Support Board
Support Board
Search Board
Control Panel
View My Posts / Threads
Direct MessagesDate/Time: Sat, 04 May 2024 16:27:13 +0000
Win 10 performance noticably slower than win 7
View Count: 2643
| [2019-01-07 08:31:51] |
| User735389 - Posts: 188 |
|
Are there any known optimizations or limitations between windows 10 and 7? I've already looked through the Help #30 section. I have a lot of charts and chartbooks (25 chartbooks, 8 charts + 1 detached window per chartbook). This runs without a hitch in win 7. but in Win 10, its unable to keep up with the FXCM 1/2 second tick datastream, pushing 1-1.5 sec per tick. A single chartbook will be able to keep up with the 1/2 sec tick. I dont see an increase in CPU usage or anything, i can't figure out where the bottleneck is. Any help is appreciated. I'm looking for a new laptop and it seems they've stopped supporting win 7 a few years ago.
|
| [2019-01-07 09:12:48] |
| |
|
One thing we can say for sure is we do not think this is specific to Windows 10. There must be something else going on, on that particular system.
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 |
| [2019-01-12 12:31:56] |
| User735389 - Posts: 188 |
|
I just performed a test. On the same machine I have both win 7 and 10. 4 core i5, 4.4 ghz. Both x64, both being fresh installs. I removed all the indicators off all the charts. Again this is 25 chartbooks x (8 charts + 2 detached windows) for each chartbook. Then just added a simple indicator on one chart, so i can debug it in Visual Studios. In win 7, it reports 0-1% cpu usage on the sierrachart process. Perfect. Visual Studio shows it takes about 10-20ms to make a round trip through all the charts. In win 10, it reports about 3-4% cpu usage. In Visual Studio shows it takes about 50-70ms to make a round trip. so that's about 2.5-3x slower performance, which was about what I was seeing with another win10 machine a few months back (based on cpu usage). The problem compounds for me when i add in all my indicators. Win 7 usually takes under 250ms for a round trip (~10% cpu usage). Win 10 pushes towards 600ms+ a round trip (near/maxing out at 25% on a single core), at which point causes more noticeable lag and longer than the fxcm 1/2 second data tick rate. I have a new win 10 laptop coming in that i'll confirm with, but unless you guys have any suggestions or can confirm with tests on your end, it does appear that sierrachart runs slower on win 10. Running multiple instances to spread the load isn't an option. I know my particular use case is probably unusual, and most users won't have an issue since SC still runs pretty fast. But something to be aware of. Anyone pushing the limits of SC on win 10 may want to consider testing on a Win 7 machine. |
| [2019-01-12 13:02:21] |
| ganz - Posts: 1048 |
|
User735389 so that's about 2.5-3x slower performance
+1
|
| [2019-01-13 07:15:41] |
| PeterSt - Posts: 36 |
|
(near/maxing out at 25% on a single core)
Which is to be read as : the thread of concern (utilizing one core at most) is at 100%. Thus SC, FYI, this is indeed maxing out. W10 can utilize tons of more optimizations than W7, although I can't readily think of what this would be in this instance. The main differences will be in the cooling features of the processor and the task scheduler. The former you can't see really, but the latter possibly can be debugged by means of forcing the system into situations that avoid task switching (read : the tast switching itself can start to cause overhead which is killing). Personally I have never done such a thing for this reason, but what I would do just for observations is telling the BIOS that only one core is in use (can be done through MSConfig as well). It is almost useless to talk about these things because it is (has to be) a matter of interaction between you and the system and what to try next etc. What I would most certainly do is observe the time it takes for things to collapse - this is not necessarily right away. Try to see whether the system handles keep increasing (indicating memory leak, though hard to imagine how this would vary per OS). Also open Resource Monitor and observe the processes in there. Open the SC processes and try to boil down what the heavy consumer is. Etc. I only know of one thing that is a general W10 issue often (sadly varying per Build) and that is Network performance. I have no experience with that in regard of CPU usage but I can expect interrupts to occur way to often, that hogging the processor and which doesn't even always show in CPU usage. In order to test this easily, start copying files just over the LAN and best would be to take a bunch of small files (but make it a 1000 or many more, if possible). Copy to and from the destination elsewhere and measure the time it takes. Always do it twice (or more) and never measure the first time. Take care that the destination has written (flushed) everything before a next attempt. Done ? then the same for the other OS. I think it will be fairly easy to detect a slow down under W10 of a couple of times. Sadly that is about all you are using (more heavily) ... the network ... Whether this (the network) is the bottleneck is still to be determined. But I would definitely want that lag out of the way. If you see network issues, try to change to another SMB protocol (tyical W10 sh*t). Etc. etc. etc. But I hope this gives you some clues. |
| [2019-01-13 09:03:40] |
| User735389 - Posts: 188 |
|
Thanks for the suggestions, but I'll probably just find a way to install win 7. I did some modifications that improved performance, like making sure the hd 'write ahead caching' was enabled, disabled a function that kept reading/writing to a sqllite db (could be an issue with the dll on w10), but it feels like i'm getting further and further into a rabbit hole that I can't get into right now. Likely will be something i can't fix anyways. Was hoping the SC guys had some ideas and pointers from their experience. If not, it is what it is. Maybe things will work better on the new win 10 laptop because of driver/hw issues, who knows. Figuring out how to get w7 to work on a new laptop will be a bit painful, but not the end of the world. BTW the speed difference is apparent immediately on startup, so not necessarily due to leaks. |
| [2019-01-13 17:20:24] |
| Merlin - Posts: 81 |
|
You're probably already doing this, but have you tried increasing the priority for sierrachart.exe? I always run it as "Above Normal" (10), and that results in noticeably faster performance, even when there doesn't seem to be anything else running on the machine. I'm also wondering whether you're running an antivirus program that's slowing Sierra down? Note that in Win 10, Microsoft's "Defender" AV runs by default if you haven't installed anything else.
|
| [2019-01-13 17:20:56] |
| User284663 - Posts: 44 |
|
Haha!! I was just googling this *unrelated to Sierra chart, just in general* as I've been sick of windows 10 after using seven, windows 10 is without a doubt wayyy slower than windows 7 and yes the fools are going to cut support, they claim windows 10 is an improvement, it's a disaster. I've been testing an open source version of Linux called Zorin OS, I highly recommend it as the OS comes pre-installed with wine, it's very stable, think windows 7 on steroids. Hey SC support, why not consider building a Sierra Chart release based on Linux I say, SC is an "engineery" type product that gets the job done, and windows looks like they are going to continue selling their retail level B/S to the masses, more and more SC users will be heading to Linux, windows is trying to hard to be hip like apple, so Linux is the new windows. |
| [2019-01-13 17:52:18] |
| |
|
Hey SC support, why not consider building a Sierra Chart release based on Linux I say, The underlying work to support that is being worked on gradually. It is a long-term process because it is quite involved.
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 |
| [2020-01-08 21:30:58] |
| JM-JO - Posts: 38 |
|
Hello fellow Sierra users, You may want to check this related topic : SC chart drawing time with Win 10 is much longer than with Win 7 |
| [2020-01-08 23:47:35] |
| ertrader - Posts: 645 |
|
Here is my experience with Linux... cannot recommend it high enough! Linux |
| [2020-01-09 00:03:09] |
| AlexPereira - Posts: 197 |
|
Are there any known optimizations or limitations between windows 10 and 7?
Did you run the test with windows 10 cpu mitigations turned off in the benchmarking? Just asking. Don't know if windows7 has the same mitigations, and depending n the cpu and application, they can make 25% to 50% slower. Here is my experience with Linux... cannot recommend it high enough!
Unfortunatly, since version 1860 and above, sierracharts has problems connecting for me. 1859 doesn't. but it seems i am the only suffering from this, so nevermind :) |
| [2020-01-09 00:13:44] |
| JM-JO - Posts: 38 |
|
Alex, Did you run the test with windows 10 cpu mitigations turned off in the benchmarking? Just asking.
Can you be more specific about cpu mitigations ? |
| [2020-01-09 08:59:28] |
| AlexPereira - Posts: 197 |
|
elji00, This was a quick search, https://www.phoronix.com/scan.php?page=article&item=spectre-melt-os&num=2 I really can't advise you to turn off mitigations for "every day usage", only for benchmarks and see if the difference comes from there. As for disabling/enabling, a search resulted in this app: https://www.grc.com/inspectre.htm , but maybe others can help, because I don't want to be responsible if something goes wrong with that app. |
| [2020-01-09 12:23:55] |
| Calculus - Posts: 86 |
|
If it's a trading PC, ie not used for much else especially online banking and such - what's wrong with using W7 even after they stop issuing security patches? That's what I'd do and I very much doubt I'd run into any problems (assuming I didn't get any emails on the PC nor any web surfing as that's where the threats generally lurk). |
| [2020-01-10 02:12:28] |
| User735389 - Posts: 188 |
|
I'm the OP. I ran into these performance issues before the spectre mess (although thats a good point, those migiationg could defintately have slowed things down even more). ive been messing with this w7/w10 thing for a few years now. I created a seperate thread somewhere about multiple monitors possibly being a factor (if the montiors/apps are running at different scaling %, the w10 scaling algos eats up cpu resources). I finally gave in to using w10, its too much of a pain to get things working on win 7 with new laptops. Had to really optimize all my studies and extend the ms per update to where its good enough and not maxing out the cpu. w7 is faster than w10 by about 1.5-3x on the base level. Practically, most users shouldn't be affected by this (maybe except consuming more battery on laptops). If you really need to max out speed and performance because you're trying to steal ticks, you'll have to go w7.
|
To post a message in this thread, you need to log in with your Sierra Chart account: