Hello,
I have a Multicore Windows Machine running a walkforward opitmization. As you can see in the screenshot it does not use all cores and does not stress the machine. I ran a CPU-Benchmark with Cinema Cinebench (http://www.maxon.net/de/downloads/cinebench.html) and here all CPUs are used an the machine is stressed by 100% CPU usage.
Is there anything I need to configure?
Regards,
Stephan
MultiCore Machine and MC-Support
MultiCore Machine and MC-Support
- Attachments
-
- CineBench.JPG
- (331.98 KiB) Downloaded 1217 times
-
- Cinebench_2.JPG
- Cinebench
- (191.79 KiB) Downloaded 1193 times
-
- MCORE_OPTIMIZATION_WALKFORWARD.JPG
- (346.51 KiB) Downloaded 1218 times
Re: MultiCore Machine and MC-Support
1. Does your signal/strategy uses keywords:I have a Multicore Windows Machine running a walkforward opitmization. As you can see in the screenshot it does not use all cores and does not stress the machine. Is there anything I need to configure?
Code: Select all
definedllfunc/externalCode: Select all
printCode: Select all
FileAppend2. What version of MC do you use?
3. Does the Exhaustive Optimization utilises all cores?
Re: MultiCore Machine and MC-Support
I.e. exhaustive optimization utilise 30-50% of CPU only?3) No, both exhaustive or genetic takes 100% of CPU Usage only 30-50 %.
How long time does strategy calculates (backtest)? If this time is very short then MC's optimizer work is equal to strategy calculation and efficiency is low.
Re: MultiCore Machine and MC-Support
When I asked "How long time does strategy calculates (backtest)?" I had in mind a time from applying signals/strategy on chart and to end of strategy calculation.The optimization takes serveral hours.
Exhaustive optimization (not WFO) utilise 30-50% of CPU only or it utilises 90-100% of CPU?
-
evdl
- Posts: 401
- Joined: 19 Jan 2011
- Location: Netherlands
- Has thanked: 85 times
- Been thanked: 124 times
Re: MultiCore Machine and MC-Support
I use an intel i7 2600k processor. And for backtesting a couple of years of minute data it will take about 1 minute to test.
The optimizing of my strategy with the same data (and up to 10 parameters) can take up to 20 hours of optimizing. In this process my processor is maxed to 100%.
bstrader has I think a dual processor setup of two Xeon processors instead of one processor. And if I understood it correctly, the question is if Multicharts can fully take advantage of the processing power of dual processors. If so all the 2x8 cores and in total 32 threads should be maxed out to 100%, I guess, and not 30 to 50%.
The optimizing of my strategy with the same data (and up to 10 parameters) can take up to 20 hours of optimizing. In this process my processor is maxed to 100%.
bstrader has I think a dual processor setup of two Xeon processors instead of one processor. And if I understood it correctly, the question is if Multicharts can fully take advantage of the processing power of dual processors. If so all the 2x8 cores and in total 32 threads should be maxed out to 100%, I guess, and not 30 to 50%.
Re: MultiCore Machine and MC-Support
Yes this is correct. It is a EC2 Server from Amazon EC2 it has 2 CPUs (Cluster Compute Instances).I use an intel i7 2600k processor. And for backtesting a couple of years of minute data it will take about 1 minute to test.
The optimizing of my strategy with the same data (and up to 10 parameters) can take up to 20 hours of optimizing. In this process my processor is maxed to 100%.
bstrader has I think a dual processor setup of two Xeon processors instead of one processor. And if I understood it correctly, the question is if Multicharts can fully take advantage of the processing power of dual processors. If so all the 2x8 cores and in total 32 threads should be maxed out to 100%, I guess, and not 30 to 50%.
And yes, the exhaustive optimization (not WFO) utilise 30-50% of CPU.
-
evdl
- Posts: 401
- Joined: 19 Jan 2011
- Location: Netherlands
- Has thanked: 85 times
- Been thanked: 124 times
Re: MultiCore Machine and MC-Support
Ok. So it is not a standalone pc with two processors. But a webbased distributed computer solution from Amazon.
I think this is the problem. In different topics on the forum is talked about sharing processor capacity. This one for example: http://www.multicharts.com/discussion/v ... =1&t=10366.
I am not sure if the solution from Amazon can be used to its full potential with MC. Because of the way data is processed. I would like to hear this also.
I think this is the problem. In different topics on the forum is talked about sharing processor capacity. This one for example: http://www.multicharts.com/discussion/v ... =1&t=10366.
I am not sure if the solution from Amazon can be used to its full potential with MC. Because of the way data is processed. I would like to hear this also.
Re: MultiCore Machine and MC-Support
Hi bstrader,
I have the same problem, but my setup is different than yours: I use a MacBook Pro 2010 (Core i5) and a Mac mini 2010 (Core 2 Duo) to run Windows XP Pro natively with Boot Camp. One of my signals somehow causes MC to use only about 50% CPU when I do an optimization on the MacBook Pro, but not on the Mac mini.
So, the same signal causes MC to use only 50% CPU on the MacBook Pro, while it uses 100% CPU on the Mac mini. I am not sure what happened. I already notified MC support about this issue, but they told me that my Mac hardware is the cause (running Windows XP with Boot Camp). I am not sure that's the case since the problem signal uses 100% CPU on the Mac mini that runs Windows XP with Boot Camp.
I think this is a bug, I hope MC support can provide more info.
I have the same problem, but my setup is different than yours: I use a MacBook Pro 2010 (Core i5) and a Mac mini 2010 (Core 2 Duo) to run Windows XP Pro natively with Boot Camp. One of my signals somehow causes MC to use only about 50% CPU when I do an optimization on the MacBook Pro, but not on the Mac mini.
So, the same signal causes MC to use only 50% CPU on the MacBook Pro, while it uses 100% CPU on the Mac mini. I am not sure what happened. I already notified MC support about this issue, but they told me that my Mac hardware is the cause (running Windows XP with Boot Camp). I am not sure that's the case since the problem signal uses 100% CPU on the Mac mini that runs Windows XP with Boot Camp.
I think this is a bug, I hope MC support can provide more info.
Re: MultiCore Machine and MC-Support
Hi,
if somebody of MC Support wants to have access to this machine to make tests, that´s no problem. I think all cores and procs are used but not very effectivly (just look at the screenshot).
Regards,
Stephan
if somebody of MC Support wants to have access to this machine to make tests, that´s no problem. I think all cores and procs are used but not very effectivly (just look at the screenshot).
Regards,
Stephan
Re: MultiCore Machine and MC-Support
There is no way to edit the bios settings, i can not change the multithreading params.
it is a ec2 instance from amazon:
Cluster Compute Eight Extra Large Instance
60,5 GB Arbeitsspeicher
88 EC2 Compute Units (2 x Intel Xeon E5-2670, Eight-Core-Architektur "Sandy Bridge")
3370 GB Instance-Speicher
64-Bit-Plattform
E/A-Leistung: sehr hoch (10-Gigabit-Ethernet)
API-Name: cc2.8xlarge
it is a ec2 instance from amazon:
Cluster Compute Eight Extra Large Instance
60,5 GB Arbeitsspeicher
88 EC2 Compute Units (2 x Intel Xeon E5-2670, Eight-Core-Architektur "Sandy Bridge")
3370 GB Instance-Speicher
64-Bit-Plattform
E/A-Leistung: sehr hoch (10-Gigabit-Ethernet)
API-Name: cc2.8xlarge
Re: MultiCore Machine and MC-Support
Could you do a simple test - run Exhaustive optimization with Channel BreakOut LS/SE. If CPU load will still be 50% then you could contact with MC's Help Desk for future investigation. If CPU load will 100% then the root of the problem is somewhere in your strategy.
-
JoshM
- Posts: 2195
- Joined: 20 May 2011
- Location: The Netherlands
- Has thanked: 1544 times
- Been thanked: 1565 times
- Contact:
Re: MultiCore Machine and MC-Support
I'd also like to know that. What are your experiences with Amazon EC2 Bstrader? Did you succeeded in using the full CPU power?I am not sure if the solution from Amazon can be used to its full potential with MC. Because of the way data is processed. I would like to hear this also.
Is it worthwhile to spend time and resources on learning and configuring this, or are the benefits of this scalable service relatively limited for MultiCharts?