Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
just a quick question - with your current setup, how long do you wait for some optimizations to end? Something like "150M combinations on 80K bars take approximately XY minutes". I would just like to compare the speed of my newly developed app to Multicharts that you use on your box.
Can't say from memory. I stopped using MultiCharts a few months ago. I've played some with NT7 recently, but not much due to lack of time. I've been spending some time beta testing IQBroker, which is built as a portfolio tester.
this is a very interesting thread.
Currently I am still on a i7-2600. not oc'ed because i needed stability....
Optimizing one of the strategies, I did give too many parameter, including time series in tick data, for it to optimize. It's been going on for hours. I am using PH Genetic Algorithm.
so i finally had this thought, how about a dual cpu rig or 3930K. this thread hit the spots just right.
Jack
To my knowledge, PH Genetic has serious flaws with NT7 and should not be used.
You should try the built-in NT genetic optimizer instead and see if there is a performance change.
As for dual cpu, I would instead recommend you spend $100 on a Corsair H80 or H100 and overclock your chip if you have the i7-2600k (unlocked version).
hey typer77, could you say how much data and how many combinations you were exploring, that took you hours? Im trying to compare the performance of my custom testing suite with Ninja. Thanks a lot.
I think one reason why it was taking so long is something wrong with my regular pc.
I have a few things loaded on it, so when I tried it on my laptop which only minimal trading softwares, it was much faster. Plus I optimized using 32bit NT7 (due to esignal feed limitation), it was slow.
On the laptop, I downloaded all historical data and ran 64 bit, much faster.
going back to your question, I had about 7-8 variables with 10 steps each.
With NT7, I think having more cores is the way to go esp using NT7's Genetic algo compared to higher clock. CPU utilization is quite low. What do you guys think? Or should it be the other way around? I am debating if i should build a system with 3930k (6 cores) or 3770k (4 cores higher clock and cheaper).
Hi Mike, I have the i7-2600... thought I wanted a stable system. last few OC'ed pc i had all went BSOD after 1 year or use. perhaps I should just build one for optimizing purposes.
However, when i optimize a large amount of variables, it always comes back to a performance 99% trade, i.e. only 1-2 trades in a years' worth of data. it's not realistic. what's the best criteria we should use to optimize?
Check out overclock.net and find threads for your motherboard and chip combination, they will guide you. You have to make certain you are using safe voltages. If you are, there is no reason your CPU would start BSOD or cause instability even after prolonged use. Especially chips these days, because with my board for example it is actually running undervoltage probably 90% of the time, only kicking in and increasing voltage when I need it.