A computer for testing and optimising the robot (Expert Advisor). - page 11
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
I suggest to make a measurement, take the robot with codebase, if there is a wattmeter, it will be great at all
I can try. In Win you can display watts with some Afterburner, in Linux - I haven't met such utilities....
The only problem is that the tester really drops into the background, and only if the Opera browser is on the first background, the activity of the cores drops immediately, but this is easily solvable, maybe MQ is just not aware of this and need to show it
So far, on 12 series, I managed to solve it only through Process Lasso. In it, I forcibly hung threads of runs on large cores. If optimisation is heavy and I can't use all cores - I solved the problem by disabling E cores in bios. )
my home computer is i3 8300. Ram 16gb. When running optimisation on my cores for 20 minutes of optimisation the rented cores run about 20% more tasks per core than on my home PC. In this case the distribution of tasks is the same, i.e. start - all cores 8 tasks, then if 2 then all cores 2.
I for example 4 cores of my i3 is not enough. so I will just change the stone to i7 9b or 10-th generation. I was thinking of taking an i7 or i9 and DDR5 memory. But I won't optimise that much. and for the rest - there's definitely no point. Invest better in a wide flat monitor. I got myself one - I can't get enough of it. It's practically a double monitor. And the graph to leaf conveniently and in 1/3 of the screen code, in 2/3 - terminal.
Speed of deployment of servers -2-4 minutes. put Radmin and configure and metatester, taking into account copying another minute let 3. The service allows you to deploy several virtual servers at once. And everywhere you can run software in parallel. You can deploy 92 cores and 196 memory at once. But it is much more expensive. Standard sliced servers are much cheaper. I won't give the service itself so as not to get banned. But it is not a problem to find an hourly VPS in Yandex.
Well, it's clear, you are comparing an ancient 8300 (with slashed frequencies) and 16gb with modern hardware ... Since 8300, up to 12th generation (and even 11th generation), production per core has increased a lot. So even a stale VPS KVM on a fresh Xeon will give you more performance than an 8300. Plus, it depends on the specific test. If the period is long - the emphasis will go to the RAM and the meta, instead of storing the generated ticks in the RAM, starts saving them to files.... Therefore, you should compare modern hardware with VPS.
So far, on 12 series, I managed to solve this only through Process Lasso. In it I forcibly hung threads of runs on large cores. If the optimisation is heavy and it is not possible to use all cores, I solved the problem by disabling E cores in bios. )
There is with the help of utilities to solve this problem, there is made a change in power management, these settings are hidden, so far I do not need it
I don't want to spoil the system, there is a variant with increased system priority.
I wish the developers at their level to do it, I think there were reports on the forum that there are applications that work as intended - do not fall in the background
ps
actually the problem is not global, for some reason I have only if opera browser on the first background, and if for example a film to watch, then works at full speed tester...
So far, on 12 series, I managed to solve this only through Process Lasso. In it I forcibly hung threads of runs on large cores. If the optimisation is heavy and it is not possible to use all cores, I solved the problem by disabling E cores in bios. )
what is the point, e-core is equal to one thread of a large core, I showed above that 24 threads go exactly at the same level in the tester
64gb for 24 threads, in multicast is enough for OHLC, not enough for ticks for sure
according to my tests one core (multithreaded) consumes up to 12gb of RAM.... 16 gb of memory is only enough for opening prices, for each tick you need to multiply RAM by at least 5 for one core.
What are you running with such a wild consumption?! I made a screenshot of the optimisation process on real ticks for three years. As soon as the optimisation ends, a new one starts automatically on a different symbol.
18 Agents, each consuming less than 0.5GB. And one more terminal runs single runs from calculated opt-files in parallel - consumes up to 1 gigabyte.
What are you running with such wild consumption?! I made a screenshot of the optimisation process on real ticks for three years. As soon as the optimisation finishes, a new one is automatically started on a different symbol.
18 Agents, each consuming less than 0.5GB. And one more terminal runs single runs from calculated opt-files in parallel - consumes up to 1 gigabyte.
I have 7 symbols working in one pass, I chose not to adjust for each symbol individually from the very beginning.
The tester is not perfect and consumes you know how.I have 7 characters working in one pass, I chose the way from the beginning, not to adjust to each character individually.
In this mode, the regular tester would only be used for single runs. It is insanely slow in multisymbol mode.
In this mode, the regular tester would only be used for single runs. It is insanely slow in multisymbol mode.
I mostly see that the system works as designed, usually not more than a month.
in optmise mode only true/false are used and there are not many two variant int ranges available
I mostly see that the system works as designed, usually no more than a month.
This is how they discuss multi-core optimisation in the thread.