What kind of graphics card does /3/ use for modelling?I am currently using an AMD Radeon HD 6700 just because I haven't felt the need for anything else. However, now I am moving onto photo-realistic rendering so I'd like a card with a shit ton of cores so 3Ds Max doesn't spontaneously combust on me.
I have a Radeon 7990 HD 3GB, but mostly for the VRAM when I am loading multiple 100+layer PSD files as textures in Maya.
>>515544What kind of processor do you have with that?
>>515530750 ti, card of the gods
A 3dfx Voodoo 2.
>>515557>not SLI'ing them BRUH
>>515548Dual quad core Xeon 3.0GHz, 32GB ECC DDR2. It's a Mac Pro.
>>515564Nice, I've been keeping my eye out for a decent priced one on craigslist. What GPU?
>>515566As I said above, Radeon 7990 HD 3GB
>>515564absolute shit tier station
>>515530i got a GTX 970 for this last christmas,
>Not modelling your lolis, dicks and other degenerate models on the go950m
>>515564AEKII? you selling?
>>515615Only if it's a swap for an AEK 1. The AEKII goes for like $50 on eBay, good luck finding an ADB>USB for cheap though.
4GB R9 290, I have 3 of them, but I've not attempted to enable the full use of them for Maya, only blender.
780ti Good enough.
980TiOnly because I use Octane render. Before that (when mainly used v-ray) 960 4gb vram. And that was overkill. But still cheapest of Maxwell architecture.So later will buy new motherboard (this supports crosfire only) and sli those suckers.
>>515741You should still be able to use both without SLI. At least, that's how it worked for my 980 and 960 with blender and Daz studio.I don't think SLI works unless it's the same model card anyways.
>>515744Whaaa? I assumed that only way to use power of two nvidia gpus is with sli.And even more - when mother board says that supports only AMD CrossFire. So I just didn't try out and lend that card to friend of mine who has worse gpu. So It's time to get it back and try it out.You had to do some hack or it is just "plug and render"? Besides render settings and etc.
>>515755For blender it was similar to pic related.Also to note:the max vRAM usable by Cycles is limited by the lowest amount of vRAM available on a card. If you have for example a GTX 580 with 1,5 Gb and one with 3 Gb and want to render in Cycles with both of them, you have only 1,5 Gb available, not 4,5 Gb.
>>515756also: http://blender.stackexchange.com/questions/31764/is-it-possible-to-use-different-cuda-capable-graphic-cards-to-render-simultaneouI'm sure similar rules apply for other renderers
>>515758thanks, man!Will try it out (as soon as I get it back) and if post will not be dead by that time, will send you virtual thanks and high five for saving time on rendering.Always wanted to use gpu render on two gpu (one on full load and other on low priority). While rendering still can use power of cpu for other works with secondary gpu. Or just on full load and faster. On same machine. Crazy!980 ti octane bench is 125960 gtx bench is 48Additional 30% on speed with no extra cost is a big deal (not talking about electricity). Only down side will be that I will feel stupid that I didn't try it faster. lolThanks again for info.
>>515530750tiIts cheap and gets the job done.
>>516068640 cuda cores though :(
>>516068Not if you want to do GPU rendering. You want something with cores in the thousands, that doesn't just get the job done, nut gets it done quick. Plus more RAM. You never want to go below x60 with Nvidia GPUs, because anything below that isn't meant for heavy graphics work.
>>5155302 gtx660's... Feels bad man
>>515768aaaaaannnnddd big thanks to you. Really appreciate info.It works like a charm. 980 Ti + 960 gtx with no sli cables on crosfire only moboNow I have to get my head around idea about SLI and what not. Because this just made me rally confused.Octane bench was 125 and now it is 184Pic related.
>>516085>>516105But that's if you're using CUDA, right? What about OpenCL?
>>516213Your question doesn't make much sense, raw processing power is raw processing power. And a CUDA app will perform even faster due to the special dedicated hardware features Nvidia's architecture has for it that OpenCL doesn't take advantage of due to it's need to be cross-manufacturer compatible.Let's compare the specs of a single step up over the 750ti:750ti640 cores - 1870m transistors - 16.3 GP/s fill rate - 88 GB/s bandwidth - 128bit memory bus7601152 cores - 3540m transistors - 31.4 GP/s fill rate - 192 GB/s bandwidth - 256bit memory busFor 67% more money ($100) you would have gotten around double the performance if not more in some situations. It's never a good idea to go below x60 for 3d work or gaming purposes
>>515564you know you're great when you have that kind of switch box
>>516551I've had it for about 12 years, and I absolutely love it.
>>516501wrong, 760 is kepler (older, completely different architecture) while 750 ti is maxwell. Maxwell allows newer extensions to be used and thus much faster paths to the metal that are exclusive to maxwell. Similar differences will be found going from Maxwell to Pascal - exclusive fast extensions
>>515564Picked this up the other day for 2 bills- 9.8x12 Intuos 4- a nice upgrade from my 6x8 Intuos 3.
>>516501>And a CUDA app will perform even faster due to the special dedicated hardware features Nvidia's architecture has for it that OpenCL doesn't take advantage of due to it's need to be cross-manufacturer compatible.That's kind of what I meant to ask. If you are using a renderer that uses opencl does it still matter as much how many cuda cores your gpu has?
>>516581yes. cores are cores, anon.
As someone who wished to get into 3D modeling and content creation I would like opt ask a few questions. What is a good (cheapest), minimalist card for a beginner who wants to start; not stretching for photo-realism at this time? The 750 Ti supports DirectX11.2; how does DirectX 12 play into things right now and going forward?
>>516773directx 12 is windows 10 only.
ive got a sapphire r9 390 (8gb vRAM), i7 3770, and 16 bg of DDR3 ram (2400)
>>516773DirectX or OpenGL versions don't really matter at all. 3D modelling software is very very slow on the uptake, maybe you'll see DX 12 features in modelling packages in 10 years.Additionally DX12 is very very different from DX11, so moving from DX11 to DX12 is not something you can do without basically touching the entire rendering stack inside the modelling software. Almost everything works differently in DX12.Other than that, it depends on the modelling package you're going to use; e.g. blenders Cycles renderer can render on the GPU; if you want to use that, and you want it to be fast, you better have a good nvidia GPU with tons of VRAM. Would recommend 4GiB or more, although you can get by with 2GiB if you don't go too crazy on e.g. the hair rendering.Other renderers may not take advantage of the GPU at all; it varies.
>>516958Would of replied sooner but had other things on my mind. So hey! thanks the reply and info. What I gather is weather a card supports DX12 at this time is not a selling point? Good to know and is some pressure off my mind. Again, I am just looking to learn and get started in Blender and SFM so I am not looking to "go too crazy." Unfortunately, I have to keep to a budget of $100 and have been eyeing a XFX R7 250 for a while. Could spend the extra $30+ for a 750ti. What exact differences does it make?Unfortunately, my monitor died this morning(currently borrowing an ancient thing from family). So I must spend money else where for now :(
NVIDIA Quadro M6000
Use a 290x, before that a 5850. All the renders I do (render to texture stuff) can only be done on the CPU so I can't utilise the GPU for much else.
>>517125If using blender, I would opt for a powerful nvidia GPU with a lot of VRAM (crucial for fast hair-rendering) -- 4GiB, preferrably more.
I'm late to the party but I'm running two 970's in SLI with an i7. Gets the job done.