Jump to content

What Features You Are Waiting For C4D (Wishlist)


Recommended Posts

Sorry if i sound a bit offensive here, but i am a bit tired of people asking vehemently for things that they actually don't understand and don't know the implications of, let alone the fundamental technological concepts.

GPUs were created to accelerate the display of vector graphics. Over time shading and rendering functionality were added. However theses additions always targeted the display in the editor and often enough the huge display speed was achieved by cutting corners. First of all the models and materials had to be especially prepared by game artists and engineers, second there are severe limitations when it comes to the number of lights, the quality of shadows .......

All this means that you basically can't use rendering and display functionality of the card to produce production ready renderings. Instead the basic functions the GPUs offer are beeing crosspurposed for real rendering.

If a GPU has access to all information inside of it's own adressable memory it can act very fast, but once it has to access the main memory of the system (PC or Mac) there is a dramatic slowdown since this data has to be provided by the CPU first and has to be passed through the bottleneck that is named PCI express. Even the fastest PCIe version isn't nearly as fast as direct memory access. SImply increasing the memory on the GPU board doesn't work as well, It is extremely costly to provide a GPU with the amount of memory you routinely need for more complex renderings (>8 GB).

This leaves us with a niche solution that costs a lot of maintainance while having limited advantages. Yes, GPUs and the sourrounding technology are improving, but so are the needs of the users and the renderengines.

So yes, the GPU can do almost anything, but it is simply not the best tool for the job, or at least not in most cases.

Where GPUs actually are of huge interest are scientific calculations, where very complex and repetetive operations have to be done on a limited amount of data. Rendering however needs huge amounts of memory and there are often dependencies in this data that does not allow for easy segmentation, again something that isn't a big issue for scientific calculations.

Another issue is that not everyone can be expected to have one or more powerfull GPUs in the system so you as a software developer have to create two solutions, one that is optimized for CPUs and one that is optimized for GPUs. This leads to a big can of worms since both systems have to provide identical solutions with completely different hardware and of course the development costs are effectively more than doubled.

You simply have fallen for the marketing texts of GPU producers.

Link to post
  • Replies 157
  • Created
  • Last Reply

I believe that we exhausted the theme here so topic is closed unless Nigel decides otherwise :)

Nope, I agree, this topic has run its course. Nigel / 3DKiwi

Link to post


This topic is now archived and is closed to further replies.

This topic is now closed to further replies.
  • Recently Browsing   0 members

    No registered users viewing this page.

  • Create New...