>24GB

Why do AMD GPUs get so much hate? Are the drivers really that bad?

>>105857456 (OP)
>Why do AMD GPUs get so much hate?
They don't.

>>105857456 (OP)
They get hate from nvidia shills who constantly make fun of people for having them because they are poor. Drivers USED TO be bad and unusable, but oddly enough AMD has gotten their shit together on that aspect. Only remaining issue lately has been purple/green screening, but its minimal or often times user error. I have always used AMD and have had minimal issues with drivers but it does get tiresome to troubleshoot issues

no one uses rocm so if you want a card for consumer machine learning stuff you'll always have to wait days/weeks for someone to do a rocm port
their rendering perf for offline renderers is also significantly worse vs equivalent nvidia cards
no one cares about vidya games for babbies, grow up

>>105857456 (OP)
>buy used 6700xt for $240 early this year around the time of 9070 release
>now I can get new 9060xt 16gb for $380
I want to kill myself.

No one makes good GPU drivers anymore, it's not like it's endemic to AMD.

>>105857456 (OP)
Yes. The drivers are terrible and many games and emulators just don't work. Nvidia are dominant for a reason. Even the army of AMD shills online use the money they're paid to buy an Nvidia card.

>>105857456 (OP)
why do you care? that's the real question. Just get whatever you want regardless of random pieces of text on your screen. Most are bots anyway.

>>105857456 (OP)
The only problem is their unwillingness or inability to shit out the best GPU.
Never had issues on either Nvidia or AMD. We just don’t get a good PPP anymore from them so it doesn’t even matter. 3000 euros for a 5090 is beyond retarded for example.

>>105857456 (OP)
they're only affordable in the us most likely, in my country they're so expensive it's just not worth it buying it instead of a nvidia card
I used to buy amd cards before when you got like twice the performance for half the price, but now they're just slightly better for the same price. just not worth it for an inferior product

>>105857842
This is deeply misleading. AMD driver bugs tend to be the same as nVidia driver bugs - when you change major architectures it takes a while to find all your years of game specific hacks and modernise them. People just remember that time they bought an RDNA1 card and it wasn't as stable as the GCN, so they bought a late gen Pascal with years of fixes and declared nVidia the GOAT, ignoring all the years of bullshit the early adopters pretended didn't exist.
And emulators are notorious for being coded to only run on the dev's own machine. If you have a similar machine you'll be fine. The problem is not nVidia vs. AMD, it's "I only have a 3060 so I only tested it on a 3060, and all my friends have 30 series too so they all agree my code is perfect."
When you code your OpenGL4.6 backend to exclusively use NV_EXT_ extensions, well no fucking wonder AMD is not supported. Fortunately most emulators are open source now and eventually someone competent comes along and fixes the bullshit. But that doesn't stop people repeating decades old myths on image boards forever.

>>105857456 (OP)
You can't do anything fun with ml/ai requiring cuda (pretty much everything) with the 24 gb so it's kinda there for no reason.

>>105857456 (OP)
on Linux they're fine
ive been using a 7900XTX since it came out, primarily on Linux but sometimes on Windows
have yet to swap it for the silly Nvidia GPU I have sitting in a box since I'm waiting on at least 580 drivers on Linux for a fix

>>105857456 (OP)
>Why do AMD GPUs get so much hate?
Where? If you filter out nvijeet shills the hate is almost nonexistent

>>105860075
Ah, but then you get TWO AMD GPUs.

>>105860075
yea no, most shit works, even on "unsupported" hardware which never meant what you thought it did and if it doesn't porting libraries to HIP is trivial even if you do a high effort port with a full divergant path for HIP in the cmake
and this isn't the first year of AI, there are a significant amount of alternative ecosystems
the tinygrad dev's user space PCIe driver and custom compute stack for the 7900 managed to put it close a 4090

>>105857456 (OP)
For the actual thing people get 24gb cards for (ML stuff), AMD never really had an answer to CUDA so there was never any point to considering them no matter how much VRAM they threw at a card.
For actually just playing games they're definitely overhated. I think they have a passable consumer-grade compatibility layer for CUDA now too but I could be wrong on that.

>>105857456 (OP)
just look at the box
Raytracing - terrible on Radeon and RDNA3
Radeon Boost - lowers resolution while playing who wants that
Radeon Anti-Lag - got thousands VAC banned
24GB VRAM - no applications to use it

$1000 for that?

>>105857456 (OP)
AMD legacy support is garbage, not a problem for AAA sloppa but if you plan on playing dx9 or older games you're in world of pain.

>>105860809
Just use DXVK like a normal person.

>>105860364
Oh yeah let me just do workarounds on shitty forks that run as slow as a 3060 on my 5070xt. I have amd and it sucks outside of getting to play cyberpunk with high raytracing for $600

>>105857456 (OP)
Were it not for heavy reliance on NV to do AI, AMD would have it in the bag.

Wrecked by 16 GB njudea 5070 TI

>>105860447
>Raytracing - terrible on Radeon and RDNA3
RDNA4 ray tracing is pretty good, RDNA3 ray tracing isn't that much worse than nvidia's lower end offerings (which cost as much as a midrange radeon)
>Radeon Boost - lowers resolution while playing who wants that
oh yeah cause nvidia won't write DLSS on the box or anything like that
>Radeon Anti-Lag - got thousands VAC banned
that was anti lag+, regular anti lag and anti lag 2 just werk and the latter is also built into many games such as cs2
>24GB VRAM - no applications to use it
but why is having more vram than you need a bad thing
>$1000 for that
as if ngreedia's 1000$ gpus are excellent value, kys

>>105860809
>that retard that thinks radeon cards can't play old games still hasn't come up with any proof for his bullshit claims
k

>>105860876
Why does Nvidia have that advantage to begin with?

>>105857456 (OP)
Radeon lacks only one thing, stronger international marketing.

>>105863508
Because cuda is proprietary

>>105863555
AMD marketing in general sucks. If its one thing Nvidia/Apple gets right, its marketing.

>>105863508
i've always heard it was historically largely because of research, outreach, deals, and free hardware with universities whereas i've heard even if you work for AMD directly you won't necessarily get hardware to work with
and their software was lackluster, they tried a lot of odd stuff before giving up and spending the last long while making a basically 1:1 clone of CUDA with renamed symbols (sometimes to their detriment, supporting AMD specific features often comes second fiddle to NVIDIA compatibility)

largely still can't actually get many people in the ML field doing research who actually write real compute kernels to use it or support it directly by default
but at least porting things is extremely trivial and every big runtime and library supports it and with modern versions of pytorch it's a lot less necessary to actually write compute kernels from source


>>105864637
buddy

>>105857842
>Yes. The drivers are terrible and many games and emulators just don't work
what? ive had 3 amd gpus and one nvidia gpu. ive never had a game "not work" on any of them.

>>105864665
Maybe the algorithm is just in Nvidia's favor, but when I look up gpus, no AMD cards comes back.

>>105857456 (OP)
>Are the drivers really that bad?
yes, AMD Windows drivers are fucking atrocious, The minute you deviate from gaming on a single-monitor setup the drivers begin to unravel. and don't even get me started on hardware acceleration issues, to make matters worse no professional software works out of the box with their cards. Saving 50$ by buying an AMD card while losing out on CUDA is a bad joke that only retards will fall for, which results in AMD cards having no second-hand resale value.
Perhaps with UDNA they will move in the right direction but honestly if they don't have an answer to CUDA (and no, ROCm only works on a handful of cards) its just over for them and our best hope is for Intel to do something.

>>105857456 (OP)
AMD drivers used to be bad. Now, their drivers have gotten significantly better where as Nvidia's have gotten slightly worse and continue to get worse. The only thing that makes me think twice about getting an AMD GPU nowadays is legacy games and very obscure issues. If you're just playing modern shit I wouldn't hesitate to go with AMD nowadays.

>>105865233
Nvidia doesn't have physx anymore so you can't trust them for legacy games either

>>105865202
>he thinks 100% of the population needs to run applications that rely on CUDA.
...jesus fucking christ how deluded are you
the primary use for all these gpus is literally gaming, you think little jimmy looking to play fortnite isn't buying my used radeon because it doesn't support cuda? in what world do you even live in

>>105857456 (OP)
VRAM capacity alone does not cut it.

>>105865495
This. 24 gigs in an AMD card is like having a 12 inch dick that hurts to touch.

>>105857456 (OP)
1000 dollars to play uhhhhhhhhhh
7800 XT can max out decade old games at 4K

>>105863508
8800 GTX could run Crysis, also having massive headstart on GPGPU development while everyone was dealing with the agony of CPU 3D rendering.

>>105857456 (OP)
This card runs great on Arch and ROCm is almost as good as CUDA when it's available. But projects tend to take a while to support it if at all, so often you need to be prepared to do some tinkering yourself or just have some patience. For gaming it has so far handled everything I care to throw at it with no real problems. Had some graphical artifacting trying out framegen/upscaling but I don't care about getting more than 144fps and I play at native resolution so this is a non issue, and it struggles with ray tracing in some high end games. There's a little bit of coil whine but it's not noticeable over the aggressive fan curve required to keep it cool. That's about all my nitpicks though I have no regrets

>>105864822
>porting things is extremely trivial
If I want to learn how to do this, where do I need to start learning? Supposing I'm a retarded end user with next to no understanding of these things

>false pretense OP thread 374893675269061764378195671064796732

>>105866001
with C++
i should have clarified, trivial for a developer
rudimentary porting of the source code itself would be maybe? be easy for someone who has no clue what they're doing since there's some tools that help with that but going from there to a binary takes
>>105865202
>and no, ROCm only works on a handful of cards
again with this absolute nonsense and just straight lies - officially supported means enterprise support specifically for prebuilt packages - clang, amdkfd, the two libraries that make up HSA support with HSA device support on amdkfd and HSA runtime support, and support for the ISA (or another one from that family, and soon that won't even apply, they just have to update the clang the prebuilt packages come with to i thnk clang 20, whichever one added generic amdgpu ISAs support) in them are literally, literally the only thing that matters

strictly speaking libhip and hipamd support doesn't even fucking matter because there's a fair amount of projects that have started using the HSA specification directly and writing to command buffers directly

>but, but it's not in the officially supported list
a list of card names which doesn't exist in the actual code in any meaningful capacity
it doesn't matter

and intel has done something you little shit, they've done tremendous work on SYCL, upstreamed parts of their offloading infrastructure to llvm, written SYCL backends for CUDA and HIP, and made a backend creation library for third party vendors

it's just the constant fucking lies
oh it's just that they only ship

AMD drivers might not be perfect, but they wont resume from standby with my entire screen upside down

>>105857456 (OP)
amd got good with rdna2

>>105865233
So if i play mostly retro games, I am better off sticking with nvidia?
I am really tempted to switch to AMD.
I dont care for AI, upscaling or raytracing the only thing that worries me are potential compatibility issues.

>>105857456 (OP)
i have an xtx it works great
>>105860873
do you mean 9070xt? also that card only got official rocm support really recently so its probably why youve had to tinker with shit ive had no issues with my 7900xtx image gen is fast, llms are fast, only thing giving me slight issues is wan but i think thats because theres no proper amd port of sage attention so my wan gen times are like 1.5x nvidia which isnt much of a problem its always bad to get gpus when they just come out in general ive had this issues with both amd and also when i had a 3090ti

>>105857456 (OP)
Their drivers used to be shit, but now days it's unironically decent, and Nvidia has been the one lacking.

>>105862632
>but why is having more vram than you need a bad thing
u are paying for it

>>105865233
people need to mention what fucking os theyre talking about when it comes to drivers because daily driving linux for the past year on a 3080 ive had no issues with games, ML, or anything else ive used it for

>>105857456 (OP)
i just upgraded from 1070 to 9060xt 16gb
>amd adrenaline blocks other programs from fan control if card is OCd
>cant change voltage in afterburner even with everything unlocked in settings
>adrenaline is bloatware full of AI slop and stream shit if you want the version that lets you overclock
>using custom resolution on nvidia simple as letting application decide
>on amd you have to create every single custom res you might want to use, has all sorts of options you need to carefully check and learn about
>less global graphics options than nvidia
>alt tabbing from games seems slower/less smooth

i dont like it, but mostly im annoyed that theoretically i have a card about 2x more powerful than my old one and all modern games are blurry TAA slop so I don't even see an improvement in fidelity.

>>105867180
>not installing minimal
lololololo

>>105867242
>adrenaline is bloatware full of AI slop and stream shit if you want the version that lets you overclock

I'm using minimal lil bro

>>105867272
If you are already using Afterburner, why would you install the Full software to OC, retardbro?

>>105867306
im not 'using afterburner' lil bro, i opened it up to see if it worked, it didn't

>>105867328
I'm not your lil bro, pal.

>>105857456 (OP)
We don't hate AMD, we just feel sad for them because they are essentially reduced to being controlled opposition to NVidia at this point.
Price is exactly Nvidia -$50, but missing extra features, and limited AI support.
Both CEOs are cousins so this explains a lot.
AMD has never been good since Poolaris and GCN.

>>105857473
This
>>105857456 (OP)
>Drivers
It's Nvidia that has the driver issues now. Almost all games have stuttering issues on all 4xxx and 5xxx

>>105865379
and again, i literally said in my post "The minute you deviate from gaming on a single-monitor setup the drivers begin to unravel".

>>105866616
its not nonsense, you can buy a 1050ti and you will be able to run any modern application and have hardware acceleration working fine, the same cannot be said for example with an RX6500, that matters. Even basic things like AV1 are a clusterfuck and you need to look at the specs to see if you're getting a functional card of handicapped silicon (its gotten better with the 7000 series since their lowest is the 7600 now but you should be able to understand my point.

>>105857456 (OP)
Oh look a shill post right at the moment Nvidia is in trouble in the consumer space. What a coincidence!

>>105868923
> 1050ti
> Will run any modern application
Lol okay. What a bunch of shills. The RX 6500 only purpose was to workaround cryptomining boom, but who cares right? Let's all shill against AMD!

>>105867171
> Linux
> Nvidia 3080
> No issues
Lol ok. Try using Wayland or VirtualBox and see how it goes. Nvidia is notorious for being worse in Linux than AMD or Intel, not because of performance, but incompatibilities with Vulcan, OpenGL and tools. They've improved a bit but it's still suboptimal.

>>105867128
>7800 XT gets official support out of nowhere as well
Nice

>>105867120
You will not have issues with AMD on legacy games. That it nonsense. But if you must know the issues you can face are PhysX will run on CPU and Nvidia uses a fucked up OpenGL that doesn't actually conform to standards, so on Windows AMD applies patches in the driver to workaround, and on Linux the MESA guys also apply configurations to those specific games to workaround. The advantage of Linux is you can apply the fixes yourself, while in Windows you can't.

Source: I've helped MESA fix some games that only worked properly ok Nvidia, but not Intel or AMD.

>>105860052
Also emulators today usually have Vulkan and that means Nvidia OpenGL hackery doesn't work anymore. In fact I think AMD created Mantle to put an end to Nvidia OpenGL bullshit and they've succeeded.

For info, Mantle was the AMD code that was given to Kronos Group and then they created Vulkan.

>>105869291
>Lol okay
I'm talking about CUDA and hardware acceleration, not your fucking childrens interactive entertainment mediums you retarded faggot

>>105869621
Whoa, keep it friendly in here!

>>105857456 (OP)
Used a GTX 970 and GTX 1080 until I switched to a RX 9070 this year.

AMD drivers and ease of interface is a bit under what Nvidia used to offer, but I made the switch because Nvidia was shitting the bed even more recently.

AMD is in a phase were they could catch Nvidia old quality of drivers in the next years, while Nvidia would need to be very retarded to not fix everything quickly when AMD catches up.

Overall, everything is fucked up but AMD is more trusty for the current gen of GPUs.

>>105857456 (OP)
Quite literally the only thing that stopped me from getting an AMD card was VR as Nvidia still had its justwerks reputation when I bought a 4080 Super, especially for VRChat. If I had waited like 3 more months and put in a bit more research, I would 100% have gotten a 7900 XTX.
But yeah my next GPU will be AMD as my next OS will be Linux and Nvidia's high VRAM cards will likely cost 2-3x than my car's worth. Fuck that.

https://www.phoronix.com/news/Tiny-Corp-Drops-AMD-Again
Even companies trying to use AMD get dogshit drivers, it's a well known problem and anyone trying to deny or downplay it is probably an AMDrone

>>105857473

>>105868923
>nitpicks the only amd gpu ever released to not have video encode/decode
???
>its gotten better with the 7000 series since their lowest is the 7600 now
...the 6500 xt didn't have this problem, the rx 550 neither
you are talking about a one off thing that happened as if it's a constant issue that amd has always had until the 7000 series, that doesn't make any fucking sense
the rx 6500 xt is also like the only dGPU ever made to only have 4 pcie lanes, why'd you ignore that? oh right because you have no idea what you are talking about
this is like saying "by buying an iphone you have to check if you can hold it a certain way and still get a signal" when that issue was only had with the iPhone 4 16 years ago
well, thanks for confirming that by "don't even get me started on hardware acceleration issues" you meant this nothingburger rather than something actually important like the fact that their h264 encoder is still behind intel's first gen quicksync or something like that, what a fucking retard.

>>105867180
so like half of your complaints is the fact that you can't use two different software for oc/fan control at the same time?
why the fuck would you ever want to do that
nvidiots really will pick the tiniest stupidest thing to claim that radeon is "unusable", crazy.
and besides, i'm pretty sure nvidia has no overclocking in their drivers AT ALL, plus they do have ai slop and stream shit there, so why are you complaining about something which is worse on nvidia? really makes no sense.
>on amd you have to create every single custom res you might want to use, has all sorts of options you need to carefully check and learn about
I didn't even get this one, you have to make custom resolutions youself... as opposed to? having them being automatically created out of the blue or something? i'm pretty sure that's not a thing, my monitor won't magically overclock itself on an nvidia card, what did you mean by this.

>>105870304
>...the 6500 xt didn't have this problem
*the 5500 xt didn't have this problem

>>105857473
fpbp
/thread even

they're good on linux where the support for rtx cards is isn't good
I think it's a bit weird though that their best current card is just mid tier and costs 700. it meants that when they finally bring out the 9090 xtx it will cost way more than the 7900xtx

>>105857737
This
>>105857615
AMD drivers are atrocious and Wayland trannies lie about this constantly.

>>105870304
the list of AMD inferiority is fucking endless, of course I know their video encoding is terrible, but that's just another grain in the heap. AMD is just the inferior product. I haven't wasted my money on AMD GPUs in like 4 generations now and I have no intention on touching them until they rewrite their drivers from the ground up, preferably by non-indians.

I simply used the 6500XT as an example because it at the time it was a card I was looking at as a possible HTPC upgrade but I realized none of them can even drive a single 4k144hz screen with HDR on linux because they don't support HDMI 2.1 due to licensing (https://gitlab.freedesktop.org/drm/amd/-/issues/1417 , nice 4+ year old unsolved "issue" without so much as a bandaid solution) and all the DP>HDMI dongles have broken firmware or weird bugs. I do want to support AMD as a company but their incompetence makes it impossible to do so.

>>105857615
>Only remaining issue lately has been purple/green screening
So their drivers literally haven't improved from when I last had an AMD/ATI GPU.

>>105857456 (OP)
>Are the drivers really that bad?
On Windows? Yes. I thought it was a meme, but I have so many issues with my 7900XT it's absurd. BSODs, weird graphical glitches in several applications, it's a miserable experience.
On Linux? No, it's pretty much perfect.

>>105871154
The golden rule of AMD drivers on Windows is "if it ain't broke, don't fix it".
Check online which driver packages are the least problematic for people, pick one that supports the newest game you want to play, and see if it works. Repeat until you find one that does, and then don't touch it until a new game comes out that won't run on your old drivers.

I will say the multi-month gaps between releases seem to have helped with stability, 25.3.1 has been extremely solid for me ever since it came out while pretty much everything before it (on the faster release schedule) was a mess.

>>105871154
and you know why that is right? because the Radeon driver is written by the community and not AMD. even Apple writes their own AMD drivers so it's pretty apparent where the problem is.

>>105871228
That's good advice. I've been meaning to roll back my driver to a more stable release, but I ended up tracking down what was causing the BSODs (the audio driver) and disabling it. I'm just living with the occasional graphical artifacting for now.

>>105871252
Yep. I never had any experience with them, but my understanding is that the old proprietary AMD drivers on Linux were a fucking nightmare.

>>105871396
oh god yes, every time on Arch things would break because their driver didn't support the latest Xorg versions so you had to versionlock things, not to mention the fucking mkinit hooks required to rebuild it constantly had issues too.

>>105859991
>my country
>doesnt name the country
thirdie detected

>>105871008
That's not a competence issue or a technical issue. The corporation that holds the HDMI rights is actively blocking it. AMD currently can't support HDMI 2.1 in an open source driver for legal reasons.

>>105871747
it is a competence issue because both Intel and Nvidia have solutions that remedy the issue, I don't care if its not the perfect solve because NV ships the binary blob, or Intel re-routes their HDMI through DP. Only AMD leaves you without a solution.

the 7900 XTX is one of the best value for money AI cards available if you're not a disgusting Windows zogslave.

hilarious how the njudea shills always start screeching that AMD can't do AI. Ok, how did I make this picture then??

>>105871899
AMD submitted technical solutions like those and the HDMI forum rejected them.
Again not a competence issue, it is on the fucked up rights holder for HDMI.

>>105867128
Yeah I have a 9070xt. I really wish they hadn't fucked with the naming scheme.

>>105872044
why do I, the end user, care about what AMD said to the HDMI forum? I want a solution from AMD directly, copying Intel or Nvidia is a solution and they've proven for over 4 years they can't even fucking do that. So stop bullshitting and being a fangay.

>>105872114
>copying Intel or Nvidia is a solution
They tried those types of approaches. The HDMI Forum rejected them. Short of an intentional violation of the HDMI Forum's IP rights, there's not much AMD can do.
It just seems like the HDMI Forum, who AMD aren't a major player in and who do not want to see open source platforms catch on are simply going to continue to fuck them over.

>>105857456 (OP)
They think it's still 2010 or who knows when they allegedly had shit drivers. Honestly I never understood the meme about Radeon. I've used Radeon through since ATI to this day. Had 5 GPUs from them. The only Nvidia I've had was in a laptop, but gaming laptops are shit, laptops in general sucks. I always take AMD, because they always had better price/performance ratio for me.

>>105872138
I think you're missing my point here, that AMD is just the inferior product and nothing "just works". There's always some compromise or tradeoff with AMD cards that just doesn't exist with the competition.

Should I paypig for a new 9070xtx build, or should I wait 2 more years?
I'm gayman on a razer blade 2019. It's fine for what I use it for, but if I had more power I would play more graphics intensive games like vr stuff

>>105872186
Just get the microcenter 2 year replacement warranty and then fucking break it in 2 years and get all your money back in store credit and get another card. That's what I do. You get all the money you paid back.

>>105872179
Except this is not a competence issue on the part of the open source driver devs, or a technical problem. In fact they wrote a working solution.
HDMI Forum blocked the release a year ago and is still road blocking any other solution.

As for 'just working' Nvidia's hardware has major issues with modern GNU/Linux distributions, and Intel still can't produce competitive hardware. Everybody has major trade-offs. AMD with the open source drivers are still the best option and the only 'trade-off' is use the superior Display Port standard over a shitty non-standard proprietary plug.

>>105872218
you've never done this

>>105872173
I had a radeon gaming laptop back in like 2009 that had a 512mb gpu in it. It worked fine and cost like $550 at that time. I mostly played css though. It was good because my parents were divorced so I could take it back and forth from my dads house.

>>105872231
I have and I'll do it again. Jam your fans and overclock the card and keep running it until it dies. Then bring it back and they give you a gift card with what you paid for it.

>>105857456 (OP)
the drivers are mostly fine now, in fact nvidia has gotten worse so IMO they're about the same now. the problem is AMD fell behind on AI tech like upscaling and frame gen. FSR4 is quite good now, but the game support doesn't even touch DLSS because FSR4 is pretty much brand new and only works on RX 9000 cards.

>>105872220
>As for 'just working' Nvidia's hardware has major issues with modern GNU/Linux distributions,
you mean that tranny wayland trash? miss me with that

>>105857456 (OP)
>Why do AMD GPUs get so much hate?
novidya paid shills
>Are the drivers really that bad?
RDNA1 is beyond cursed and still has problems
RDNA2 is rock fucking solid everywhere and has been from launch by some damned miracle
RDNA3 had loads of issues that much I know., no idea now, no idea about what comes after it

>>105857456 (OP)
>Why do AMD GPUs get so much hate?
/g/ literally stands for "/g/ay for AyyMD"

>>105857456 (OP)
>>105871900
cant force AA thru the driver.
AMD is useless

>>105857456 (OP)
Because they aren't NVIDIA GPUs, the ones people actually want. They can't use proprietary NVIDIA technology so they have less features, have worse drivers, have worse energy efficiency like most AMD products, same as 10 years ago but much much worse. Kepler was the coup de grace for the Radeon brand, Ampere/Lovelace was the funeral.

>>105863508
It all started with the GeForce 8000 series, nearly 20 years ago.

>>105857456 (OP)
its due to nvidiotism

>>105872138
*laughs in GDMI*

>>105863508
ATI flubbing up with R600 a.k.a 2900XT and been playing catch-up to Nvidia. They overtook Nvidia a number of times but it was not enough to stop Nvidia's momentum. Nvidia made the distance with Maxwell and onwards.
I would argue 970 being the centerpiece of this despite its "3.5GiB+0.5GiB" meme.

>>105867180
Why the hell are you overclocking and overvotling modern GPUs? It is retard to bragging rights-tier at this point. Modern chips are already binned close to their breaking point with maximum boosting. It is actually more beneficial to undervolt your chips as they sustain higher boosts due to the reduce thermal load.

>>105857456 (OP)
https://m.youtube.com/@northwestrepair/videos
Cringe thumbnails aside, he does solid repairs and says that he hates amd cards because most of his amd card repairs are warped or bent, which makes them prone to failure even if you fix them

>>105857456 (OP)
struggled a lot with AI genning with their stuff. use of AMD stuff in AI is less supported, and at times outright undocumented. Other than that, it's amazing, and you get good, stable performance, especially if you use an amd cpu.

Also, nowadays you can just offload you genning to server farms like salad iirc.

>>105874731
So you can kill the card faster and get a new one

>>105857840
Underrated, imagine a world of optimized drivers (for real) and games like the first silent hill or doom 3, buying 12gbs cards would be overkill if you dont own 8k tvs.

>>105869358
i am using wayland dumbass im running hyprland and have been doing so for awhile without any problems, two monitors as well

>>105872742
I like my 9070xt. I get to play games I couldn't before now.

>>105857456 (OP)
>Are the drivers really that bad
No iv'e had zero issue's with my 7900XTX since launch

>>105871900
Exactly the 7900XTX can do AI and it can actually do it extremely well

>>105879343
But it cant, its slower than a 3060 on stable diffusion

>>105872742
> They can't use proprietary NVIDIA technology
marketing gimmicks
>less features,
For gayming SKUs not really. On the professional scene this has merit
>have worse drivers
Not anymore, Nvidia cease giving a crap post-Ada Lovelace
>have worse energy efficiency
Not really, this is mostly from node difference not architecture. Ampere was ass in efficiency because it was stuck on Samsung's shitty 10nm node. Ada Lovelace and Blackwell fixed this by being back on TSMC's 5nm/4nm nodes.
>Kepler was the coup de grace for the Radeon brand
WRONG, it was Maxwell into Pascal. Kepler was recovery mode that had aged very poorly. Pascal was the actual nail in the coffin. Turing is when Nvidia ceased giving a crap. Ampere was panic mode because RDNA2 gave Nvidia pause from its XSX/PS5 teasers. Ada Lovelace is when Nvidia went back into form.

>>105879360
on windows that's for sure because AMD couldn't arsed to support that but it's definitely fast on Linux
t. 7800 XT owner

>>105879360
on windows that's for sure because AMD couldn't be arsed to support that but it's definitely fast on Linux
t. 7800 XT owner