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)
Haven't had time to do any benchmarks. But MC has no problem filling all 48 cores with full load when you are backtesting more than around 50,000 bars. If you test fewer bars it won't fill the pipe. Not a problem for me, most of my tests are over a million bars.
I change affinity so that one or two cores are left untouched, so the rest of the machine continues to respond well.
A guestimate for speed: My i7 920 was overclocked to 4.0ghz, so clock for clock the i7 was faster. But there were 8 cores, and now there are 48 cores. So no matter how you look at it, the new machine is considerably faster.
If I was building a new single processor workstation, it would not be AMD. It would be Intel.
In my opinion is better built a lan by windowserver2008 hpc and share the calculation from all machine.
Do any one try to built grid computing by multicharts?
I would love for that to be true. But unless I am completely wrong, that is not at all how a HPC works.
When you run an application like MultiCharts, it does not see a combination of all the nodes. Instead, the head node simply assigns the task to one of the cluster nodes. The application does not see an aggregate of all cluster nodes, in other words the application does not span multiple cluster nodes.
I think Mike is right, unless having a special software version, HPC ready, the threads will not be dispatched on all nodes. I did some large backtests, maybe 10 years ago, on a Linux HPC farm, but it was Perl home made code, and the threads were not dispatched on all slave nodes, that was real processes.
Hi All,
I use Worden's Stockfinder, which has a VB based programming interface called "Realcode". I have programmed a set of indicators. Till recent i had it running on a Intel I7 920, overclocked at 4Ghz. Stockfinder/realcode uses a lot of CPU resources, specially since I use at least 500 bars per chart, and have some indicators that loop over each bar and do some extensive calculations.
With the I7 920 I had my layout maxed out at 3charts. I desperately wanted to load more charts, therefore, since Stockfinder is a multithreading app, I thought I would throw as many CPU cores at it as possible.
My first attempt was a Supermicro H8QGi-F board with 2 6282SE Opteron CPU's (max is 4 CPU), for a total of 32 threads. Used MS server 2008, to my surprise it performed worse then my I7 920. returned the board on a RMA, and replaced with a ASUS Extreme IV formula, with a I7-3930x and overclocked it at 4.9Ghz. I got it to 5Ghz, but it would crash once in a while, and at 4.9Ghz is rock-solid. Now I've got 11 charts running with minute data.
Moral of story : whatever app you're using, and thinking of upgrading hardware in order to improve performance, then just the total amount of cores might not be the single answer.
A simple test is to have your current CPU run on a single core, and step from a low clockspeed to the max clockspeed supported. In my case I found out that a single chart with all my indicators on it runs very poorly if my CPU clockspeed gets reduced below around 3.5Ghz, regardless of total cores. after clocking over 4Ghz, I can expand my layout with more charts, almost on a extra chart per extra core bases.
Since the AMD Opterons run at about 3Ghz clockspeed, it really didnt matter how many cores I had available, performance was poor. But when I switched to a CPU that can run at higher clockspeed, things run fine, and my layout started taking advantage of the extra cores.
next mission is to try a EVGA SR-2 with dual 5690's OC'd at about 4.5 to 4.6Ghz (from reading that seems to be the ceiling for these CPU's), and thus be able to run even more charts on my layout.
It was a dog. I sold it to a Folder. I built a new workstation instead, i7-2700k overclocked to 5.0ghz on water, 32gb, 2x256gb crucial m4 ssd raid 0. More threads and cores are great, but the AMD box was a real dog as a workstation, the i7 box just crushes it.