Crazy cache of testing agents - page 6
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
No, of course not.
Justify, please, of course I believe you, but there are facts...
Justify, please, of course I believe you, but there are facts there...
As long as you are trying to simplify the calculations and take data at the base by eye, everything will be wrong.
Read the claude fact, calculate repeatedly, discarding mega digits as wrong and then you'll be close. Not forgetting to fully justify the formulas.
It's useful to do the math yourself.
As long as you are trying to simplify the calculations and take data by eye in the base, everything will be wrong.
Read the claud bac, calculate repeatedly, discarding the mega digits as wrong and then you will be close. Not forgetting to fully justify the formulas.
It's useful to do the calculations yourself.
Well look here https://www.mql5.com/ru/articles/341
How do I know the cost of calculations I want to run in MQL5 Cloud Network? Won't the amount be too large?
You can run a test task with a running time of a few minutes. After its completion you can see the amount of money blocked for it in "Accounting" and estimate the cost of the whole task. For comparison, the cost of the optimization shown in the article"MQL5 Cloud Network speeds up calculations" is ~ 0.06 USD.
You can find it here https://www.mql5.com/ru/articles/341
Once the optimization is complete, log in the Journal and see that 14040 passes on 8 local agents took 1 hour, 3 minutes and 46 seconds.
Let's find the yield per second 0.06/(1*60*60+3*60+46) and multiply by the number of seconds in a year 0.06/(1*60*60+3*60+46)*(60*60*24*365) - we get that we will pay $494.55 for one year of offline PC.
In the test intel core i7 950 processor was used , I bought in 2016 3 PCs for 30k rubles generally comparable http://cpuboss.com/cpus/Intel-Core-i7-950-vs-AMD-Phenom-II-X6-1055T, i.e. just the same 0.5k $.
The service life of PCs is growing lately - let it be 3-5 years, so in claud I will spend 500*3*3=4,5k for 3 years - for 4,5k I could buy 9 PCs - and it would significantly increase the speed of work on the project. This is my point is that the price is still high for the cloud, for this reason it is not in demand. Personally I would be ready to pay a maximum of 50 c.u. for the cloud, similar to the PC usage per year.
...
So in the cludes I'll spend for 3 years 500*3*3=4,5k - for 4,5k I could buy 9 PCs - and that would significantly increase the speed of the project. This is my point is that the price is still high for the cloud, for this reason it is not in demand. Personally I would be ready to pay a maximum of 50 c.u. for the cloud, similar to the PC usage per year.
Have you forgotten about the landlord's interest? What he has you at cost should surrender their computers? What about the interest of the organizer (trading floor)?
So it is not quite clear what you are trying to prove. Owning an asset is always cheaper than renting it - that's a given.
Have you forgotten the landlord's interest? Does he have to rent out his computers at cost price? What about the interest of the organiser (the trading floor)?
So it's not quite clear what you're trying to prove. Owning an asset is always cheaper than renting it - that's a given.
The fact is that resources are idle, while people would like to get some money. We have a situation where there is no demand because of the high price of the service.
The commission to the organisers is fixed - 10% according to their statement.
In fact, now a man selling resources earns very little - I was able to earn $ 5 for a long period of time (several months - the work is very rare), and so you could earn a month stably money because of the emergence of demand.
Well, let's say 50 is not enough, but 150 is already a decent amount. 500 is too much.
Perhaps, if the interest in the service grows, the price will also grow, but right now we need to attract the interest of the masses.
And, then, I don't need 1000x acceleration - I'm interested in 10x acceleration and that should be cheaper. Right now there's no way to influence the cost of a service you buy, depending on its most important characteristic - the time it takes to get the result.
Do you use a claude yourself?
Would you rather make the same $5 but have your cores spinning round the clock? That is exactly what will happen if they lower the cost.
Have you solved real problems using the cloud? What is the point of speeding up the calculation by a factor of 10 instead of 1000? You want the process, not the result?
If it were possible, the result would be sold instantly, in one click. Instead of 10 hours at $20 an hour, I would pay $200 and get results instantly - that would be cool!
And if you only need to run part of the optimization, why run the whole thing? Limit the range of parameters.
You really don't understand your mistake, or are you trolling?
Take a simple monocurrency EA and run a real ticks test on EURUSD over the last year, look at the cache volume.
Take the same EA but run a test on GBPCAD and look at the cache volume again. For an explanation of the phenomenon of a threefold increase of the cache, refer to the magazine.
Now take the same EA and run the GBPCAD optimization on your 24 cores. Multiply the previous figure by 24. And while you are contemplating the number of zeros, think about whether you would agree that only one of 24 cores could work normally at a time (and the rest would stumble, trying to read the same block of data from the cache), or would you really like that after warming up (cache creation) the optimizer loads all the cores to the limit?
Yes, the hard drive can be a bottleneck in the tick data cache preparation process. Yes, it is not always reasonable to use all the cores in optimization (if there are not many tasks). But the tester does work well.
I'm not saying that MQ has nothing to strive for, or that Renat is right, always equally confident in defending his brainchild (even if he was wrong, and knows it). But you also include a little brain in the process of searching for the truth.
Respectfully.
Would you rather make the same $5 but have your cores spinning round the clock? That is exactly what will happen if they lower the cost.
Have you solved real problems using the cloud? What is the point of speeding up the calculation by a factor of 10 instead of 1000? You do not need a result, but a process?
If it were possible, the result would be sold instantly, in one click. Instead of 10 hours at $20 an hour, I'd pay $200 and get instant results - that would be awesome!
And if you only need to run part of the optimization, why run the whole thing? Limit the range of parameters.
Why $5? If I pay $150 a year it will be $12.5 per month - $5 I have earned in 4 months - i.e. I am expected to earn 12.5 times more per month or I can turn on agents for less than two hours and get the same income as before - saving on electricity.
I haven't solved any real tasks using the cloud - I'm just trying to figure out how much it would cost me to solve my tasks in the cloud compared to desktop PCs.
The thing is that development of ATS takes time not only for optimization, but also for analysis of results, making changes, adding algorithm, besides a person needs sleep - so it's important for me to take the free time between working on ATS and waiting for optimization results - now it is about 10 times faster - for my rhythm. That's why it's not clear why I should pay for extra speed that I don't need.
I don't optimize all of the parameters - all of them separately, and then, after analysis, I combine the best versions of the results of individual components in the overall optimization - this saves a lot of time.
The fact is that resources are idle, while people would like to make some money. We have a situation where there is no demand because of the high price of the service.
The commission to the organisers is fixed at 10% according to their statement.
In fact, now selling the resources one earns very little - I was able to earn $ 5 for a long period of time (several months - the work is very rare), while it would be possible to earn a month to a stable money because of the emergence of demand.
Well, let's say 50 is not enough, but 150 is already a decent amount. 500 is too much.
Perhaps, with the growth of interest in the service will rise in price, but now we need to attract the interest of the masses.
And, then, I don't need a 1000x acceleration - I'm interested in a 10x acceleration, and that should be cheaper. Right now there's no way to influence the price of a service you buy, depending on its most important characteristic - the time it takes to get the result.
Do you use the claude yourself?
Yes, I agree, it all comes down to the issue of fair pricing, and the problem is non-market pricing. There are buyers, there are sellers, but there is no market as the price is set by someone at the top. So there is no balance of interests and no efficiency.
In comparison, in Amazon's cloud service, the price is regulated by the market, according to the exchange principle. Buyers and sellers of resources can set their own prices or work at current market prices. Accordingly, supply and demand are always balanced.
And here they are inventing some clever abstract formulas to calculate the price, which has little relation to reality.
I myself have been using the cloud for mathematical calculations until recently, but then I started having problems with the cost of it all: the deducted sums sometimes did not correspond to the amount of work performed. I've still not been able to work this out with tech support. So for now I don't use it, or I use it very carefully.
Yes, I agree, it all comes down to the issue of fair pricing, and the problem is non-market pricing. There are buyers, there are sellers, but there is no market because the price is set by someone at the top. So there is no balance of interests and no efficiency.
In comparison, in Amazon's cloud service, the price is regulated by the market, according to the exchange principle. Buyers and sellers of resources can set their own prices or work at current market prices. Accordingly, supply and demand are always balanced.
And here they are inventing some clever abstract formulas to calculate the price, which has little relation to reality.
I myself have been using the cloud for mathematical calculations until recently, but then I started having problems with the cost of it all: the deducted sums sometimes did not correspond to the amount of work performed. I've still not been able to work this out with tech support. So for now I don't use it, or I use it very carefully.