A computer for testing and optimising the robot (Expert Advisor). - page 13
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
The problem has been observed for a long time and not only on our Expert Advisor.
Memory leakage is a serious problem. Steps to reproduce give - there should be a result.
The point is that 8 threads on P cores are one and a half times faster than 8 threads on E cores
E-core has a whole core, they are not multithreaded, I inserted a screenshot6 when all 24 agents are running smoothly, where did you get 1.5 times?
https://c.mql5.com/3/423/8v5i05crveq_2023-11-28_024849045.png
what kind of CPU do you have now to measure the speed?e-core has a whole core, they are not multithreaded, I inserted screen6 when all 24 agents are running smoothly, where did you get 1.5 times?
https://c.mql5.com/3/423/8v5i05crveq_2023-11-28_024849045.png
what kind of CPU do you have now to measure the speed?Meaning that a single P core running on a single thread is 1.5 times faster than 1 E core running on a single thread. I.e. if you have 8 meta threads on 8 large cores, which will be distributed one thread per physical core, it will be 1.5 times faster than 8 threads on 8 E cores. Firstly, because of the high frequency of P cores, secondly - because of the number of ports, branch predictor, larger queue in the scheduler, etc.
here is a person overpaying for memory, even though his memory controller does not support such a frequency. I didn't go further into it. It is better to read the theory than to listen to someone.
In the end, i7-14700K (went Sdeck-shopping 2.5 months!), Asus PRIME Z790-P WiFi, Kingston FURY Renegade Silver [KF560C32RSK2-64] 64 GB.
When overclocking the memory to 6800 without errors withstands multiple passes of Memtest86+, at 7000 individual errors.
I stopped at 6400-32-38-38-80-1.35V.
Air, 8 fans in the case, at low-medium load on the CPU, up to 60°C, they are not heard at all, I picked up the profile. When running Memtest86+ for many days, the temperature does not rise above 77°C, then, of course, it becomes noisier, but only on some tests (3, 7 and 9).
I haven't put it to work yet, I'm planning to have time before the March expiry and trade the June ones on a new machine.
But, once again, I do not want to optimise, but mainly to work on real charts of tens of thousands of robots in test mode and dozens in combat mode. Well, and for preliminary testing of new ideas in the tester for general suitability.
not to write martingales, grids and over-sitting :-)
or add to their parameters "trade completion date and position closing date", which should be set during testing.
not to write martingales, nets and overcalls :-)
I don't know how to do it differently))))
or add to their parameters "trade end date and position close date" to be set during testing
Thank you!