MetaTrader 5 on Intel Xeon Phi 7250 - 272 cores in one computer - 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
I wanted to buy it last year, but I couldn't find it in the shops.
You're not looking hard enough. I just found one in Moscow in less than a minute for 282,000 rubles.
And you can get 32 cores for 0.3 quid an hour on Amazon... What's the point of buying a home server?
Better make AWS AMI with agents. Surely there will be demand...
And you can get 32 cores for 0.3 quid an hour on Amazon... What's the point of buying a home server?
Better make AWS AMI with agents. Surely there will be a demand...
Wouldn't the cloud from MQ be cheaper? After all, such power is only needed for optimization, and the cloud is probably more than 1500 agents, too lazy to look now.
The cloud from MQ has limitations, can be slow (waiting 30 minutes to get results after testing), and it's hard to calculate the cost of the cloud (still don't understand how 5 quid was spent).
The cloud from MQ has limitations, can be slow (I waited 30 minutes to get results after testing), and it's hard to calculate the cost of the cloud (still don't understand how I spent 5 quid).
Mmmm, I apparently had a simple algorithm, tested a few times, half a minute each, at a cost of no more than a quid. Well, it's an individual thing. And about"hard to calculate cloud costs (still don't understand how 5 quid was spent). " There, yes, only by experience, it seems.
And you can get 32 cores for 0.3 quid an hour on Amazon... What's the point of buying a home server?
Better yet, do an AWS AMI with agents. Surely there will be demand...
How about 3 quid an hour?
But even if you take this fantastic 0.3 bucks/hour for 32 cores x 2, x 24 hours, x 30 days, a month comes to fantastic 13.824,00 $/month, which is more than 2 times higher than the cost of the pedalats discussed in this thread. But to get the real (not fantastic) cost of renting 64 cores/month on Amazon, the result needs to be x 10.
Maybe for 3 bucks/hour...?
But even if we take fantastic 0.3 bucks/hour for 32 cores x 2, x 24 hours, x 30 days, it adds up to fantastic $13.824,00/month, which is more than 2 times more than the cost of the pedestal discussed in this thread. But to get the real (not fantastic) cost of renting 64 cores/month on Amazon, the result needs to be x 10.
Something with your arithmetic... $0,3*2*24*30 == $432
There's something wrong with your arithmetic... $0,3*2*24*30 == $432
Pardon me, I multiplied everything by 32 for some reason... I missed it.
But in practice, for $0.3 per hour no one will give such power, for $3 per hour they will give it, i.e. $432 x 10 = $4320 per month in total. Roughly. Rather more expensive, as the agents in operation are constantly eating up 100% of kernel resources and this will increase the fee.
Z.I.S.: info from Amazon price list (not including SSD cost):
One hour of an agent's time in the MQ cloud costs $0.02,
272 agents x 0.02 x 24 x 30 = $3916.80
The pepelats only cost5,500 euros according to the topicstarter. One-off.
Sorry, I multiplied everything by 32 for some reason.
But in practice, for $0.3 per hour no one will give such power, for $3 per hour they will, i.e. $432 x 10 = $4320 in total, per month. Roughly. Rather more expensive, as the agents in operation are constantly eating up 100% of kernel resources and this will increase the fee.
Z.I.S.: info from Amazon price list (not including SSD cost):
One hour of an agent's time in the MQ cloud costs $0.02,
272 agents x 0.02 x 24 x 30 = $3916.80
The pepelats only cost5,500 euros according to the topstarter. It was a one-time fee.
Here are the results of tests in theMQL5 Cloud Network:
Forum on Trading, Automated Trading Systems and Strategy Tests
Analysis of Testing and Optimization Results in the MetaTrader 5 Strategy Tester
Anatoli Kazharski, 2018.03.01 14:44
3. How long does it take to optimise the parameters in the cloud?
For comparison, let's try optimization with the same parameters inMQL5 Cloud Network. There is a charge for using this service. We will record after each optimization how much money will be frozen on the account to pay for this service.
In this test, we will set simultaneous use of local CPU cores and cloud agents to speed up optimization.
Symbol: EURUSD
Frozen funds:
As you can see, the optimization was significantly faster(6 min 41 sec) than last time, but only on the local computer(28 min 56 sec).
Symbol: EURCHF
You may encounter a situation where the service takes a very long time before the optimisation job is submitted to the network. This is due to the fact that agents need time to download the necessary data on the desired symbols. In this case, that is what happened.After a long wait you can stop the optimization process. There will be entries in the log as shown below.There were no calculations in the cloud, so no funds will be withdrawn.
Let's try to run the optimization again on this symbol.
This time the process was successful but the result was not impressive. The optimization took22 min. 14 sec. It took32 min. 50 sec. In fact, this is also due to the process of data uploading by agents on the network. The gain will be on larger tasks and inSlow complete algorithm mode.
Frozen funds:
Now let's see if it makes sense to run optimization in the cloud with multiple symbols.
Symbols: EURUSD,GBPUSD,USDJPY
It took24 min. 56 sec. But in our last run, it took2 hrs. 15 min. 3 sec. In this case the gain is already noticeable.
Frozen funds:
Symbols: EURCHF,AUDCAD,AUDNZD
This time, the optimization of the network did not start for a long time either. All this time local agents have been working. You can see what is happening in the log (see the listing below). As soon as all agents in the network that can perform optimization for you download the required data, optimization is started.
In the end, the optimization was finished in1 hr. 19 min. 49 sec. And the last time, but on the local machine only, it took3 hrs. 13 min. 37 sec.
Frozen funds:
The total for all four optimisation processes came out to$7.46. The table below shows a summary. Instead of6-7 hours, it took2 hours and 13 minutes to optimise, which is about three times less.
Here are the test results inMQL5 Cloud Network:
The only thing is that there are only 8 local agents in the test, while in the cloud, during optimization a batch of tasks is distributed to 256 or even 512 agents at once. It's obvious that calculations in the Cloud will be faster than in local agents.
For accurate speed comparison local agent set should be 272 or more, then the test will be more accurate. IMHO.
Anyway, if there's something to calculate, it's definitely more profitable to buy a pepelats.