Let me start with the summary:
CPU: i9 9900ks processor
Memory: 64GB
Hard Disk: 1TB solid state
Graphics card: 1660ti
LAN: at least gigabit
======================================
At that time, the company bought AMD , which is the number one on the ladder list . However, in the process of running the graph, a considerable part of the time is single-core operation, so the high CPU frequency is also a very critical point .
I have bought a graphics card that is more expensive than the 1660ti. Some people have tried the 2080ti on the Internet. However, I don’t know if it is the problem of CC and Pix4d or how to drop it. Anyway, it takes about the same time, and the graphics card is not used in the whole process. , So buying expensive for desperation will not increase too much income .
It also piled up more than 100GB of memory, but this 100GB of memory has not been used up once, because if the graph is too big, we will run it in chunks to separate the computational pressure to speed up the time.
Of course, there is another situation that when using Pix4d to run orthophotos, the 100GB is not used, because Pix4D has a problem with the program design, too many point clouds, it will exceed the limit of the Int data type, and thus crash.
So it will reduce the number of points to run, so no amount of memory will be tasteless.
This 1TB solid state is necessary, of course, on the main service of the CC cluster.
If you want to speed up the progress of the calculation, the only way to do this is to increase the host or use the computer of a colleague at night. A single high-priced host will not bring much benefit.