>tfw waiting a few extra months for my new rig paid off Get

>>522039
I bought a $500 video card once. I felt really dumb when I got it desu. A few years later it was so worthless I just threw it away.

>>522039

Why did I buy this crappy 970. Getting a 1070 and 2tb ssd by the end of the year.

>>522040
>A few years later it was so worthless I just threw it away.
That's any technology.
Based on this logic you shouldn't be buying a SSD ever because the prices are coming down like the Hindenburg all the time

>>522040
>A few years later
$500 to work comfortably for a few years is pretty good desu senpai

>>522044
It literally got dominated the next year. What a waste. Now I max out at $200

>>522047
~350 is my tolerance border. A new GPU must be more energy efficient and 2x as powerful as the old one.
I have an old GTX660 probably will get a 1070 at the end of the year. Or maybe an ATI card, i don't know yet.

Why would I buy nvidia? Cycles opencl is getting better than cuda by the minute, vray has opencl, octane will get opencl in the next version. There was also that time when the 390x beat a titan in opencl cycles. Is there any info on the new ati cards, I wanna buy a 390, but I want their prices to drop first.

>>522098
amd has no drivers
>opencl
muh sides

Aren't gaming GPUs not designed for the demanding task of rendering?

>>522100
gtx 1080 is the fastest consumer card ever created

>>522039
>modelling rig
hahahahaha
i work on 20k rouble laptop

>>522102
>compensating for babby benis

>>522104
It's not that my penis is small.
It's that shit's expensive and I've got no money.

>>522101
I know but aren't workstation GPUs like titans specifically designed for things like rendering? I know they can both do the job well but my question is isn't a gaming GPU not designed for the highly demanding workload like rendering? Like Quadros.

>>522106
That used to be the case when people wanted to load up a CAD project of something which had 20k-100k+ objects and they needed interactive FPS. Now with command lists (NVcommandList) you can get 100k objects in 1080p at 120fps on a 750ti, This'll probably be core in openGL 4.5 and is similar to Vulkan performance (but better in vulkan)

https://github.com/nvpro-samples/gl_cadscene_rendertechniques

https://github.com/nvpro-samples

>>522107
*4.6

>>522099
http://blenderartists.org/forum/showthread.php?376184-How-is-it-that-the-R9-390x-beats-GTX-TitanX

>>522104
>american education
20k rubles is around 300 dollars

>>522112
3 copekh bill still?

I've been running a 680 for the last ~4 years. I'm tempted to get one of these after a few manufacturing cycles and driver updates.

I don't really understand how much of a pleb somebody can be to argue against <$1000 investment for a product that will last at least 2 years, that i will just claim as tax deductable anyway.

Have been using my 4 year old potato rig with a 660 Ti to learn this shit, now it's time I need to get a serious one, if I don't want to wait half an hour to render in Cycles or Arnold for a test-screen.

I'll probably get a 1070 this year and sell it next year to get a 1080 Ti or whatever is the best option for 300-400 bucks by then.

>>522122
>f I don't want to wait half an hour to render in Cycles or Arnold for a test-screen.
You should just buy a subscription to a renderfarm or get a pay as you go one. It'll be far cheaper. Rendering without a farm is madness tbqh.

>>522126
What if I built 3 computers render through networking?

>>522122
What are you rendering that Arnold needs 30 minutes for a test?
Are you sure you know how to use Arnold?

>>522111
>http://blenderautists.org

>>522149
That's a real site. Wow.

>>522148
I need my plates to be compositable in footage anywhere from Full HD to 4k

>>522163
use a farm. You don't have access? You must be a real poorfag then, maybe apply for food stamps

>>522149
Nice reply, you sure showed me.
The way I see it, opencl can run on both amd and nvidia, so why use cuda at all? Makes no sense.
Anyway, Vray, Lux, Indigo, Octane 3 and of course Cycles are all gpu renderers which use opencl as well.

>>522106

Gaming GPUs support most of the features you need for 3D rendering. They're not designed for that, no, but they can do it just fine. It's not like you really need double precision floating point just to make pretty pictures.

I don't know about commercial software, but at least with Blender you're much better off with a gaming card than something like a Quadro because that's what most users have, and so that's what the devs prioritize. A good gaming card will outperform the best Quadro on Blender, just because that's what Blender is coded to use.

>>522198

CUDA outperforms OpenCL on NVidia cards.

When Cycles was being created, AMD's drivers couldn't work with the Cycles kernel, and Intel wasn't really in the picture. CUDA was the obvious choice. OpenCL is still fairly new in Cycles (well, _working_ OpenCL anyway), so it's still immature. I know if I had to drop $600 on a card today, I'd buy NVidia because I know Cycles has good support for it.

Give it a couple years and I doubt it'll really matter.

File: 1462842269817.png (299KB, 480x480px) Image search: [Google]

299KB, 480x480px

>>522039
>6770
>Bought it 6 years ago for $150
>Still runs most games on medium settings

Still want a new card for them cuda cores though. I pretty much gave up 3D that wasn't technical drawings because of render times.

>>522544
Me too, friend

This already motivated me enough to dust off my cgi-software knowledge.

>>522094
this is a good metric to go by, i also don't buy the latest hardware because of how hard the prices go down