The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.
I see that nVidia gfx cards have this thing called CUDA cores...what is that? Does ATI have something similar?
besides from miniscule gaming performance difference between the two, is there any other difference? I know nVidia has something called PhysiX...does ATI have something similar?
Will they both perform the same when using CAD/RENDERING software?
Thx
p.s. I am talking about consumer grade gfx cards..not workstation cards...
The new generation of Graphics Cards feature programmable streaming units (formerly called shaders). That means these shader units can be used for simple programmable functions. Some of these for filters used in 3D rendering, or physics in games. This is often referred to as GPGPU. PhysX is CUDA accelerated Physics Engine. But it is also ported to x86 to run on CPU. To play physx games, you do not need Nvidia, as you have to install PhysX software anyways. This is among many reasons why PhysX is not an optimal choice, it's another layer of performance hinderance.
You have to remember, programmable shader units are simple, they cannot do complex functions that the CPU is able to tackle with ease. But GPGPU can be used if you need to crunch tons of simple calculations. Security firms are onboard with GPGPU, and some education firms if they have lots of calculations to crunch. But don't get into the hype, do the research, because that's what it is, it's hype.
CAD work is OpenGL which is a API. It's like Direct3D from Microsoft. OpenGL is an open source API maintained by the Khronos Group. This means it doesn't matter if you have AMD or Nvidia. OpenGL is not specific to either and both graphics card support OpenGL 4.1, the latest revision.
As for rendering you may want to research CPU vs GPU. GPU for some simple rendering with simple functions and simple calculations it's very fast. But you lose out on the complex filters and quality.
My suggestion if you buying a new system. It's cheaper to find a machine that has a powerful i7 Sandy Bridge Processor. Just find whatever you can get for your money on the CPU. As long as it has a AMD or Nvidia graphics you will be more than fine. The Sandy Bridge is powerful and can handle anything you will throw at it with superior quality than any GPU rendering can offer. High end GPUs for high end gaming are expensive and power consuming. You don't need this hinderance if you are not focused on gaming but focused on CAD/Rendering. Any Nvidia/AMD graphics can do this as long as you have a powerful CPU to render this with optimal quality and flexibility.
For a simple function that streaming processors can rip through for your rendering, you won't need 800 of them or 300 from Nvidia. You only need a few as the bulk of the work to maintain quality will be done on your CPU. This is much like encoding right now. You can encode on the CPU but run simple filters that are faster on the GPU. It's a hybrid of work. That's what you should be aiming, not all just GPU or all CPU.
It's almost unanimous across the board for rendering software companies, GPU rendering is unacceptable for final production work.
Problems? See this thread at archive.org.
