First of all I would like to thank KODI, OPENELEC and OSMC teams for making these things for us, for their hard work and continuous support!
I went through and tested OPENELEC 5.95.4 and OSMC 2015.08-1 on my Raspberry Pi 2.
What I’ve found out was that OPENELEC was faster in almost anything, take a look at the times I recorded (Both systems were overclocked to the same values):
Boot:
OPENELEC - 13s
OSMC - 18s
Adding TV series to the library:
OPENELEC - 92s
OSMC - 121s
I was adding TV series from a HDD attached to the Raspberry PI 2
It felt like OSMC had better (or newer) drivers for WiFi usb adapters.
I had better speeds with Edimax EW-7811UTC.
5GHz speeds were equal (6-6.4 MB/s) but the 2.4GHz was a big difference.
With OPENELEC I achieved speeds 2.7-3 MB/s and with OSMC I achieved 5-5.3 MB/s.
These are the speeds I got using that same Wireless adapter with my laptop:
5GHz 18-20 MB/s
2.4GHz 5.3-5.5 MB/s
It feels like OSMC has more/less a better driver, but everyone lacks the real speed for 5GHz.
I understand that it’s Realtek’s fault for making crappy linux drivers.
Hopefully we’ll get better drivers in the future and get the current ones updated and improved.
You can check if you are on a 5Ghz frequency with iwlist wlan0 scan. I am working on some further improvements to the Realtek Wireless drivers which should improve things further.
We’re working on reducing the boot time as well. The scanning time for TV series is new to me. The most immediate thing I can think of is that OSMC is on Isengard and OE is on Helix. Were things faster before? We’ll take a look and see how we can improve the scanning speed. It looks like OpenELEC has had a database optimisation for some time that we have not had and that they are upstreaming that.
@sam_nazarko
I’m definitely on 5GHz, it works on both OSMC and OE, the only problem is speed which is a lot lower than one would expect, but I understand this is due to bad drivers.
Boot time is not a problem, couple of seconds is nothing to worry about.
You are right, that database optimization might be what makes OE faster with scanning.
Basically OE is 25% faster, that is what looked odd.
Please let me know if there’s anything I can do to help out!
@CaNsA
RPI 2 + Seagate HDD + 2A power supply.
That’s the hardware I used.
My router is D-Link DIR-868L.
For time testing I used Ethernet cable.
Adding to library:
I choose to set up the content of a folder which is on the HDD,when it asks me if I want to refresh information I click on yes. I measure time from that moment to the moment it’s done downloading information.
I have identified some issues in the RTL8812au driver which are impacting performance, so if you are using a device based on that chipset you can expect some better throughput soon.
@sam_nazarko
I have tested this yesterday, I can run some more tests today if needed, but OSMC was noticeably slower than OE while scanning.
Looking forward to the library scanning time improvement.
Wow, you are quite a rare person, usually everyone just accepts things as they are even if they don’t work well, I’m looking forward to your better drivers!
I’m currently waiting on Edimax EW-7822UAC and Netgear A6200-100PAS, I should get them next week and I’ll test them out.
Quite curious to see how they will perform.
That’s great. When improved scanning times make it in to OSMC they will be announced in the update announcement on the blog, I’d appreciate it if you can update me when that happens and let me know if things are any better.
@sam_nazarko
Will do, sounds good!
On the side note, do you and the OE guys help each other out regarding drivers, add-ons and such or is it every man for himself?
Again I’m not a developer or a programmer, I was struggling just to make that digital clock screensaver for Kodi which still hasn’t made it into the official repository (working on it), but I’m willing to help out if needed…
If you are interested, I can produce some test builds and we can see if library scanning improves. Aside from enabling mmap I think I have found some pragma tunables which can be patched in Kodi as well. The more testers we get the better and if your library is consistent then this will help immensely.
Sure thing Sam,whenever you’re ready let me know.
On the side note:
How complicated is fixing wireless drivers?
I imagine pretty much, especially if they were crap right from the start.
The most interesting thing about the WiFi drivers that users sometimes forget is that I don’t have all of them. Most drivers can be fixed by:
Moving in-tree to out of tree.
Backporting fixes
Adding VID/PID combos
I am hoping to significantly improve our scanning performance in the September update. This needs to be done in two stages:
Testing my changes to Kodi which should improve scan performance
Adding the accepted, known fixes to SQLite which are now upstreamed.
The reason to separate the two is to prove that my first patch is indeed beneficial, and we are not just seeing benefits from the second. Conversely, another reason to separate them is to also ensure that both sets of changes are beneficial.
I am now compiling my changes to Kodi and Jenkins should make those builds available soon (assuming they compile OK).
If you’re happy to do so, then I can produce a couple of test builds (this is the first of a few) and we can work together on improving scanning performance.
Sounds good Sam, you can reach me here or on my email (I guess you have access to my account).
Whenever you’re ready let me know.
I’m setting up a special library just for testing purposes.
That should give us reliable results.
Of course it’s not, boot time was not something I was really interested in anyway, but since I measured it I shared it with everyone.
The more important thing was scanning time.