Best GPU

Well... i guess it depends on use case, cause if you are doing like deep learning stuff I am sure those AMD workstation cards can give nvidia a ran for their money

File: Screenshot_20170718-153747.jpg (381KB, 1377x1111px) Image search: [Google]

381KB, 1377x1111px

Yeah I mainly just love the way it looks. The price for this fuckin thing right now though, christ.

V O L T A

>>61519775
No rgb no buy.

File: 201404201840321.jpg (1MB, 2448x2448px) Image search: [Google]

1MB, 2448x2448px

The 7950. Still a relevant card to this day. Prior to that would be the 8800gt.

>>61519807
>deep learning
>amd
LOL, fucking dumbass

>>61519775
I use AMD because Linux, nVidia can fuck off until they fund an equivalent Open source driver, heck even their propietary drivers are garbage.

>>61519775

>three variable length telescoping rods

name a better GPU /g/

>>61519775
1050 Ti

It's honestly depending on what you want to do with the GPU.
But if we are are forgetting price here then sure that'll be the best GPU.

>>61519775
Semi-related question: is there any game that really benefits from having 16GB instead of 8GB, on a setup with 8GB GPU as well?

>>61520993
>what is ROCm
Also ASICs will make GPGPU meme irrelevant even in meme learning.

File: 3Dfx voodoo card.jpg (250KB, 958x680px) Image search: [Google]

250KB, 958x680px

>>61519775
No amount of green plastic will make any card as iconic as the 3Dfx voodoo 1 chip.
It is the chip that created the "master race".
Everything else was a joke next to it.

File: 9700.jpg (2MB, 2000x1203px) Image search: [Google]

2MB, 2000x1203px

>>61521256
*blocks your path*

>>61519775
XFX 480
/thread

>>61521336
oh shit nigga - is that 9600?

>>61521412
Acktually, it's 9700pro.

>>61521336
That monstrosity might indeed be quite a challenge.
I bet Nvidia still feels the pain.

>>61521416
eh, close but no cigar.
>silky smooth doom3 and half-life2
>tfw dedicated tv-out
what crazy times those were

>>61521417
>"ATi cannot compete! we're unstopabble!" - nvidia, circa 2002
Boy was nv30 a bad response to R300.

>>61521450
At least you can't hear people criticizing your videocard when you got an FX5800.

>>61521614
https://www.youtube.com/watch?v=qspdnAYiiug&t=12s
You couldn't hear fucking ANYTHING with this fucker working.

File: aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9SL0wvNjkzMDU3L29yaWdpbmFsLzA5LTNkcy1NYXgtUGVyY2VudGFnZS5wbmc=.png (28KB, 711x533px) Image search: [Google]

28KB, 711x533px

>>61519775
Vega FE is comparable to P6000 in rendering/pro tasks while costing 5 times less AND without certified drivers.

Vega WX Pro will bury it.

>>61521642
Yes, but AMD needs to do the same thing they did with EPYC: invest in ecosystem.
Also Vega10 supports SR-IOV as a nice bonus.

>>61521651
Why wouldn't they do that? AMD isn't ran by 2 people, ecosystem is crucial in both CPU and GPU.
Gaming or Pro.

Rome wasn't built in a day, and AMD has started a lap behind against Nvidia in this case

>>61521672
They completely ignored "good drivers" and "good ecosystem" part of GPUs before pajeet took over.

>>61521185
>meme learning
Classification algorithm which infers whether X is a meme.

>>61521688
Yes, good management helps.

>>61521642
Why are people calling this thing a failure?
I hear people screaming how Raja should be fired and how they should have never launched Vega.

>>61521721
>Why are people calling this thing a failure?
GPUs are mostly for kids and kids are retarded.
>Raja should be fired
>R300 dude
>should be fired
Kids. Fucking kids.

>>61519775
it would be if it wasn't 500w

>>61521721
Some "people" think GPUs can only be used in gaming, so they're left confused when it doesn't do good in gaming but costs a lot.

tl;dr children

>>61521780
It's usually somewhere along 300watts.
3840 ALUs are not free you know.

>>61521780
It's somewhere between 200-300W depending on workload.
Pretty efficient.

>>61521801
>>61521801
For $6000 it better fucking be.

>>61521809
It's still full GP102, you pay so much burgers for nvidia tax, ECC, moar memory and tech support.
And certified drivers.

>>61521013
>heck even their propietary drivers are garbage
I disagree, NVidia's binary blob GNU/Linux drivers are totally fine. The only big and important issue with them is that updating the rest of the system will break your desktop and you have to re-install their driver. It also forces you to use older kernels and sometimes older xorg versions. AMD cards just works - except for HDMI audio - without installing any binary blob bullshit and it doesn't break when updating. That's why I prefer it.

>>61521736
there's some "people" thinking that R300 refers to the R9 3XX series too, take that into account

>>61522814
These people must be extremely new or retarded to not remember classical ATi codenames.

>>61519775
GP100
/thread

>>61522881
>less FP32 ALUs
HOL UP

>>61522881
Useless unless you do fp16 or fp64
It's effectively slower than a gp102 in anything not using those.

>>61521786
>Some "people" think GPUs can only be used in gaming, so they're left confused when it doesn't do good in gaming but costs a lot.
Workstation gpus have external factors surrounding the hardware to justify their costs. AMD always delivers on the "hardware", and marketing side of gpus, but completely fail on making that hardware useful. Workstation AMD gpus are as useful as sticking a brick in your pcie slot.

>>61523006
Yeah, that can never change, ever.

>>61523023
Yes, goy!
Buy Quadro.

>>61523006
Seems like AMD is putting enough effort into ecosystem recently, does that not matter anymore?

>>61519875
>meat honey
>very likely that nobody in history has ever actually tasted it before
>it could taste amazing
>it could taste absolutely horrific
>it could be harmless
>it could be the most potent toxin a human could possibly ingest and 0.0005g will kill you on the spot

>>61523023
>Yeah, that can never change, ever
It could change, but amd never gave enough shit to.

>>61523058
>does that not matter anymore?
It matters after it matures, but right now no.

>>61521336
Unironically this

>>61519775
Intel HD graphics 4000

File: 1499411187833.gif (931KB, 300x300px) Image search: [Google]

931KB, 300x300px

>not supported by nouveau
fooken dropped m8

>>61524024
>using nouveau
Just buy fucking AMD is you want opensource drivers.

>>61524033
My GTX 970 does that just fine thank you
>inb4 3.5GB

>>61524080
>3.5
Ha
HAHAHAHAHAHA

File: 1485420006055.jpg (22KB, 360x336px) Image search: [Google]

22KB, 360x336px

>>61524080
>he bought the meme card

File: 1500827625684.webm (3MB, 1280x720px) Image search: [Google]

3MB, 1280x720px

>>61519775
>not the hbm2 ones

File: Screenshot_2017-07-23-14-02-35.png (49KB, 1080x138px) Image search: [Google]

49KB, 1080x138px

>>61524096
>>61524134
works on my kernel faggot

>>61524220
how is this shit even profitable? what are those cards doing ?

>>61524847
> how is this shit even profitable?
It's not.
>what are those cards doing
Number crunching.

>>61524847
Running linpack


tl;dr nothing of real value

>>61524847
shilling is always profitable

File: ss chrome (2017-07-23 at 11.55.55).jpg (97KB, 511x169px) Image search: [Google]

97KB, 511x169px

>>61523089
Some people have already tasted it. It's supposedly "sweet"

>>61520991
>7950
My negro.

>>61524847
https://www.youtube.com/watch?v=3RqF8m65r8g

Nvidia Volta GV102 TITAN V
Nvidia Volta GV102 GTX 2080 Ti
Nvidia Volta GV104 GTX 2080
Nvidia Volta GV104 GTX 2070

OP BTFO

>>61526636
Bigger dies? Wew.

>>61526650
Only the GV100 will be big

Aerospace Engineer Student here, while quaddros where better than geforce for the more technological tasks, a graphics card like the titan can do tasks just as well as the quaddro and only for lower end cards is a quaddro actually any use

>>61527284
Everything will be at least 20% bigger.
That's the only way for moar perf for now.

>>61524847
it's probably not profitable, it's a government program so they're running some biowhatever simulations or other stuff on the GPU. They don't need to be profitable, just help mankind.

>>61527322
But Pascal is faster with less cores and smaller dies than Maxwell

>>61527499
It's clocked much higher and moar ALUs for every SKU.
GP104 is ~300 ALUs less than fucking GM200.