Managing textures

Hello folks,

A while ago I posted a topic about waterfall displays and you guys gave me great advice so I’m hoping I can get more of the same! :slight_smile:

For the system I’m working on, the worst case scenario would be to have 32 2048x1024 waterfall displays plus various other plots and symbols at once. All the waterfall displays need to be updated even when not on screen and when they are required, their data needs to show up in a flash!
Question is, do all these texture reside on my graphic’s card memory? If that’s the case then I’d need to have a massive amount of memory on my card since the system is dual display and each display can do whatever it needs with the info…
If this is the case (that each texture would stay in memory) would there be a way to transfer the information back and forth from the PC’s memory to the graphic card’s?

For example, out of the 32 waterfalls, let’s say you can only have 3 up at a time, in theory, you would only need to update 3 textures, but update the 32 buffers on the pc (as opposed to update all 32 textures). Then if I change page, I overwrite the 3 textures with stored data from the pc. Is this feasable if I need the updates to happen FAST?

Is there a good way to do this? Or am I just missing something completely? :confused:

Let me know if I’m not being clear! :wink:


Sounds like what you are doing is a custom program for a client and not meant for wide distribution. If so I would just set the requirements for 512MB card and be done with it.

Anyway, your graphics card should start using main system memory for textures if you run out of video memory - So it should do a swap for you.


Thanks for the info.

I didn’t want to take any chances, so when asked my opinion, I suggested they buy a Sapphire HD 2900 XTX OEM with 1GB of ram… it’s definitely overkill, but whatever…

How will the swapping affect performance, though?
And is this done automatically?
As you probably have guessed by now, I’m really no expert…


Yes, the swapping is done automatically. Just pretend you have enough texture memory available, the driver is supposed to take care of it…

As for the performance, there is no definite answer to that, except that it will most likely be lower :stuck_out_tongue: .

The GL spec is (intentionally) not clear at this point, the details on how this is implemented are up to the driver. It may depend on a lot of factors, so the only way to be sure is to try it out :wink: