Hello. We have got an AMD Ryzen Threadripper PRO 5975WX 32-Cores, 64 RAM and NVIDIA GeForce RTX 3080. Once all components and updated drivers installed, we have made a speed test. We are far away from a same processor as is seen in the attached image.
I have several questions:
-Comparing the two performances in the benchmark, the faster is with version FLAC3D_700.154, should we change to that version?
-Could we get more speed from the graphics card? It seems that is not providing much.
-When testing for the benchmark, CPU was running at 100% utilization however our model not (much smaller) as we can see in the image above. Could we do something to improve CPU utilization?
-Hard disk is encrypted and has an antivirus installed, could it be a problem?
-Would the processor perform better using Linux?
The specific version of FLAC3D shouldn’t make much difference. You will see some natural fluctuation in the speed test, even if you run the same version on the same computer twice.
A better graphics card will not speed up your simulation
Thank you Jim for your answer. We are going to try installing eight 16 GB RAM modules and populate all eight RAM slots on the mainboard as said on the website. Regards.
I would suggest using the same machine to test, although I noticed that these two machines are with the same essential hardware specifications, but others, e.g., software (OS et al), RAM usage, even machine temperature, can have some effect as well.
BTW, my test with my machine on these two versions showed almost the same speed.
Now the supercomputer is up and running reaching the 4th position in the benchmark. The success has been to fill all the free RAM slots, having 256 GB in total. Thank you Itascans!
Hello Chen, I recently use the 64-core (128 logical processors) 5995WX to run models and found FLAC3D could only use 64 logical processors. Is Itasca will improve this in the near future?
In theory, FLAC3D should be able to utilize as many threads as are available on the machine. If this is not the case, you can type “program threads auto” in the FLAC3D console or include this command in a data file before running the model.
Thank you very much for your quick response, and I really appreciate your support. I added the “Program threads auto” in the data file, but it still works as before. When I run one mine-scale models or two single pillar models like this screenshot, the software just uses 100% on half of the cores. When I run two mine-scale models, all the 64 cores can be used to 100%. Therefore, I am thinking there may be a limit on the number of cores for each model.
I don’t think FLAC3D has set a limit on the number of cores for each model. Currently, we don’t have a similar machine with 128 threads for double-checking. Some processors have an optimization to use only partial cores if the program or model doesn’t require all cores. We’re not sure if this is applicable to the 5995W or not.
