OpenCl and the tools for it. Reviews and impressions. - 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
Try MQL5 Cloud Network - the speed of optimization is as high as never before seen in MT4. On massive calculations, it is amazing how hundreds of results per second collapse when using a couple of thousand independent agents.
Tried it. The speed of optimization compared to MT4 with the genetic algorithm enabled is only depressing. MT5 won't dream soon about the optimization speed of MT4 of the current builds.
Have you tried MQL5 Cloud Network with 2000 agents on a dual-core?
Or have you simply launched a single local pass and rested there?
It is with the MQL5 Cloud Network that we have achieved a simple acceleration of hundreds and thousands of times. Without GPU, on any Expert Advisor. Take my old example with video: Example of calculation acceleration using MQL5 Cloud Network
We have made it possible to instantly raise thousands of agents with one button (the warm-up time is about 20-30 seconds) for fast calculation of tasks. Instead of hundreds of hours, you can really get it done in half an hour.
Moreover, the prices are ridiculous - anyone can check them, and everyone who signs up at MQL5.com gets a bonus of $2.
This is a real revolution. Now any user can put almost any supercomputer to shrugs through the use of the cludes. And with the introduction of the GPU (a build will be available on Friday) it will be possible to beat supercomputers in packs.
The performance is enough (or rather you convince yourself of that) for simple cases. The performance is not nearly enough for the most massive and detailed calculations.
The mistake with comparing MT4 vs MT5 testers is that people turn a blind eye to the completely different level of detail and the ability to accurately multicurrency test.
You're wrong about dependency on hardware:
You are blatantly wishful thinking, clearly knowing you are wrong.
Renate you are blatantly wishful thinking. About MT5 supporting architecture of different bit sizes and using all local cores, I don't argue. Another question is how does it use it? I.e. it is possible to use all of them, and at the same time it is absolutely useless.
The OpenCL hardware-independent joke should be placed in the Humour section. You can tell this joke to traders who have video cards without Cuda built into their motherboards. This card is more than enough for trading, and for watching quality video and pictures. But for computer games this hardware-independence is not enough. Trading shouldn't be confused with computer games.
Renat, have you personally experienced this distributed computing network? I have the impression that you are buying the advertising your company is spreading.
My experiments with Clouds Network with genetic algorithm turned on gave me the following results:
1. There are not 2000 agents in Clouds Network. There are at most 1500.
2. When you connect Clouds Network, at best of all agents, only 200 are connected, at worst none.
3. Distributed calculations work jerkily, i.e. first I get autorized mode, from which many networks never leave. Then one network connects. It receives a batch of tasks, engages a portion of agents, quickly processes these tasks and shuts down, switching permanently to finished mode.
Compared to those farms that handle 3D graphics, including free ones, Clouds Network is pathetic. It's not about any kind of revolution. Everything is extremely ugly made.
The prices are really ridiculous and everything else looks very sad and even depressing.
Let's break it all down to get to the root of the problem:
The optimization mode on MT5 is very slow when the genetic algorithm is on. I made an Expert Advisor on MT4 and tested and optimized it. Optimization time is less than 5 minutes on dual core (of course MT4 has only one core involved, but other tasks don't get in the way because they can run on the second core). I have rewritten the same Expert Advisor for MT5. I have tested it for optimization. The optimization time is more than one hour, or nearly 2 hours to be exact. What is the difference?
Of course, we cannot optimize EAs at such a speed.
I will connect the Clouds Network. Indeed, the speed has increased but even with distributed calculations it is still much slower than on MT4. There are several reasons for that:
1. All tests for MT5 were done on M1. I had H1 involved. Most likely the miscalculations on velocity lies here?
2. Distributed computing, even on a local computer, slows down the processes, while MT5 performs all the optimization processes only through distributed computing.
3. Clouds Network significantly accelerates the optimization process, but if it is not used, you will not be able to optimize a complex Expert Advisor on a local computer in a reasonable time.
4. Clouds Network only allocates one agent from each network when searching for forward tests, resulting in an extremely slow process.
5. The most resource-intensive part of autotrading is the optimization. This is the result.
And this is the third time it's been repeated on "hardware dependent software".
You can of course argue the contrary. But the point remains that MT5, unlike MT4, is iron-dependent for autotrading. We do not have to guess why:
1. Optimization speed on MT4 without additional crutches is very high even on relatively weak and obsolete computers.
2. MT5 does not allow optimization without additional hardware, i.e. additional cores, high-speed RAM, GPU or without using someone else's hardware, through distributed calculation in local networks or through Clouds Network
Hence we conclude that the platforms are hardware-independent or hardware-independent.
We can perform a simple experiment:
1. Let's take an average computer without video cards with Cuda mode. Of course, you can experiment on a supercharged and expensive computer, but we are talking about being hardware-independent. And an average trader is unlikely willing to install a server rack to optimize a Pipser on a cent account? It is desirable that the computer is remote in the Internet from the Clouds Network servers for the N-th number of hops, rather than sitting on one hub in the local network. Because ordinary users of terminals to this very hub may not connect to it, and quality of connection in distributed computing is very important.
2. Let's take two Expert Advisors for MT4 and MT5 with the same logic, but it is desirable to have many input parameters, i.e. to use the genetic algorithm.
2. Let's set the dates for the last year. Note: on MT5 if you select "Interval" for the last year and "Choose period" by dates for the same last year, the results are completely different. Not sure why, but that's probably a question for the developers.
3. Let's set the timeframe to H1
4. Normal trading mode, at opening prices.
5. Set takeprofits and stoplosses so that the number of trades is 250 - 400.
6. We switch on forward test 1/2. We do not have it on MT4, but it will help someone to see sluggishness of this mode on MT5.
7. Let's start both advisors on MT4 and MT5, and check time. And on MT5 we launch the process twice: with local agents (if one has strong nerves and no time to wait for the completion of optimization) and with Clouds Network.
8. Conclusions based on the results can be drawn as to which platform is more suitable for autotrading and suitable for an average trader and which is not.
Let's break things down to get to the root of the problem:
The optimization mode on MT5 is very slow. I made an EA on MT4 and tested and optimised it. Optimization time is less than 5 minutes on dual core (of course MT4 has only one core involved, but other tasks do not interfere, because they can run on the second core). I have rewritten the same Expert Advisor for MT5. I have tested it for optimization. The optimization time is more than one hour, or nearly 2 hours to be exact. Is there a difference?
........We can make a simple experiment:
1. Let's take an average computer without video cards with Cuda mode. Of course, you can experiment on an expensive supercharged hardware computer, but we are talking about hardware-independence. And an average trader is unlikely willing to install a server rack to optimize a Pipser on a cent account? It is desirable that the computer is remote in the Internet from the Clouds Network servers for the N-th number of hops, rather than sitting on one hub in the local network. Because ordinary users of terminals to this very hub may not connect to it, and quality of connection in distributed computing is very important.
2. Let's take two Expert Advisors for MT4 and MT5 with the same logic, but it is desirable to have many input parameters, i.e. to use the genetic algorithm.
2. Let's set the dates for the last year. Note: on MT5 if you select "Interval" for the last year and "Choose period" by dates for the same last year, the results are completely different. Not sure why, but that's probably a question for the developers.
3. Let's set timeframe to H1
4. Normal trading mode, by opening prices.
5. We set takeprofits and stoplosses so that the number of deals was 250 - 400.
6. We switch on forward test 1/2. We do not have it on MT4, but someone will see the brakes of this mode in MT5.
7. Let's start both advisors on MT4 and MT5, and check time. And on MT5 we launch the process twice: with local agents (if one has strong nerves and no time to wait for the completion of optimization) and with Clouds Network.
8. Conclusions based on the results can be drawn as to which platform is more suitable for autotrading and suitable for an average trader and which is not.
Yura. You're right. That's the way it is for now.
But we will have to study mql5 anyway. :) Because the situation will soon change. The developers have already understood it. And if they are imbued with it, they will do it. They have also developed the MT4 tester, right?
Yura. You are right. So far it is like that.
But you still have to learn mql5... :) Because the situation is about to change. The developers have already got used to it. And if they have, they will do it. The MT4 tester was made by them, right?
It's a pity they didn't get into it earlier, spent a lot of time and nerves on different crutches for the tester.
MT5 is far superior to MT4 in terms of functionality, but in terms of optimization speed it's a wild terror.
If they got into it earlier and at least caught up with MT4 in terms of speed, almost everyone who is engaged in auto-trading for a long time would have switched to MT5 by now. And if traders move to the new platform, then brokers will have to move there too - no need to spend money on advertising and trying to persuade someone to move to MT5.
TheXpert:
There's a column for "specific performance per quid spent on card purchase" :)
It's a shame they didn't get into it sooner and wasted a lot of time and nerves on all sorts of crutches for the tester.
MT5 is far superior to MT4 in terms of functionality, but in terms of optimization speed it is a wild terror.
If they got into it earlier and at least caught up with MT4 in terms of speed, almost everyone involved in autotrading for a long time would have switched to MT5 by now. And if traders move to the new platform, brokers will have to move there too - no need to spend money on advertising and trying to persuade someone to move to MT5.
I don't know about that. God willing. And I'm still impressed with the usefulness of the graphics card in my computer.
TheXpert: Why not a 6950?
Is it a question for me? If it's for me, the answer is simple. It's not too expensive and acceleration is unlikely to be significant compared to 6930. Well, maybe 20%.
P.S. Oh, I see, question for MD.
Why not a 6950 then?
This was the second candidate. Although the 6930 has a slightly better price/performance, the 6950 is ahead in absolute terms of performance.
In the end I didn't manage to stay within the money (8100r vs 5650r).
// But I can change my memory from 1600 MHz to 2000 MHz, which is a (very) good compensation because slow memory-to-RAM-video exchange could have eaten up the whole advantage.