Support Board
Date/Time: Sun, 24 Nov 2024 22:52:41 +0000
[User Discussion] - Technical question about limitations of replay backtest spreadsheet automation
View Count: 1461
| [2013-12-07 04:29:14] |
| enemyspy - Posts: 306 |
|
Hi, I am currently conducting multi-chart backtestingon a number of different automated trading formulas. Obviously as stated in the manual if a high resolution is required the replay test is conducted slowly. So my question: Is this limitation due to computer memory/processing power? Or is it a limitation of how fast the software can conduct the calculations? What I am getting at here, is if one were purchase a top end motherboard with the greatest amount of cpu cores, Maxed out the RAM slots, with an SSD, could that significantly improve the processing speed over a typical quad core 2 gig RAM, HD setup? Date Time Of Last Edit: 2013-12-07 04:34:14
|
| [2013-12-08 07:11:12] |
| vegasfoster - Posts: 444 |
|
A faster computer will perform the calculations faster. More memory is typically only useful to a point, so if you are on a budget I would get 8gb and put the rest toward the processor or an SSD. If you aren't on a budget, then get 16gb of ram, an Intel 4770k with water cooler so you can overclock, a motherboard that supports Raid 0, and two SSDs. :)
|
| [2013-12-08 09:52:54] |
| enemyspy - Posts: 306 |
|
Thanks Vegas I really appreciate your response and will take your advice on those specs.. Also, I was looking at the system performance in the task manager, and noticed that while I am running the replay at a 10 second resolution and 100x speed: the cpu usage seems to fluctuate between 30 and 50%, and approximately 25% of the memory is still available. It is not actually replaying at 100x speed, it seems to be going much slower. 2 months seems to take about 12 hours. So not being a technical wizard when it comes to system performance I have 1 more question: 1. Should it be maxing out the processor, or can the settings in the bios or something else be capping the amount of memory/processing power that it will allocate towards the task? Date Time Of Last Edit: 2013-12-08 09:54:21
|
| [2013-12-09 02:50:41] |
| vegasfoster - Posts: 444 |
|
Sierra Chart is only using a single core, so you should focus on single core performance over additional cores, e.g. spending big bucks on a Xeon setup may not help you. Also, if you are going from a Core2 to an I7, then the performance gain will be noticeable, but not as much you might hope for. This is also the reason I mention overclocking, because this should give you a real performance gain in this application, but results can vary greatly based upon the setup and if not done correctly it can result in damage to your system. It has already been requested that they add multi-core support.
|
| [2013-12-09 05:59:10] |
| enemyspy - Posts: 306 |
|
I see, That is good to know, I looked actually was looking at some info on cpus today and was wondering what would better more cores, or less cores and greater speed. So just to clarify, If I were to use a Zeon than only a single core of it would be used by sierra chart? and therefore just like how people only use 10% of their brain same deal with that processor, and therefore the only help is a bigger core? |
| [2013-12-09 08:25:29] |
| |
|
Is this limitation due to computer memory/processing power? We do not see how multiple cores can be used when performing a single back test. Could be used if there are multiple back tests going on. However, then there will be synchronization issues which can be very complex to deal with. It is very easy to talk about using multiple cores, but multithreaded programming gets to be very complicated and unless it is done right, it results in an unreliable program. It is also possible there simply is no benefit to multithreaded programming based upon the synchronization methods used. So you have to be very careful when you hear about multicore/multithreading support. It can mean unreliability unless it is done by the most best of programmers. And also if not done right it can be meaningless and of no benefit. If you are looking for Sierra Chart to support multiple cores with back testing and make it a whole lot faster, don't wait around for that. We do not see how a single back test can be distributed across multiple cores. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing Date Time Of Last Edit: 2013-12-09 08:28:44
|
| [2013-12-09 08:50:38] |
| enemyspy - Posts: 306 |
|
Thank you for that answer as welI am simply seeking to understand how it works, and why. So would running seperate versions of sierra chart that do not share files be possible to run for 4 seperate backtest, 1 seperate installation/core, all working independantly of one another? Also would it be possible to network historical data/statistics in a loop between two instances of sierra chart on seperate computers so that the tick by tick calculations are conducted on a very low amount of rows, and sent to the historical stats on the other computer so that they will be available to be sent back to the low time frame processing computer after being analyzed by the historical computer?Thereby giving it a larger memory of the tick calculations without driving 1 system too hard? Date Time Of Last Edit: 2013-12-09 08:53:30
|
| [2013-12-09 16:30:37] |
| |
|
So would running seperate versions of sierra chart that do not share files be possible to run for 4 seperate backtest, 1 seperate installation/core, all working independantly of one another? We do not understand your second question. It is outside the scope of our support. Sierra Chart Support - Engineering Level Your definitive source for support. Other responses are from users. Try to keep your questions brief and to the point. Be aware of support policy: https://www.sierrachart.com/index.php?l=PostingInformation.php#GeneralInformation For the most reliable, advanced, and zero cost futures order routing, *change* to the Teton service: Sierra Chart Teton Futures Order Routing |
| [2013-12-09 16:48:55] |
| enemyspy - Posts: 306 |
|
ok that is useful to know. So, there is no support for/ability to network 2 instances of sierra chart located on separate computers? and would I be correct to assume that it would run into the same syncing issues as if one were to attempt to design multicore support? |
To post a message in this thread, you need to log in with your Sierra Chart account: