage to their card than any miner could

>>718853852
current and voltage also cause degradation though.

>>718853852
>silicone
kek

>>718853798
User error.

>>718835796
>Master race is starving for new cards with more vram and just won't buy current cards in meaningful numbers.
>Guaranteed opportunity for Nvidia to milk some extra dollars.
Anon...

>>718847485
Stability,
the last driver without any issues is 566.36, which was also the last release before blackwell. it's been almost 9 months and the black-screen issue is still not fixed. There is also issues with voltage spikes causing hotspot temps to get to the point that it triggers auto restarts.
And many more issues that were not present on 4000 series cards.

https://www.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/573126/geforce-grd-58108-feedback-thread-released-81925/

It's at the point where AMD drivers are more stable.

What is the answer from AMD... b-bros?

>>718833770 (OP)
If it really launches end of this year or January 26 then I might be getting one. Especially if they've finally upped the memory.

If I have a 2070, what card should I be looking for to get my money's worth? Nvidia or AMD, don't care

>>718854240
Nothing, their cards are selling for $100-$150 over MSRP.
Also, the 9070 uses GDDR6 not 7 so they can't upgrade the memory capacity without doing a double-sided PCB or reworking the architecture. Other than that the RDNA 5 GDDR7 refresh is due next year around Q2

>>718851719
7800 XT great price to performance ratio, 3060 Ti is also good value but you will probably have to buy used

Whens the supposed release?

>>718854642
A used 3090.

>>718854378
No you won't.
24gb is enough to run many AI models, these will be bot'ed and scalped to shit, first by retailers then shitskins on ebay. With 24gb the 5080 will be capable of running just about any model the 4090 can just a few % slower, chinks will swarm the supply like the locusts they are.

>>718854014
Not buying you a jacket.

>>718854813
end of q4, no set date other than "before the holidays"

>>718854642
5060ti, 5070, 5070ti
If you're in the United States AMD isn't even worth looking at.

>>718855005
>before the holidays
Lmao so that nvidia can slap on the holiday tax i.e. 30% more

Fuck sake

>>718854642
What is your monitors res? I would avoid anything with less than 12gb VRAM at this point if you plan on keeping it for as long as you did your 2070

>>718839765
I wish I could unsee this

>>718855172
Pixio PX7Prime res is 2560 x 1440 at 165hz
I'm due a makeover since my Ryz 7 2700X is also showing it's age

>>718855061
nigga, they will slap more than 30% just because the goyem will guy it. The 5080 is xx70 class silicon selling for 1300, which is 300 more than the already scalper tier 1k msrp for a midrange die

https://www.youtube.com/watch?v=2tJpe3Dk7Ko

>>718834718
>Is this Nvidia's new FUD to prevent people from switching to AMD?
AMD barely has 1/6th of all GPU sales (excluding consoles because consumers don't really get to choose the hardware for those), this is just nvidia's usual tactic of taking all the bottom-binned crap lying around and trimming the specs down to fill in whatever price points they haven't occupied yet. I'm not even sure why you think nvidia would be worried about people switching to AMD, they've long since built up their brand as being the only "acceptable" one for PC gaming and every retard on the internet has spent the past decade and a half chanting the "lol nodrivers/nofeatures" mantra as if that's stayed true the entire time, turning people even further away from AMD cards no matter what.
Besides, the nvidia and AMD CEOs are cousins who both operate under the wishes of the chinese government. There hasn't been any actual competition between the two companies for decades, just a bunch of theatrics performed by both parties to fuck people over with a lose-lose situation if they want their graphics cards to go back to being normal again.

>>718855464
>gAMDers nexus
yeah ok

>>718851683
Same here

is a 5060ti going at 450 euros good?

>>718851719
5070 Ti but expect to learn why everyone eventually converges to waiting for a new gen and buying the most expensive card

5080 super is gonna be like 1.7k

>>718855893
Not him, but who does that when a new card is 2000+ yurodollars? I'm not going to spend 10-15 days wages pre-tax on a fucking card.

>>718854642
9070XT

>>718833770 (OP)
Wait for the 5070Ti super

Might as well just grab a 5090 honestly, they are plentiful now and 32gb vram will have you set nice and comfy for a while

>>718856174
>yeah so go ahead and set your house on fire purposefully

>>718855549
go suck leather jacket mans rice dick somewhere else

3080 to 5080S soon

>>718852396
crypto shit has been dead for years now.

>>718856209
don't forget paying 2.8k for the privilege.

Should have bought the RTX 4070

>>718851815
Keep in mind that out of the box, most GPU fans don't even turn on until you're playing a game so it shouldn't be that much power draw.

>>718855790
not really, Some of the 5070's drop to 499 occasionally and the 16gb 9060 XT is less than that

>>718856341
Yeah, I know your brown faveloid ass can't buy it, I was addressing the white men of good characted who have jobs that allow them to afford buying computer hardware

I'm not upgrading from my 3070, literally everything works.

>>718856564
>just buy bread for $200
>Rippoff? lol, you're poor!
You don't know the difference between "you're" and "your", you have no right calling anyone but yourself brown

>>718853852
Except miners wear the fuck out the VRAMs, they aren't supposed to be extressed like that. They end up all burned and are the main point of failure after coolers.

>>718853852
Most people that actually uses PCs don't shut off their PC

>>718857171
then it's a good thing mining has been dead for years now. 40xx cards on the used market have not been mined on.

AMDlets spreading FUD again.

>>718856013
People didn't use to but we're currently (finally) transitioning (sisters) to 4K and away from rasterized rendering and the top end are the only cards that can do that.
Yeah maybe they can technically make the 6070 so that it will be able to reach the target of 60fps pre-framegen at 4K with path tracing. But why would they when AMD can't compete and force them to? Nvidia will be milking this for as long as they can.