Support Board
Date/Time: Sun, 16 Mar 2025 02:46:47 +0000
Post From: Parallel replay on CPU and GPU
| [2021-12-08 13:09:29] |
| User785723 - Posts: 58 |
|
I am running 2 Sierra Chart instances, each with 19 sub-instances so a total of 40 Sierra Charts in parallel on an AWS EC2 machine with 64 CPUs. The CPU usage is below 80% but the replay runtime on each parallel instance is 5-6 times more than the case when just one non-parallel instance is replaying. The purpose of parallel replay is for hyperparameter tuning. Does anyone know why the replay speed gets so low when the number of parallel instances increases although hardware resources are not full? Disappointed by an efficient parallel run on CPU machines, any ideas on how to run Sierra Chart in parallel on GPU and if this helps in terms of replay speed? |