A computer for testing and optimising the robot (Expert Advisor). - page 8
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
use Genetic Optimisation in MQL5 Cloud.
There custom characters are not available, which in particular speed up optimisation by several times.
I don't use them.
Speeding it up a few times doesn't do anything. It's not enough.
I've certainly thought about buying a comp with 96 or 128 cores, but crazy prices, up to $20K.
They use a 1200 watt power supply for them.
I don't use them.
Accelerating it a few times doesn't do anything. It's not enough.
I certainly thought about buying a comp with 96 or 128 cores, but crazy prices, up to $20K.
They use a 1200 watt power supply.
So why buy one top-end one? Buy a farm on processors. or just a lot of computers and network them!
PS Does no one feel the difference between allocating 100-300 rubles a week for optimisations and freezing 200 000 rubles in hardware?
At the same time you can regulate the time of optimisations by normal scaling.
In general, a collection of demo millionaires!
So why buy one top-end one? Buy a farm on processors. or just a lot of computers and combine them into a network!
PS Doesn't anyone feel the difference between allocating 100-300 rubles a week for optimisation and freezing 200 000 rubles in hardware?
At the same time you can regulate the time of optimisation by normal scaling.
In general, a collection of demo millionaires!
I don't understand anything )
So why buy one top-end one? Buy a farm on processors. or just a lot of computers and combine them into a network!
PS Doesn't anyone feel the difference between allocating 100-300 rubles a week for optimisation and freezing 200 000 rubles in hardware?
At the same time you can regulate the time of optimisation by normal scaling.
In general, a collection of demo millionaires!
There is the most important point here, which for some reason no one has ever said,
which of us are actually testing hundreds of parameters, you're testing metres through 2mm.
you can't jump faster than a single per thread/core, and in any VPN there are dead cores, my home one will be two/three/four times faster,
it follows - if you have a 24 thread processor, and you can buy a 13900, it will be much, much faster.
I personally test not tp with distance from 20-100 in 2 point increments, but with parameters, there functions are connected.
---
dead cores, dead memory by latency, there may be 4 systems hanging on one of them, even though the frequency of 4.5 is indicated (it's just like on the fence).
okay AI data set, but it doesn't seem to be done by proces at all there, v/k cards make special ones, and ordinary ones will be better.
there's a major point that for some reason no one's ever said,
which of us are actually testing hundreds of parameters, you're testing metres in 2mm.
you can't jump faster than single per thread/core, and in any VPN there dead cores, my home one will be two/three/four times faster,
it follows - if you have a 24 thread processor, and you can buy a 13900, it will be much, much faster.
I personally test not tp with distance from 20-100 with 2 points step, but with parameters, there functions are connected.
---
dead cores, dead memory by latency, there may be 4 systems hanging on one of them, even though the frequency of 4.5 is indicated (this is just like on the fence).
okay AI data set, but it doesn't seem to be done by proces at all, v/k cards make special ones, and ordinary ones will be better.
I guess you haven't used VPS for optimisations for a long time. If you take the veteran zomro, there is no overceling. But everything runs on DDR3. and optimisation is quite fast. Almost like at home. in other counterparts even write on what processors run VPS. and very few people on VPS use software that loads processors at 100%. Usually the load is a few per cent. Take even me. For trading I use VPS and if not optimised, the load spikes up to 20%. There are several terminals and Expert Advisors are not the easiest.
It must be a long time since you used VPS for optimisation. If you take the veteran zomro, there is no overceling. But everything runs on DDR3. And the optimisation is quite fast. Almost like at home. in other counterparts even write on what processors run VPS. and very few people on VPS use software that loads processors at 100%. Usually the load is a few per cent. Take even me. For trading I use VPS and if not optimised, the load spikes up to 20%. There are several terminals and Expert Advisors are not the easiest.
As for the speed of work of indicators, it's complicated here, indicators from the terminal just fly, but who knows how to make such indicators in reality?
Most of them, except for the cycle and pre-cycle, seem to work at the level - they seem to work and that's it.
I'm at the level of poking and prodding there too.
and loads the processor, I have a good volume load and through the indicator spy on 7 symbols 1-2%, for him a separate fanless mini pc,
and modern computer as I wrote, it's a home PC, but it can in one pass test in 20 minutes all my options. Almost any VPS, especially a penny VPS, will test an hour in one pass.
If I need more than 24 passes at a time, I'll buy a 13900 processor or more.
--
Dimitri You don't think that 50 cores will process 24 jobs faster, especially dead ones.
+
Our then it does not concern the service, there the user's processor is loaded and for this it is charged, but as in those systems) if it is a server, frequency is deplorable, memory is deplorable, and a number of other savings can.
while I was chatting, the yen got stolen.
I couldn't sleep because of it.)
If you are constantly optimising different variants and algorithms day by day... then renting servers by the minute or daily is unprofitable for you. especially if we talk about optimising on real ticks.