Should i buy Nvidia or Amd

Hi logan, hi community.

At the moment I'm building a gaming/3d modeling computer for my self and i just want to ask you guys for your advice on what graphics card should i buy.

My specs are:


asus sabertooth

16gb ram 1866hrz ram

and  i'm planning to buy a graphics card now but i don't know which one will be the best for the pc i want to build.

Should i buy for Nvidia for the cuda cores and the advanced physixs or should i buy a new amd and later on buy one nvidia quadro as well and have them both on my rig. (I don't know if it's possible but my idea is when i want to play game i will just activate the amd from the device manager and the other way around for the 3d modeling. but i don't know if it's possible) (also if i do the second one there a slight chance that i will buy a quadro, because i don't really want to spend to much on this rig).

Thank you very much guys and sorry for if i made any mistake with my English, it's because it's not my first language and I'm still learning.

err ... So I take it you need cuda apps to keep working?

if you use software for your 3d modeling that take advanice of cuda cores then buy an Nvidia card. if you use software that can also use open CL then go AMD.

I'm not really sure what are you trying to say but ill say thay... I'm still a student and i just checking to see which one of this graphics cards is going to give me the best performance when i'm using maya, zbrush, and 3d max.

Also i think you need cuda cores for animating.

Well i'm sure maya uses openCL, but i'm not sure about zBrush and 3dsmax

okay you have to do some research on that thing then.  because if both the program´s you use  can also use open CL then i would recommend an amd card. because they contains more stream processors. then Nvidia cards have on cuda cores. most of the time.

A good question here is how good are the cuda cores of gtx 780 gonna be over the R9 290x

thats a question nobody can ansewr yet, because the cards are not out or benchmarked yet, but in line of expecting i think that the R9-290X will be a very close call to the Titan. and probably will beat a GTX780. but thats just rumours. untill real benchmarks shows the numbers.

About 3d modeling and stuff, still the question if  all the software you use can also use open CL then i think the AMD will be faster.

I don't know i think i will go over to Nvidia side because i want those cuda cores and PhyscX. Because you can use those PhyscX not only on games but also on 3d modeling for games

Given that OpenCL is more efficient than CUDA, decide for yourself based on the latest objective benchmarks by openbenchmarking:

Given that a Radeon HD 7850 is under 200 bucks and performs roughly 2.5 times better than a 400 bucks GTX 680, I think it's nowhere near a close call or a difficult decision...

Ofc the amd could beat the nvidia in a linux environment with an open cl benchmarking

PhysX is no selling point. Even when it was released in 2004, processors could simulate it and take the load off of the GPU.

If anything is OpenCL enabled, then AMD is a strong option. High numbers of stream cores and raw computer performance really help when modelling. CUDA is really only helpful in image editing from what I've seen. 

Yup, PhysX software is not even an issue with half decent relatively modern CPUs, it doesn't put any significant load on the CPU.

Problem with CUDA acceleration in video processing is that the applications that use CUDA, or QuickSync for that matter, also reduce the image quality. CPU rendering still rules for maximum image quality, and acceleration through OpenCL, if set up right, offers the same image quality as CPU rendering. Typically, with CUDA-based rendering acceleration, complete frame refreshes cause all kinds of rendering artefacts, and with QuickSync, the pixel quality is deplorable. CUDA accelerated rendering applications may render faster, but that doesn't mean it's the same quality result as a CPU or CPU/OpenCL rendering by an application that uses those. Adobe knows why they are moving to OpenGL and OpenCL instead of staying with CUDA.

Wendell's dual Xeon + dual Phi rig actually makes a whole lot of sense for high quality single system video rendering. It's optimized CPU-rendering for best image quality, and fuck acceleration technologies in whatever applications, it will CPU render like a boss. The graphics card just doesn't matter anymore at that point.

Thanks Zoltan you really gave me a good piece of your knowledge