Pl disclose benchmark improvements with new versions

Questions about MultiCharts and user contributed studies.
hilbert
Posts: 224
Joined: 17 Aug 2011
Has thanked: 76 times
Been thanked: 64 times

Pl disclose benchmark improvements with new versions

Postby hilbert » 17 Oct 2014

Dear MC Team,

I strongly feel MC should indicate by how much performance has improved because of a change they have done in a new release. For example, Whats new under Features -> Stability for MC 9.0 build 10014 says "Optimized CPU usage by MultiCharts.exe process and optimized memory usage by TradingServer.exe process".

Rather than just saying this, it is much more helpful if users can get some sense of improvements like how these optimizations affect the speed/time taken for various activities? Opening/closing workspaces, loading data in charts, plotting indicators, backtesting speed, optimization etc.?

MC team must have these benchmarks internally. Revealing the percentage speed changes, reduction in time etc. make marketing sense as well as allows users to get a sense of how much improvements are we talking about. E.g. lets say as a user I have refrained from doing much tick data analysis because of speed limitations. However, if backtesting speed or speed of loading data in charts has increased by 2x or 3x, I might want to increase the tick data analysis that I do for my strategies. Just saying speed has improved is not very helpful.

Thanks for listening! :)

User avatar
arnie
Posts: 1594
Joined: 11 Feb 2009
Location: Portugal
Has thanked: 481 times
Been thanked: 514 times

Re: Pl disclose benchmark improvements with new versions

Postby arnie » 19 Oct 2014

I like the benchmark idea.

For the 9th version of MC, my personal experience tells me that in terms of memory usage I don't see much difference.
In terms of time loading a chart or a workspace things seems to be also OK compared with previous versions.

What I've been seeing though is an increase of CPU usage.
My CPU is all over the place. 8%, jumps to 30%, jumps do 60%, comes back to 20%, 8%, jumps to 50%, like I said, all over the place.
Lately I've been only running MC on the desktop to see how memory and CPU behaves. If Chrome is opened I can notice a clear rise in CPU usage but with MC alone it's been a roller coaster ride :(

I've a i5 CPU and I recently searched for an i7 since my motherboard is compatible to it but unfortunately the socket 1156 it's no longer available, not in my country :(

So my question to MC support is if there's in fact an increase of CPU usage in this 9th version?

Tomorrow I'll do a test.
I stripped my workspace from every study I use, leaving only the charts.
Lately, when the markets open my CPU spikes to 100%. I want to see how it behaves without a single study. I want to see how the studies impact MC CPU usage. If even without a single study MC spikes to 100% this means that this 9th version became really heavy on the CPU side.

User avatar
Henry MultiСharts
Posts: 9165
Joined: 25 Aug 2011
Has thanked: 1264 times
Been thanked: 2957 times

Re: Pl disclose benchmark improvements with new versions

Postby Henry MultiСharts » 27 Oct 2014

Dear MC Team,

I strongly feel MC should indicate by how much performance has improved because of a change they have done in a new release. For example, Whats new under Features -> Stability for MC 9.0 build 10014 says "Optimized CPU usage by MultiCharts.exe process and optimized memory usage by TradingServer.exe process".

Rather than just saying this, it is much more helpful if users can get some sense of improvements like how these optimizations affect the speed/time taken for various activities? Opening/closing workspaces, loading data in charts, plotting indicators, backtesting speed, optimization etc.?

MC team must have these benchmarks internally. Revealing the percentage speed changes, reduction in time etc. make marketing sense as well as allows users to get a sense of how much improvements are we talking about. E.g. lets say as a user I have refrained from doing much tick data analysis because of speed limitations. However, if backtesting speed or speed of loading data in charts has increased by 2x or 3x, I might want to increase the tick data analysis that I do for my strategies. Just saying speed has improved is not very helpful.

Thanks for listening! :)
Hello hilbert,

Here are the results of this bug report. Please keep mind that these results are subjective for a certain hardware/MultiCharts configuration. The results can also vary depending on the market activity.

Within 13 hours MultiCharts64.exe in version 8.8 Release (Build 8593) accumulated 1600 Mb. Max CPU usage 100%.
TradingServer.exe accumulated 2400 Mb.

Within 3 hours MultiCharts64.exe in version 9.0 Release accumulated 100 Mb. Max CPU usage 7%. Average CPU usage 5-7%.
TradingServer.exe accumulated 3 Mb.

Within 13 hours MultiCharts64.exe in version 9.0 Release accumulated 1200 Mb. Max CPU usage 94%. Average CPU usage 5-7%.
TradingServer.exe accumulated 170 Mb.

User avatar
Henry MultiСharts
Posts: 9165
Joined: 25 Aug 2011
Has thanked: 1264 times
Been thanked: 2957 times

Re: Pl disclose benchmark improvements with new versions

Postby Henry MultiСharts » 27 Oct 2014

I like the benchmark idea.

For the 9th version of MC, my personal experience tells me that in terms of memory usage I don't see much difference.
In terms of time loading a chart or a workspace things seems to be also OK compared with previous versions.

What I've been seeing though is an increase of CPU usage.
My CPU is all over the place. 8%, jumps to 30%, jumps do 60%, comes back to 20%, 8%, jumps to 50%, like I said, all over the place.
Lately I've been only running MC on the desktop to see how memory and CPU behaves. If Chrome is opened I can notice a clear rise in CPU usage but with MC alone it's been a roller coaster ride :(

I've a i5 CPU and I recently searched for an i7 since my motherboard is compatible to it but unfortunately the socket 1156 it's no longer available, not in my country :(

So my question to MC support is if there's in fact an increase of CPU usage in this 9th version?

Tomorrow I'll do a test.
I stripped my workspace from every study I use, leaving only the charts.
Lately, when the markets open my CPU spikes to 100%. I want to see how it behaves without a single study. I want to see how the studies impact MC CPU usage. If even without a single study MC spikes to 100% this means that this 9th version became really heavy on the CPU side.
arnie, did you try running your workspaces without any studies?

Please also try disabling the UAC in Windows, all antivirus and background applications.
In MultiCharts go to Help->Support->Open logs folder-> 9.0.10014.400->Logs. Check the folders. Make sure there are no files with "Error" in the name and there are no excessive backup files with .zip extension (excessive means a new zip file is created each 30-60 seconds).

User avatar
arnie
Posts: 1594
Joined: 11 Feb 2009
Location: Portugal
Has thanked: 481 times
Been thanked: 514 times

Re: Pl disclose benchmark improvements with new versions

Postby arnie » 27 Oct 2014

arnie, did you try running your workspaces without any studies?

Please also try disabling the UAC in Windows, all antivirus and background applications.
In MultiCharts go to Help->Support->Open logs folder-> 9.0.10014.400->Logs. Check the folders. Make sure there are no files with "Error" in the name and there are no excessive backup files with .zip extension (excessive means a new zip file is created each 30-60 seconds).

Yes, I did test my workspace without a single study.
The performance improved a bit, on average I could see my CPU running between 10% and 40%.

Image

I then decided to remove the two volume delta charts and on average the CPU ran between 6% and 20%.

Image

So no doubt that volume delta charts can push the CPU by at least 20% depending naturally on how fast the market is trading. Today hasn't been a fast market.

Can you guys confirm that volume delta charts can indeed weight on CPU resources?

Regarding background applications, I've been working on it but it's hard. One way would be to make this PC for trading alone, having this way the only thing installed the broker and MC.
Another way would be to install in this computer a second OS (Ubuntu) so Windows would be for trading alone while Ubuntu would be for everything else.
This is something that I'm going to test soon since I have Ubuntu on my laptop and that thing is simply amazing and fast.

Something that I have some issues with is having Windows without an anti-virus and a firewall.

Buying a second machine is out of the question at the moment simply because I have no room for it :(
Attachments
tm02.png
(54.19 KiB) Downloaded 558 times
tm_01.png
(49.89 KiB) Downloaded 563 times

User avatar
Henry MultiСharts
Posts: 9165
Joined: 25 Aug 2011
Has thanked: 1264 times
Been thanked: 2957 times

Re: Pl disclose benchmark improvements with new versions

Postby Henry MultiСharts » 29 Oct 2014

arnie, Volume Delta chart can consume CPU resources for chart display. You can notice a decrease in resources usage if you minimize the chart.
The more price levels are plotted on the chart - the higher will be the CPU usage. Please make sure the MinMovement and PriceScale values for this instrument are correct in the QuoteManager in order to avoid excessive price level splits (excessive CPU usage).
Regarding background applications, I've been working on it but it's hard. One way would be to make this PC for trading alone, having this way the only thing installed the broker and MC.
Another way would be to install in this computer a second OS (Ubuntu) so Windows would be for trading alone while Ubuntu would be for everything else.
This is something that I'm going to test soon since I have Ubuntu on my laptop and that thing is simply amazing and fast.

Something that I have some issues with is having Windows without an anti-virus and a firewall.
That is required for testing purposes only. That is possible that one of the applications is using the CPU time, that is why it is recommended turning off the applications one by one for eliminating such possibility.


Return to “MultiCharts”