eSignal CPU usage

[clearpicks]

Hello,

I just upgraded eSignal to 10.4.1608. I did a simple test of eSignal CPU usage. There was only one 2min chart (6days, 8:00-16:15). These was no any EFS running on the chart. If the symbol is $SPX, the CPU usage was almost 0% no matter the chart size. However when I changed the symbol to ES M9, the CPU usage would greatly depend on the chart size. For example, when the chart size was 1600x1200, the CPU usage was almost negligible. However with the increase of the chart size, the CPU usage would increase dramatically. For example, when the chart size is 6400x1200, it would be about 40%. Since these is not EFS running and the CPU usage is low when the chart is small, it seems not the problem of EFS engining or data manager. It was most likely caused by the charting engining. However because it is only the area arround the last bar need to be updated except for the time a new bar is created, such issue of CPU usage depending on the chart size seems abnormal to me. I can not recall whether I saw the same problem in older versions (I never used 10.3), but I guess not. Moreover the CPU usage when I open four ES M9 1600x1200 charts would still be much less than the case of one 6400x1200 chart. Can someone do a milti-screen test to reproduce the issue I just described? If this is really an eSignal issue, I might have to roll back to older version.


- Clearpicks

[eSignal_AveryH]

Clearpicks,

Thank you for the post, we'll investigate this and report back our findings.

Thank you
AveryH
eSignal Support

[sandpiper]

clearpicks,

I am really pleased that you have spotted this one as well.

I spotted it and reported it about 3 weeks ago but had to revert to an older eSignal version to carry on some development and just haven't got round to re-installing the offending eSignal version.

My exeperience was exactly the same as yours. Even to the extent that if you display symbols within the charts where the market is closed, i.e. there is no tick activity, the CPU usage rises considerably as the size of the charts increases.

Your observation that 4 small charts use less CPU than 1 correspondingly large chart also matches my findings.

One thing did occur to me though and that was why I was going to re-test it on my other machine. The machine I ran the test on was an AMD Athlon 64 4000+ single core and I wondered if the problems mingt be related to this (I should say that thhis has never caused me problems with eSignal in the past). On the contrary I reckon it is about one of the best single core setups for running eSignal on.

However, I was intending running the offending eSignal version on an Intel quad core just to make sure. What machine are you running on?

For your information, I rolled back to 7.9 and then forward again to 10.1.1291 which is the latest version I could find that includes tick based time templates but does not have the problem you describe.

[clearpicks]

I did the test on an Intel P4 3.0GHz Hyperthread desktop because only on this computer I am able to hook up 6 monitors while all my newer computers are laptops. The issue exists in both eSignal 10.2.1391 and 10.4. Maybe I should roll back to 10.1 as you did untill this issue gets solved.

[clearpicks]

Where can I download old eSignal version such as 10.1.1291?

[sandpiper]

There used to be a knowledgebase article with links to older versions but my IE link to it no longer seems to work I'm afraid.

Hopefully Avery can let us know the correct link.

[eSignal_AveryH]

Hello Clearpicks,

We're continuing to investigate the report of CPU usage when expanding charts across multiple monitors. In the meantime you can download a previous version of eSignal from this article on the eSignal Knowledgebase.

AveryH
eSignal Support

[clearpicks]

AveryH,

It seems the article link you posted could not be opened.


Clearpicks

[ScottJ]

Link opens fine here. The article # is 3579 so you can just type in that # in the KB and it should come right-up.

Thanks.

[clearpicks]

I just downgraded to 10.1.1291 and the CPU usage is noticeably lower than that in 10.2 and 10.4, and for most of time it varies between single digit to 20 percent with three charts running, on each of them there are two EFS running. One chart is 4800x1200 while the other two are both 800x1200. If without any scripts running, the CPU usage would be less than 4 percent even if with a chart of size 6400x1200. In 10.2 or 10.4, no matter whether there are EFS running or not, any single chart larger than 3200x1200 would make the CPU usage keep hanging above 45 percent ( since it is a heperthread single core CPU, it would means almost 100% CPU utilization ).

Clearpicks

[clearpicks]

Hello AveryH,

How is the test going? Have you found out the reason causing the problem?


Regards,
Clearpicks

[ScottJ]

We only move forward with our desktop code so we'll see how 10.5 works for you when released. The information provided here was sent to our developer team.

Thanks.

[clearpicks]

I just tested the issue in eSignal 10.5 beta. Unfortunately the issue still exists. In eSignal 10.1.1291, after I modified all EFS' so that they all compute on close, the CPU utilization has been below 5% for almost all the time. However after installing 10.5 beta, with the same layout, the CPU utilization is always hanging above 30-40 percent (sometimes 50 percent, in my case of single core hyperthreading P4, it is equivelant to 100% CPU utilization). Now it is still premarket time and the ES trading volume is still light. I bet after 9:30 EDT, the CPU utilization would be much higher. In the test, there is one large chart ( 6400 x 1200 ). If the chart size is decreased, the CPU utilization would decrease accordingly, while in eSignal 10.1, there is no such issue and the CPU utilization is much much lower. So there must be a bug in charting engine which exists in all versions after 10.1 which is causing this issue.


- Clearpicks

[sandpiper]

Avery,

Any news on this issue?

Regards.

[jgr]

Clearpicks,

Have you tried your test case with the 10.5 RC??