Why do Nvindians hate AMD GPUs?
It's All Fucked Shirt $22.14 |
Shopping Cart Returner Shirt $21.68 |
It's All Fucked Shirt $22.14 |
Why do Nvindians hate AMD GPUs?
It's All Fucked Shirt $22.14 |
Shopping Cart Returner Shirt $21.68 |
It's All Fucked Shirt $22.14 |
it is not about hate,
it is about being shit-skin ( likely indian ) outsourced shill and You have a script that You have to follow
or in some cases ,
being so unsatisfied with Your computer that You just have to be mad at other computer parts , which could be considered mentally ill, but at least not brown
I don't think about you at all
>I don't think about you at all
And yet you reply
Because AMD is no longer playing price wars. Nvidia is free to upsell their SKUs because the whales can't stop themselves.
AMD has never made a good chipset, it's not about hate op
You have to go back a decade+ for chipset issues.
amd makes shit gpus
>t. I own one
Radeon cards really can't do much outside of gaming, and aren't competitive beyond traditional rendering, they suck at raytracing. For compute, you can buy any nvidia card and expect CUDA enabled applications to work. Not so much on AMD, ROCm isn't supported across the entire product stack, and even if your card is supported, support in applications, ML frameworks, isn't a given. Even In a best case scenario, a top-end radeon like the 7900xtx is outpaced by a much cheaper NV card like the 4060ti.
muh gaymes
muh gaymes
muh gaymes
muh gaymes
If that's all you use your computer for, a radeon is a great choice. It's all they're good for.
well unless you want to use modern features in modern games.
AMD already has alternatives to NVIDIA's ""recent"" DLSS offerings. For rasterized gaming AMD cards perform very well for their price.
Hopefully Intel continues with GPU development so AMD can't start charging out the ass like NVIDIA.
Obsession with being seen as high status/caste. They think associating themselves with the number one brand will make people think they're rich and successful.
nobody has even mentioned wealth or status, i think you are projecting your insecurities about your own finances
I buy amdshit because I am obsessed with being seen as a low caste sissy bawd that can be scammed by anyone. I think associating myself with the number three (intel is second btw) brand will make people think I don't deserve proper firmware and driver support.
>They think associating themselves with the number one brand will make people think they're rich and successful.
That was the last thing on my mind when I bought my 3060. If AMD had something similar to CUDA I would go with them.
>If AMD had something similar to CUDA I would go with them.
They release a new CUDA-'alternative' every year. (and then they drop support, because frick you)
Making bad software and then dropping support for it.
Reminds me of a certain race
they do
they've had one for a while, called HIP/ROCm, the base of the project started out as an attempt at modernizing their opencl runtime but is now literally just a clone of CUDA with cu and cuda cut out and hip pasted in and is so much like CUDA it significantly neglects AMD specific hardware features
the quality isn't great and neither is the performance (the runtime libs are slow enough that transpiling PTX from CUDA libraries to AMDGPU ASM is faster) but using it is braindead enough a few projects support it
my guess is in a few years AMD's going to give up and let ROCm be eaten by intel's SYCL implementation
it's an open source implementation of an open standard slowly getting support for all of intel's already high quality libs, and they've intentionally designed it so other manufacturers can use it easily
right now intel's made a rudimentary AMDGPU backend on top of HIP but i imagine eventually AMD'll just write a backend on top of their kernel drivers, pass the rest of the work off to intel, and stop maintaining ROCm entirely
that or just use vulkan, which these days has enough compute capabilities to act pretty close to a functional "alternative" compute runtime with near feature parity to CUDA (including AI shit) and it's supported by anything that can larp as a GPU
doubt it'll make a difference though, CUDAgays are lazy fricking morons who need a single pass compiler to baby the shit out of them and are basically the iToddlers of graphics programming
even when 9 times out of 10 adding support for another manufacturer is like two lines of cmake since both intel and AMD have transpilers they don't do it
Intel's commitment to open source has been somewhat questionable lately (XeSS is still closed source), and compute is important enough that dropping ROCm for something controlled by Intel likely doesn't make sense. There would probably have to be a situation like with Mesa where there is an underlying common framework but with vendor-specific components.
SYCL is a khronos standard, intel just maintains the biggest backend for it, and i believe all of it is open source (including a modified version of a formerly proprietary library they bought)
and i thought intel was trying to get the sycl runtime merged into llvm proper
or at least parts of it
the XeSS thing is weird and concerning though, i'll give you that, especially because they've got open source plastered all over its marketing
>and compute is important enough that dropping ROCm for something controlled by Intel likely doesn't make sense
i mean the thing is that it's not just their SDK that sucks, the base HIP runtime (the equivalent of something like radv or gallium) sucks too
it's not just very poorly implemented, it doesn't even let you use like half the features of their cards
>There would probably have to be a situation like with Mesa where there is an underlying common framework but with vendor-specific components.
that'd be the ideal, with AMD basically just maintaining amdkfd just like they do amddrm
but literally none of the typical devs or foundations have seemed remotely interested despite 99% of the work already being done since clang already implemented the language
rusticl's the biggest development in the space for years and it's opencl which means it'll never see widespread adoption
>it's not just very poorly implemented, it doesn't even let you use like half the features of their cards
I think AMD is just trying to get past the first hurdle of getting some developer adoption. When you have as little PC marketshare as they do getting developers to add support for things specific to their hardware is rough.
>People only buy Nvidia to brag about it on anonymous image boards to seething AMD poorgays
Do people actually believe this?
>AMD cards perform very well for their price
>They (nvidia owners) think associating themselves with the number one brand will make people think they're rich and successful.
>Hopefully Intel continues with GPU development so AMD can't start charging out the ass
I'm starting to see a pattern here...
This is really the only reason schizos buy AMD
My card works fine.
>This is really the only reason schizos buy AMD
I have been quite happy with AMD cards though. Had a RX560 then upgraded to a RX580 which is still running great in my son's build, the hand-me-down computer. I put it together in early 2021 with the intention of swapping it out by Christmas 2022. No need to change shit right now.
>They release a new CUDA-'alternative' every year. (and then they drop support, because frick you)
If they focused on something that works instead of piece-meal hiring of H1Bs for driver positions it would be really frickin neat. If you're reading this Lisa, I love you.
Bad software. Maybe one day AMD will prompt AI with 'rewrite our software, but make it so it doesn't suck ass'.
i understand correlation does not equal causation but there seems to be a trend of mentally unwell individuals who bitterly defend amd without even getting paid for it. whats up with that?
The divide between AMD and Nvidia hasn't always be so wide. It was during the GCN days where AMD really stumbled and fell in comparison to Nvidia in every single category imaginable, creating the gap we know today.
Nvidia and AMD basically trade blows each gen.
6000 series beat 3000 series. It was a TSMC 7nm monster vs a dogshit 8nm Samsung.
Nvidia took it personally and then made the 4090 which is a 5nm TSMC monolith beast. AMD took the gamble making world first consumer GPU chiplets which caused them to not hit their expected targets. Expect RDNA4 to be the return to form but Nvidia will fight hard no doubt so competition is back.
The problem is that even when AMD has had very competitive hardware in terms of price performance, or even has definitive wins for performance, it doesn't translate into sales. AMD's long game projects of integrated GPUs, SOCs, and GNU/Linux support are starting to pay off with products like the Steam Deck
The rumors that their next-generation will be an OEM focused mid-range chip produced in bulk, with maybe a high-end chiplet workstation card seems credible. AMD's only hope in discrete graphics is to start to match Nvidia in chip production volume with OEM contracts.
Nvidia's 3060 may have been very disappointing compared to similarly priced RDNA2 hardware, but Nvidia sold more 3060s than all of AMDs RDNA2 hardware combined.
>Nvidia sold more 3060s than all of AMDs RDNA2 hardware combined.
you're literally moronic.
there are more ps5 owners than 3000 series owners
I'm talking about PCs. No doubt having the consoles has helped AMD, but the margins are lower and not getting OEM deals has really hurt AMD.
I hate Nvidia for being price-gouging c**ts and that put me in the red camp for the longest time, but I owe exactly zero allegiance to corporations so now I hate AMD too. If I can't like them for being notably cheaper, then what the hell do you want me to like them for?
AMD gpus are mid but their cpus are fire
AMD GPUs were garbage during GCN which spanned nearly 10 years. Liking AMD GPUs or having AMD favoritism basically makes you a newbie. It is now cool to like Intel.
>AMD GPUs were garbage during GCN which spanned nearly 10 years.
Wrong, GCNs was spanking their Nvidia counterparts. Nvidia only started to come around with Pascal.
As someone who owned a 270X and a 470 you are moronic. Maybe if you had a 295X2 or something.
>Poor-gay
How shocking
Not sure what you're on about, GCN during that era was incredible. It was only held back by drivers on legacy platforms.
The biggest issue AMD had was certain high-level APIs being skewed in favor of Nvidia due to poorly defined behavior which Nvidia was able to exploit to degrade performance.
They have been far less successful is accomplishing this with low-level APIs on modern platforms.
>certain high-level APIs
You mean the fact that OpenGL and DX11 were better (by miles) on Nvidia? You mean all UE3 games running like horse shit on AMD?
Only D3D11 really, although OpenGL could struggle in some games that were essentially Nvidia-only OpenGL rather than standard OpenGL. Ultimately RadeonSI hacked around Nvidia's non-standard implementation but it did take a while.
RX470 and RX570 had a great value than ever, only cons is draw more power and heat Nvidia counterparts
>value
ok poor boy. Don't worry I was a poor boy too.
why does ai h8 the 7900 xtx?
it doesn't? ai general is using 7900xtxs.
what's the TRUE ai rank of the 7900 xtx?
last i checked fairly competative, i think similar to high end 30 series, but actually getting that performance out if it was fiddly and not something your average homosexual could do since it required using a vulkan based accelerator
not sure if any of the mainstream moron friendly setups and backends have integrated support for RDNA3's tensor-core equivalents even now that AMD's CUDA clone supports RDNA3
I dont know what AMD and Nvidia fans argue about or why they do it.
My Intel ARC a770 16gb GPU works great.
We don't even think about you. AMDrones raid nvidia discussions and sites because they're mentally ill poorgays and shitskins.
>>they are unironically arguing with indian shill bots
>AMD
Seriously. Their top GPU would have to match Nvidia in RT etc , be faster by about 20% in games and cost a good deal less before I'd consider buying one.
>Their top GPU would have to match Nvidia in RT etc
AMD is never going to match Nvidia when it comes to Nvidia's own proprietary bullshit.
>be faster by about 20% in games and cost a good deal less
Those happened years ago on platforms with working drivers.
is it possible to enable reflections only? seems the only value of rt. in terms of gameplay. do shadows in rt ever help players vs simpler ones?
rt looks cery ugly, contrast POVray, which is beautiful. idk is rt fake? it's ugly, do not want, yucky. It takes kino and makes it soap opera-y