Boris Lund Regular  Posts: 59 Joined: 12/9/2021  User Profile | I will share my findings regarding maximum performance for OT. It does not really support multicores so a giant multicore system is out of the question. But it does support dual core at least for the ARM training. I have both AMD 5800x and 5950x and they are performing equally despite the superior multicore performance from the 5950x. The most performance gain I got was from loading the whole program onto a 12GB ramdisk. For those interested in ARM training there is a giant bottleneck from writing to the arm file. The more rules it finds the more it writes to the file. Moving everything into ram I got a performance gain of 50%. Even from having the fastest SSD possible. Unfortunately nothing in OT benefits from more than two cores so buy the best singlecore performing system and load it onto a ramdrive. On https://www.cpubenchmark.net/singleThread.html I can see the I9 12900KF is at top right now. But that is only 20% faster than what I got, while utilizing multicores would likely speed it 500% up. We all wish for true multicore utilization in OT. Even I am not training ARM, I am optimizing a lot to find the optimal strategy settings with the optimal stops. If that happy day should arrive then we have AMD Threadripper 3995wx on the table, or even the possibility of rented cpu space since that is about scaling cores but it hits the treshold very fast with singlecore performance. What else apart from ARM could benefit from multicore support is ofc. the optimization but also the strategy scan. It would be awesome if it could grab 10 symbols at a time and parallel process them. | |
