AMD won the GPU war with this graphics card. You don't need anything better. If someone tells you that you do they have a insidious agenda.

>>105670425 (OP)
AMD still lost tho

>>105670425 (OP)
Ummm, no. AMD didn't do anything. Nvidia still controls the overwhelming majority of the marketshare and mindshare in the discrete space. This is despite them not giving a flying crap post-Ampere. They know they got their legions of paypigs willing to eat-up their yield rejects.
5060 and 5070 families are outselling the RX 9060 and 9070. Paypigs and impressionable-type are convinced that silly gimmicks are worth any premium. You need to get the Nvidia brand you will look like a loser in the /v/tard circle.

>>105670425 (OP)
I do need 182% better performance bra

>>105670638
the math behind the 5090 makes no sense. its 2x the fps of a 7900xtx but 4-5x the price of one?

also a 5080 is still like 2x the price of a $650 9070xt

Intel will be the final winner in the GPU war

>>105670425 (OP)
I just wish kernel drivers wouldn't be so ass rn

>>105670425 (OP)
yes a 300w card priced at €750, they won

>>105670671
That's always the case with new cards, just give it some months and they'll be the best drivers

>>105670671
>>105671662
yeah currently nvidia is broken on windows if you own blackwell. its the price you pay for bleeding edge.

>>105671739
Idk why people buy on release, if it hasn't been out for at least 6 months you don't know about any potential issues

>>105671662
>>105672021
This Pageflip issue I have happens on older cards too now.

>no DLAA
its worthless

It's true

>>105672066
Idk what issue that is, my cards works fine, I don't buy nVidia though

>>105672131
AMD issue. Supposedly, the gpu restarts but the DE doesn't refresh so the screen freezes. Have to restart the display manager from tty. https://gitlab.freedesktop.org/drm/amd/-/issues/?sort=created_date&state=opened&search=pageflip&first_page_size=20

OP needs help
>>>/v/713247638

>>105670425 (OP)
That's cool; I'm still waiting for Intel Arc™ B770.

amd won the gpu war for me because of this apu

>>105672171
Haven't had that issue on dwm or Hyprland. I do know there is an issue that causes the driver to use the incorrect clocks for some AMD cards and you have to set them yourself to fix it which is what I am doing. But looking at it I think my work laptop has this issue, it runs Ubuntu.

XT is bloat. Embrace 9070 effeciency.

>>105670425 (OP)
People use their GPUs for more than just (rasterized) gaming now.

>>105670638
...you mean 82% better performance?
this isn't how you read this graph, how fucking retarded are nvidiots?

>>105672111
fsr 3/4 native aa exists, retard.

>>105672267
SAAR DO NOT REDEEM THE GPU SAAAR
you're brown

>>105672113
>no high end to compete with 4090 and 5090, and that's where big money is
>low end still dominated by 3060 and shitty prebuilds
>9070XT is the only good card they have

Like seriously, if 9070XT was 550$ and 9060XT 16gb was 300 they would have a chance. As it is now nothing will change.

But 9070XT gets sold out anyways so they are selling everything they produce.

>>105673533
AMD is on par with RT now too. It's only PT where they are behind and Redstone is supposed to change that. Not like it matters AMD needs to be way cheaper to beat Nvidia brand.

>>105670425 (OP)
It lacks the sweet CUDA cores, sucks at ray tracing and runs too hot.

>>105673982
>AMD is on par with RT now too
Only in light, console-tier RT settings.
The 9070XT's RT performance isn't even better than the 7900XTX's.
The fastest 9070XT in the world is barely better than a Redditor's stock 5070Ti in RT.

>>105673533
ML jeets are not people

>>105670425 (OP)
Are AMD GPUs good for Blender yet?

>>105674109
Nvidia crashes during workflow. AMD is stable.

>>105674061
nah

>>105670425 (OP)
>8% marketshare

>>105670425 (OP)
It's way too slow for me. I have a 4090 from 2022 in my main desktop. My old 3080 Ti is in my home server and I kind of wanted something better there, but the 9070 XT isn't even much of a leap from that, it's not worth bothering with and not worth the price that it's actually being sold for (like 750-800EUR in my area).

>>105672484
Idk I'm relattively new on AMD and I'm on KDE. Some people say it doesn't have that issue on Xorg but...Xorg doesn't work for me. It's laggy when I move the mouse.

>>105674061
Radeon GPUs could run CUDA shit, they just legally aren't allowed to.

>>105674273
So you're saying Radeon GPU can't support CUDA.

>>105672175
based hero. i am a fan of OP

>>105674280
They can run CUDA code but doing so violates CUDA license. It's like how they can't do HDMI 2.1 on Linux because it violates HDMI Forum license. Nvidia bypasses this by owning CUDA and storing their shit in proprietary blobs.

>>105674334
and thats why Nvidia will always be broken on Wayland too.

patiently waiting for a 9060 nitro of regular size (193mm max)

>>105674334
its the same shit from taiwan anyway. cant you just flash nvidia software onto it?

>>105674334
>They can run CUDA code
AMD cards can't run CUDA code by itself, it needs a translation layer that is outlawed.

>>105674271
I've been using AMD on both Xorg and on Wayland and haven't encountered the problem. It might be dependent on your setup or how your distro compiles the kernel drivers. I'm using Gentoo and haven't had any issues except for the clock issue.

>>105673967
>>no high end to compete with 4090 and 5090, and that's where big money is
Wrong, it what was use to be "Mid-range" that made all of the bank via sheer volume.
High-end has always been halo market that barely pays itself for R&D costs.

>>105673533
Nah, the vast majority of gayming-types use their SKUs for rasterization, even with SKUs with shitty RT. They turn that off as soon as the novelty wears off as it is not worth the performance cost.

>>105674061
"CUDA cores" is just marketing speak for shaders

>>105674613
Somehow the issue is gone if I run glxgears so it's something with opengl and drivers.

>>105674681
It's not about games anymore AI farms are buying tons of high end cards.

>>105670660
I'm not poor, do not care about price
also can't do any AI shit with AMD
>but sars AMD is a very good value like android sars!
don't care!

drivers after 25.4.1 are still broken in the game I'm playing tho

>>105674076
>Only in light, console-tier RT settings.

That is 95% of RT in games. AMD is only bad in path tracing in few Nvidia sponsored titles. Mostly because they are still lacking ray reconstruction, OMM and SER.

We are halfway through the year and literally the only game like this is new Doom with PT patch. Hilariously without PT AMD is actually faster there.

>>105673967
Actually you should imagine this being no luxury consumer GPU because they have a big bucks GPU for AI farms. Or do you thik the 5090 is the Best GPU that nvidia can make?

>>105674280
They can use HIP on ROCM to run cuda code. AMD is moving to catch nvidia fast.

>>105670425 (OP)
Lmao. 20 years now that amd is being destroyed on blender3d. People who buy amd cards aren't doing much with their computer; a little gaming and that is all. You are wasting your time.

>>105673855
They don't call us nvidiots for no reason

>>105676198
Wrong, they are getting dedicated SKUs (loaded with VRAM capacity) not shitty gaymng SKUs (intentionally gimped on VRAM capacity)
The simple truth is that gayming market isn't that lucrative anymore. It is little more than dumping ground for sub-par yields that couldn't make the enterprise cut.

>>105676780
Protip: all gayming SKUs are for media consumption.
>20 years now that amd is being destroyed on blender3d.
Nah, it has respectful performance. If you are serious about 3D professional graphics. You aren't wasting time with shitty gayming SKUs anyway for either camp.

AMD did good this generation. They are pretty much exactly on par with Nvidia this gen. I hope they keep it up.

>>105670425 (OP)
I don't need better performance but I do need drivers that don't constantly shit the bed in VR

>>105670425 (OP)
>costs just as much as the 5070Ti, but with less features and support
We all lost, the jews won

I don't see why not now that nvidia dropped physx compatibility. Nvidia *just work*ing was the whole reason I choose them in the first place. And I play a lot of old games

>>105676339
are you sure about that?

>>105677080
Nvidia is poopie yucky and just keeps getting worse. If they circumcize their cards any more they will be trans next generation

>>105677480
why choose fedora over arch?

>>105677535
i have everything i need on fedora

>>105673967
>As it is now nothing will change.
but things are changing, they are losing even more money now

>>105670425 (OP)
>You don't need anything better
I don't even need it. That is Overkill

>>105677480
Oh wow, you opened a window! Surely that would be impossible without a mighty AMD GPU?
Show us tokens/second, you coward.

sidious?

>>105670425 (OP)
baseded

>>105670425 (OP)
it's wild how powerful tech youtubers are when it comes to influencing public opinion. this card is more expensive, performs worse and has far less vram than a 2 year old 7900XTX. it is quite dramatically outperformed by a 5070ti (a roughly equivalently priced card) in a 20 game average at 1440p, the card they all said was "bad"
i don't even think they're paid to shill it, they don't actually own nor want to own one - they just want nvidia to think that they do and lower their prices
this is an objectively poor graphics card

>>105677328
It was a side-effect of them dropping 32-bit CUDA support that was meant for enterprise customers to be forced to upgrade their hardware. There's nothing stopping Nvidia from creating a translation layer for legacy 32-bit Physx on their modern platforms.
>. Nvidia *just work*ing
This is mostly a meme. They have a laundry list of stupid issues over the years. They just have better damage control.

>>105679631
wrong and stupid you cant even back it up

>>105679631
>this card is more expensive, performs worse and has far less vram than a 2 year old 7900XTX.
Actually, 9070XT is almost as fast at raster but it faster at ML/MT and is a smaller piece of silicon to boot. It is cheaper than 7900XTX at launch even with the initial scalping going on. 7900XTX came out over north of $1K shorting after it came out. It only became cheaper when etailers wanted to clear inventory out when Blackwell launch was nigh.
7900XTX's only refuge is it is equip with more VRAM.
>it is quite dramatically outperformed by a 5070ti (a roughly equivalently priced
You may have double-check that. At worst, 5070Ti slightly edges it out, but they are roughly on par with each other under most settings content.
5080 and above are the only gayming SKUs are a cut above the rest.
The real take is that gayming market is upscaling and pricing most of its original demographic out. It has become a "geekier" version of Air Jordan where main customers (ironically poor-fags) are getting them as status/fiscal flexes rather than the merits of the product.

>>105679631
Even compared to the 7900xt it's pretty mediocre, like 10% more performance for almost as much more money, but with less VRAM
Yes it's better at RT but if you actually cared about that you wouldn't buy AMD in the first place

>>105672171
Nothing you can do about it. AMDGPU will randomly shit itself because it's a pile of shitty C code trying to tie down complex semantics. The Asahi tranny had a mental breakdown from how bad DRM and AMDGPU is.

I still have random page flip timeouts and vsync related power saving crashes.

>>105673867
>FSR
Ghost trailing blur fest. XeSS is the only nonshit choice. DLAA is shit too.

>>105679924
least delusional intel slopper

>>105677284
>>costs just as much as the 5070Ti, but with less features and support
Actually, 5070Ti is a bit more because Nvidia and retailers are justifying it with the Nvidia brand.
>features and support
Not really on both counts. Nvidia really dropped the ball on support side as of late. It is probably because they just don't care and know their paypigs are willing to eat it all up.

>>105679935
I have a 6950XT. XeSS is objectively the only nonshit "upscaler" available. For AA, TAA just works and has minimal ghost trailing. No reason to use the others unless you need upscaling.

>>105679910
RX 9070XT was intended to be a 7800XT/7900XT replacement. Cheaper to make and have slightly better performance. Nvidia did exactly the same thing with most of the Blackwell line-up. 5090 is the only outlier (It is bigger, faster and more expensive to make then 4090 it had replaced)

>>105679947
Absolutely dogshit opinion
>6950XT. XeSS is objectively the only nonshit "upscaler"
As a 6950XT owner your only choices are Xess and pre-transformers FSR, no shit you prefer Xess, it's the only one of the two that isn't a simple sharpening filter. You don't get a choice of using DLSS or new FSR. AMD abandoned you.
>For AA, TAA just works
TAA is so fucking terrible you may as well just smear vaseline on your screen. If you seriously use TAA and think nothing is wrong with it then you're a fucking moron.

>>105679961
>Cheaper to make and have slightly better performance
Yet more expensive and less VRAM, which is the main component that both companies use to segment their consumer and workstation cards.
Nvidia being shit doesn't excuse AMD from being shit, especially when their marketshare is in the toilet.

>>105679987
>delusional upscaler tranny claims TAA bad
It's literally the only AA that works in 2025.
Your only other choice is upscalers and XeSS is the least shit option that works everywhere.

anyone who bought rdna tech now while udna is next just shot himself in the head really

>>105680023
>don't buy now, new tech is tight around the corner
Still rocking that 2500k+980ti combo, yeah?

>>105680262
yes its a new uarch which means the support for rdna will drop faster on EOL thats a given since amd doesnt have the money to support both teams

Please Intel save us from this duopolistic tyranny!

>>105680273
It's also a new arch, which means that you should wait for the second or third gen of it if you don't want to be a beta tester for your own money. And by the time it arrives, it would be reasonable to upgrade from RDNA anyway.

>people with amd always lie to each other about the performance
lol funny how it's always them

>>105680023
It's not out until like 2 years. Assuming it's not delayed and they hit performance claims (LOL sure).

By that logic you would never buy anything.

>>105678835
Nvidia market share is slowly declining. The 92% shit you heard recently is GPUs shipped not market share. Of course they shipped more when 5000 series launched. 3/4 of it going to AI servers.

>>105676462
>That is 95% of RT in games
Not in higher RT settings available in the PC, even excluding PT.
People buy a gaming PC & dGPU to experience better-than-console experience.

>Hilariously without PT AMD is actually faster there.
There's nothing hilarious about a game developed by a Microsoft-owned studio that focused its performance optimization on console APUs that have Radeon hardware.
That effort trickled down to skew the PC performance in favor of Radeon cards.

>>105682476
>Of course they shipped more when 5000 series launched. 3/4 of it going to AI servers.
Serious AI projects don't want those GeForce cards because NVIDIA deliberately dropped NVLink support.
They don't even want the repackaged variants of these GPUs like the L40/S.

These cards are hardly worth it for even AI startups if they can rent some more performant & cost-efficient servers.

>>105682531
There is still a ton of demand especially in SEA and India.

>>105682477
All AAA titles are optimized for PS5 first. You only see Nvidia getting ahead in few Nvidia sponsored games.

Actually in latest tests gap between 9070XT and 5070Ti RT shrank to 2-3%.

Also 9070XT is a 5070 competitor not 5070 Ti competitor. The thing is AMD is 100$ above MSRP still because it's selling great while 5070 is rotting on shelves so it's dipping below MSRP.

>>105682555
Singapore is the only notable country in that region because NVIDIA and their Chinese customers are exposing a loophole.
They send more H100s etc. through that route than these GeForce cards.
You can't prove that a significant amount of RTX 50 cards are being delivered in the rest of SEA/India.

NVIDIA's reducing production of these GeForce cards to make more export-compliant AI GPUs for Chinese customers.
Yet Radeon isn't making significant gains even during this favorable environment, because people are still more willing to buy a GeForce card.

>>105682622
Anon 4090 and 5090 had such a massive demand in china USA specifically forced Nvidia to make shittier D models. Not every firm interested in AI can afford best shit.

>>105670638
I swear every one of these charts omits the 6900XT. It's weird...

>>105679924
>Ghost trailing blur fest
None of that happens you disingenuous faggot.
>.t 6900XT owner that uses FSR

>>105682589
>Actually in latest tests gap between 9070XT and 5070Ti RT shrank to 2-3%.
In games with less prominent RT options. The gap will only increase as devs add more sophisticated RT options.

>Also 9070XT is a 5070 competitor not 5070 Ti competitor.
Not according to AMD's official documents.

>>105682622
>Yet Radeon isn't making significant gains even during this favorable environment

It's not favorable at all.

AMD only had 9070XT. In the same period Nvidia was shipping 5090, 5080, 5070Ti and 5070.

And despite that Nvidia share in Steam surveys dropped below 75% when it was 80% not long ago.

>>105682662
Old news, NVIDIA already cut the 5090D's memory configuration to 24GB and they're producing fewer of these cards in favor of US export-compliant AI chips.

>>105682679
>In games with less prominent RT options

So 99% of them. Even in Nvidia sponsored titles 9070XT is not that much faster unless you use PT. Indy is literally one of the worst case scenarios for AMD and even there it's doing fine.

>>105682685
>It's not favorable at all.
5070Ti and especially 5080 are overpriced.
5070 12GB isn't very popular due to its memory configuration.
RTX 50 launched with unstable drivers that is finally starting to look stable.

This is one of the most favorable market environments AMD Radeon will have.

>>105682704
You buy a video card to own it for at least 3~5 years.
The RT options will only get more demanding over that time.

>>105682696
The point is Chinese corps were buying so many of those chips US government had to intervene.

>>105682705
And they are selling almost all cards they can make. 9070XT goes out of stock every day.

>>105682713
Those card purchases aren't being reflected in the Steam Survey you retard.
Why the fuck would they be dismantling GAYMEN coolers & PCB in favor of server form factor?

>>105682723
No, they're not. Not enough people are willing to pay over $700 for a 9070XT in the US.

>>105682709
AMD is getting Redstone later this year. Also 5070 Ti barely gets 60fps using DLSS in PT games. You need framegen for decent FPS.

If that's already happening today imagine what games in 3-4 years are going to do.

Who cares if 5070 Ti is faster in PT when both will get terrible fps in few years? And wile RT is becoming mandatory PT won't. Consoles or cheaper GPUs can't use it and won't be able to for many years.

>>105682727
That's the entire point retard. Half of Nvidia sales are top end overpriced cards going to AI farms. They never end up on Steam survey yet Nvidia % keeps going down.

>>105682736
They were until just a while ago. Thank fucking god all the retards willing to pay 900$ already got them and prices are finally going down.

>>105682760
Are you an ant? the pixel barely points down.

>>105682775
>AMD is getting Redstone later this year.
You don't know the final quality of this feature. More importantly, its adoption rate.
>Neural Radiance Cache
>Ray Reconstruction
>Frame Generation revision

All features GeForce cards have today, with 3/4x MFG.
"They're trying to catch up!" Is not the great sales pitch you think it is, and the market reflects this reality.

>>105682816
Market reflects that reality by 9070XT selling great everywhere.

>>105674253
Sometimes being the odd one out is smarter

>>105682832
>Market reflects that reality by 9070XT selling great everywhere.
The 9070 XT that cost $700 or over aren't selling out, that's the reality kid.
Meanwhile, people are still keeping track of the $750 5070Tis that instantly sell out.

>>105682746
>Also 5070 Ti barely gets 60fps using DLSS in PT games. You need framegen for decent FPS.
Oh no, you have to use features you paid for that are also part of AMD Radeon's Redstone upgrade.
RTX GPUs have a clear advantage in heavily RT titles, and the advantage grows further with x2/3/4 (M)FG.

Not to mention you can't Optiscaler cope away the lack of Anti-Lag 2 adoption that is required to run FG with low-input latency.
Just keep waiting, bro™.

Former 5700xt user here
It held me down when I was poorer and I anticipated a 90 class card from them but like the 5700xt they shat the bed and pretty much diminished what could have been a god tier generation and easy win.
5090 will still anal prolapse anything AMD can offer me on linux and yes I full time tumbleweed.
They just had to make a higher vram card and I would have stayed but the software and features is not enough for a user of my caliber

>>105680546
its literally cdna with graphics output

>>105682867
This Steel Legend on Newegg was sold out literally two days ago. They just got new shipment probably. 9070XT around 700$ usually sells out fast.

>>105684089
>sells out fast.
You are coping, badly.

>>105683035
Anon if you need to use DLSS balanced to get around 60fps today how do you think new AAA PT games are going to run in 2 years?

You will just turn it off like 9070XT users. Also in FPS games like Doom even 60fps latency feels slow.

>Rumors of a 9070xt 32gb version
>Deny the rumor and say you'll never make it
How do you fuck up that bad? So many retards will buy that for +$300 over the 16gb MSRP

>>105684131
>even 60fps latency feels slow.
Doom has no Anti-Lag 2 or FSR4 support, you can't even use Optiscaler in it.
Keep on citing a console-optimized title with no FSR4 or AL2.

>>105684123
Yes they are available now they were sold out 2 days ago. Pulse was sold out few days ago too.

Obviously most people rushing to get them already have them so finally prices and availability is stabilizing.

>>105684154
Except over 450 people/bots won't commit to the purchase. Coping. Badly.

>>105684153
Vulkan games don't get FSR4 yet. It literally just came out and there is no SDK yet.

>>105684172
So keep waiting™.
Why don't you explain the slow rate of adoption for Anti Lag 2?
You'll be getting an inferior FG'd experience without it even after muh Redstone™ update.

Also, Radeon is letting their board partners engage in price scalping as well.
Nice little "Launch Price" promotion baited simple-minded retards like you.
Over 550 people/bots are hesitant on buying the Pulse 9070XT for $750 (discounted from $880!).

Radeon and GeForce are engaging in the same profiteering, AMD just can't charge as high because people hesitate to pay over $700 for their cards.

>>105684207
Anon 9070XT was selling out for around 850$ since release. It's only recently that it's getting closer to 700$.

>>105684228
Steel Legend never sold at above $730 in Newegg.
And you've moved the goalpost from
>9070XT goes out of stock every day.
I accept your concession.

>>105684132
They have to deny it until it's ready for launch, that way people will keep buying the 16GB one.

No they fucking didn't.
Stock wasn't there. The price was fake.

Doesn't matter if they can hit MSRP/RRP 9 months later. Hit it day 1, have it in stock, have it at MSRP 3 months later.
That simple.

Also you don't win on a generation without vbios flashing and removed power controls. If amdsmi is your only outlet for taming the card then you're now nvidia tier.
RDNA4 is quite frankly a shit gen, availability, strategy and software is where it died. Having a DLSS2 equivalent frankly isn't enough.

>>105672484
>I do know there is an issue that causes the driver to use the incorrect clocks for some AMD cards and you have to set them yourself to fix it which is what I am doing.
This clock issue happens on Windows too with the 9060 XT btw.

>>105672113
Not a single top tier nvidia gpu in the list. You people buy bottom of the barrel scraps from nvidia, shitty gpus with little to no vram and call yourselves gamers. Fuck off.

>>105674255
Are you retarded? Not even a 5090 is that much better than a 4090, why the fuck would you even consider changing a 1600 USD gpu for a 600 GPU? Do you even understand money? What a complete lack of cerebellum is that? Do you even readproof what you write? No, for real, did you just had a stroke?

So XT or XTX?

>>105685904
cry

>>105685904
The xx70 would be mid range....if it were 2016 again. Sadly it's 2025 and ada xx70 certainly isn't mid range anymore, either is that 4060 Ti.
Ada was fucked, ampere was fucked. Turing was a correction to pascal and maxwell. Maxwell was weighed down by 0.5GB of VRAM, but Pascal, Pascal was free, maaan.
Turing had to jump the price up and stick with 8GB to compensate. The 3070 was there to send a message. A message to people who kvetched about buying Turing. The 5070 is sent to us from Jesus reborn, Jensen will make sure we pay for this if the 5070 doesn't top that list soon.

>>105685904
I literally bought a 9060 XT.
The only way I could buy more bottom of the barrel scraps from ATi, is if Frank Azor rose up from the ground and took 4GB away from the 8GB 9060 XT and sold me a 9050 XT 4GB.
Following nvidia's naming scheme? Shit. Yes, it makes sense to consumers. But how will I know the difference between the 6600, 6650 and the XT variants now in RDNA 4 and UDNA 5?
It's all so confusing brother.

>>105684132
There is a 32GB version of the 9070XT, but it's not marketed as a consumer GPU, it's their workstation class W9700.

>>105686041
>buy me a nice W9700
>sheeeeit
>want to use ROCm
>install ubuntu like good boy AMD wants me too
>stuck on 6.11 kernel
>have to distrohop
>distrohopping and dualbooting
...yeah. Nvidia suck dick for linux support, everything breaks all the time but AMD's recent woes during RDNA4 aren't enough to get me to buy a W9700 over a 5090 or even a 2nd hand 3090...
If Intel would sell me a B60 today, I'd even take that over the W9700 and enjoy my life without ROCm woes or dealing with the nvidia driver and linux power management.

>>105686131
I thought Ubuntu was on 6.14 now, yeah?
But yeah, ROCm support is extremely frustrating.
At least projects like GGML have shown Vulkan to be pretty capable for GPGPU programming.