REAPER for Windows vs REAPER for Linux

[Glennbo]

Today I had to boot Windows for a moment to check something, and while there I updated REAPER. Some interesting observations I have from running REAPER almost back to back with the two OSs with the same project.

First, the latency between ASIO and JACK were both asymmetrical, which I knew was the case, but mistakenly thought they were the same.

They are not, and this is the difference.

ASIO Windows = 2.4/2.0ms
JACK Linux = 1.4/2.9ms

Then the BIGGER difference came to light. My most recent project which has a lot of instances of Guitar Rig, and instance of Kontakt, plus a bunch of Windows 32 bit plugins wouldn't play without artifacts in Windows, but plays fine in Linux.

Had I done the song in Windows, there would have been no 32 bit plugins, but it makes me wonder if 32 bit bridged plugins in 64 bit Windows is less efficient than LinVST bridged Windows 32 bit plugins running in WINE on Linux.

Sure seems like that is the case, but I will have to do a bit more testing between the two to be 100% sure.

When I went to 64 bit Windows I remember hating using bridged plugins because it seemed they ran my machine to the edge, and I ended up buying a lot of replacement plugins so I could stop using any 32 bit plugs. Today's adventure seems to reinforce that feeling.

[JamesPeters]

I posted my findings here, comparing a basic "stress test" of Reaper I'd done in Windows 7 x64 versus native Reaper in Linux.

Since then I've noticed my system in Linux is the most stable/smooth I've ever experienced, including how fast/smooth Reaper operates.

[Glennbo]

Originally Posted by JamesPeters

I posted my findings here, comparing a basic "stress test" of Reaper I'd done in Windows 7 x64 versus native Reaper in Linux.

Since then I've noticed my system in Linux is the most stable/smooth I've ever experienced, including how fast/smooth Reaper operates.

I remember your results from that.

This evaluation mostly compares 32 bit Windows plugins bridged to 64 bit REAPER in Windows vs 32 bit Windows plugins running in WINE and bridged to 64 bit REAPER in Linux.

It seems like, and I ran both with the same project a second time to confirm, that bridging WINE hosted 32 bit Windows plugins seems more efficient that bridged plugins in Windows, which was NOT what I expected to see.

A test I'd like to see some comparisons with would be to run all native 64 bit Linux plugins in 64 bit Linux vs running the exact same native 64 bit Windows plugins in 64 bit Windows, using no bridging of any kind on both OSs.

That would of course limit the test to mostly native REAPER and JS plugins, but might just be worthy of me taking the time to create a real project with lots of stuff going on. It would need to have a lot of both audio and midi tracks, and lots of identical 64 bit plugins distributed across a lot of tracks to simulate real world use.

[JamesPeters]

That's basically what I did, LOL For that particular test I only used ReaPitch. Since the initial test was done with Reaper 64-bit, the plug-in was 64-bit native to windows. The Linux version was 64-bit native to Linux. ReaPitch is a VST plugin.

I have noticed since then that all the Linux versions of native VST that I used to run in Windows run at approximately the same CPU or slightly better depending on the plug-in. Same thing with JS plugins.

And my CPU core loading is very even where is in Windows a lot of the time it was not, having one particular core handling more of the work meaning Reaper would sometimes have problems before I actually hit 100% CPU. That's probably why the test I described on that other thread work so well now, the core loading being so even in Linux.

[JamesPeters]

That's basically what I did, LOL For that particular test I only used ReaPitch. Since the initial test was done with Reaper 64-bit, the plug-in was 64-bit native to windows. The Linux version was 64-bit made it to Linux. Those are VST plugins.

I have noticed since then that all the Linux versions of native VST that I used to run in Windows (their native Windows equivalents) run at approximately the same CPU or slightly better depending on the plug-in. Same thing with JS plugins.

And my CPU core loading is very even where is in Windows a lot of the time it was not, having one particular core handling more of the work meaning Reaper would sometimes have problems before I actually hit 100% CPU. That's probably why the test I described on that other thread work so well now, the core loading being so even in Linux.

I look forward to the results of your test. I would recommend in Windows having task manager open so you can see the CPU load per core. Then compare that with a similar utility in Linux. There are lots of them as it turns out. Also make sure both systems are set with the CPU frequency at maximum, no throttling.

[Glennbo]

Originally Posted by JamesPeters

That's basically what I did, LOL For that particular test I only used ReaPitch. Since the initial test was done with Reaper 64-bit, the plug-in was 64-bit native to windows. The Linux version was 64-bit made it to Linux. Those are VST plugins.

I have noticed since then that all the Linux versions of native VST that I used to run in Windows (their native Windows equivalents) run at approximately the same CPU or slightly better depending on the plug-in. Same thing with JS plugins.

And my CPU core loading is very even where is in Windows a lot of the time it was not, having one particular core handling more of the work meaning Reaper would sometimes have problems before I actually hit 100% CPU. That's probably why the test I described on that other thread work so well now, the core loading being so even in Linux.

I look forward to the results of your test. I would recommend in Windows having task manager open so you can see the CPU load per core. Then compare that with a similar utility in Linux. There are lots of them as it turns out. Also make sure both systems are set with the CPU frequency at maximum, no throttling.

The test you did is likely indicative of the results I will probably get. I just always like to use testing as an excuse to write some more music.

Also, I like to know in advance what's going to happen when I add a lot more variables, like midi and plugins of different types and functions to the equation, so I won't be surprised when doing something really important.