Is an NVIDIA geforce 2GB card suitable for point cloud processing or would I need a 4GB. I think I read somewhere before that past a 2GB card there is not much benefit and that money would be better spent on RAM.
Totalsurv, post: 375682, member: 8202 wrote: Is an NVIDIA geforce 2GB card suitable for point cloud processing or would I need a 4GB. I think I read somewhere before that past a 2GB card there is not much benefit and that money would be better spent on RAM.
Assuming the software you're using is relying on the GPU to do the work (and it usually does), my advice is to DON'T SKIMP on Vram.
Get baddest Geforce (if that's you're preference) you can. I think you'll get more bang for your buck than the equivalent amount of ram in the machine.
That said, ram is so cheap now, you should load the machine up with that too.
hello Totalsurv,
i don't know if what you are doing is anything like generating point clouds that number in the millions of individual points through photogrammetry. but, if it is, then your GPU may be doing almost nothing during processing. my computer has an i7-3770 quad core processor, 16GB ram and a pretty ordinary Radeon HD 7450. when processing, the GPU does nothing. the CPU, however, is usually maxed throughout the whole process.
i recommend downloading Process Explorer to replace the windows task manager so you can monitor your system when doing your processing to see what part of your system is being maxed, and hence, where gains may be had. some software won't even request much CPU or RAM or GPU no matter how much you load the system with.
Look closely at the programs recommended spec. You may find that a Quadro card is necessary for the best performance. A high spec Geforce card is great for gaming etc but may not be best for working with point clouds.
Sent from my SM-G935F using Tapatalk
i7 processor with at least 16 cores, 32 megs of ram
best quadro with the most vram you can get
ssd hard drive(s) for software and data
The minimum system requirements for the software will not be enough to work on actual projects.To be efficient, you need to navigate through the data easily.
