Can you tell me about the power required of our computers to use the software?
Can we generate a catalog of ".LAS"?
RE: informations - Added by Hackenberg Jan 9 months ago
HiTo give you an advice for your first question:
*Computree does not support any kind of graphic card computing (as far as I know), so you survive with a standard up to date graphic card for visualizing purposes only.
- The numbers of processors does only scale in linear time. Double this number and you half some Steps computation time. But only few Steps support multi-threading and only those will scale. SimpleForest QSM modeling can process multiple tree clouds in parallel, but often you can simply model data over night/weekend time in batch mode. As the scaling is limited and you can counter with good time planing I do not recommend investing here. For developers its a different kind of story - code compilation costs efficiency here.
- CPU type scaling cost even more, just plan better, but also go for lower costs. New processors are quite expensive, but do not bring you that much benefit.
- Memory, if you really want to invest money a bit, do it here. 4GB cannot handle larger, but still commonly used clouds. Nothing is more painful than running out of memory during the modeling.
- There is no general configuration without knowledge what you process and how much you want to process it.
- Get some test data fullfilling your needs and run CT on multiple different computers of friends or collagues. Just try out.
- If your project has good funding, dont limit yourself here. As I explained, you can go with a cheap up to date machine, but if you know you will do such computations on daily routine the next year and you have it you can also go for expensive machine -an uptodate gpu from nvidia, 6+ up to date xeon cores and 32+GB will cost multiple thousand Euro, but in such a scenario those costs are less than loosing multiple man month by waiting for computational results with you working machine blocked.
- How much you want to process: are you aiming for just a quick modeling run with default parameters or you want to do multiple runs
RE: informations - Added by Piboule Alexandre 9 months ago
To complete Jan’s statements a little:
For computer power:
- Computree is able to use graphic cards capabilities (for display optimisation). However, it’s able to work will small graphic card (just need to have updated drivers of the graphic card with recent OpenGL). But if you want to display big clouds, it is a good idea to have a relatively good graphic card (with enough dedicated ram). NVIDIA are a good choice.
- CPU is a matter of computation time, depending on amount of data and type of process you want to use. Some Computree steps are able to use multi-thread, but not all (depends on plugin dev.).
- For memory, scale it to data you want to process. I would say 4 GB is a strict minimum (probably too low in many cases), 8GB or 16GB are better. More if you want to manage big Terrestrial Lidar clouds. For ALS, as the process use loops, no need to go too high.
- Enough disk space is a good idea, ideally in SDD as data loading and exporting are often the longuest part of the process (especially in ALS).
But as Jan said, all depends highly on the use you want to do. You can first try on your current computer and see how it works.
For the LAS catalog:
- We currently do not exactly provide a “catalog”.
- For now we have a process to create buffered tiles, which are after used in loops in processing (loops are available directly in the GUI, no need to use batch mode).
- The only drawback is that for now the buffered tiles take some time to produce and occupy some disk space.
- However we are working on something a lot more quicker and optimized in space occupation, close to what you call a catalog (we are considering differents options).