Go Back   Cockos Incorporated Forums > REAPER Forums > Recording Technologies and Techniques

Reply
 
Thread Tools Display Modes
Old 04-04-2017, 12:13 AM   #1
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default RME RayDAT low latency performance

Im using an RME RayDAT with an audio rack PC. I have had it for quite a while now, and real time performance is driving me nuts. Im not sure if its Windows (im running Windows 7 64 bit), the PC itself, the DAW (Reaper 5.4 64 bit), the VSTs (I mainly use Guitar Rig 5 64 bit)

Its driving me insane. I have spent around $4000 AU on the PC, plus another $1000 on the RayDAT, I have experimented with different PCIe slots, BIOS settings, disabling Windows Services, and Reaper ASIO settings, and I still cant consistently get a stable, click free performance at 32 samples buffer size when the real time CPU goes above around 30-40%. The whole reason I spent all this money was for that exact purpose - to run as many instances of Guitar Rig 5 at 32 samples buffer size as possible, and Im losing all hope I will ever get anywhere with this goal.

I need a PRO who knows PC/CPU/Motherboard/BIOS/Driver/Windows architecture back to front who can help me optimise my machine/software, and Im happy to pay for their time/services, but for the love of god I cant seem to find anyone. There must be someone out there who can help me go through my rig step by step, and tell me what the best possible settings/drivers/optimisations are for my specific hardware. The only people that seem to know anything about this kind of low latency performance optimisation are people who build custom DAW PCs, and they arent keen on sharing their secrets, which is fair enough, but it leaves me with pretty much no options.

I honestly feel like giving up and selling my PC and the RME, the frustration is filling me with so much negativity its crippling me creatively/musically.

I guess this post is a last ditch effort to find someone who is willing to spend some time working through this with me, for a reasonable hourly rate, before I give up entirely and sell or smash all this garbage to bits. Too dramatic probably but Im really at the end of my rope.

If anyone thinks they can help, here is a list of the parts in my rack PC

https://au.pcpartpicker.com/user/mat...s/saved/gfvRBm

As mentioned Im using Windows 7.1 64bit, with Reaper 5.40 64bit, and Native Instruments Komplete 10 (mainly Guitar Rig 5.2 and Kontakt)
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Old 04-04-2017, 01:55 AM   #2
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

Have you disabled everything you're not going to use in Device Manager? If you are looking at purely low latency performance then that will include bluetooth, Wi-Fi and Ethernet. Just right-click and disable. Nothing is permanent with this, you can re-enable them at any point.

Are these live, rec-armed tracks? If so, then try enabling "Live multiprocessing using X CPUs" in Preferences->Buffering. X should be the number of real cores you have. Or half that, whatever works best.

It's worth pointing out that the RTL of your interface at 128 samples is around the same as a Komplete Audio 6 at 32 samples.
snooks is offline   Reply With Quote
Old 04-04-2017, 02:01 AM   #3
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default

Oh Ive gone apeshit over the last few months with Reaper, Windows and BIOS settings... including but not limited to:

Disabling/Enabling Hyperthreading
Disabling Speedstep, C-States etc
Disabling ALL non-essential devices (LAN, Wifi, Bluetooth, onboard audio)
Disabling almost every Windows Service
Turning on/off HPET
All kinds of Reaper settings (multithreading, thread priority, MMCSS, buffering options, bla bla)
Updating drivers
The list goes on...

the issue is that there are SO many variables here and im by no means an expert, so im just constantly guessing, stabbing in the dark hoping that some magic combination will be the answer, but itll take me a million years cause i have no idea what im doing

As for the speed of my interface, the thing is its just a digital interface, I still have to use an external AD/DA converter, which adds another 50 or so samples. the RME running at anything above a buffer of 64 samples buffer isnt acceptable for my needs.
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Old 04-04-2017, 03:15 AM   #4
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

I think it would be good if somebody with a pro audio desktop could share what their system craps out at. For me it's around 25% RT CPU at 48KHz, 32 samples. My max DPC is around 90 microseconds, ticking along at 20/30 or whatever. So it can't get that much better - 0.1ms (DPC) isn't 75% of 0.7ms (size of buffer in time) anyway. Your 35% is about 50% better than that, which might be the PCIe vs USB or the desktop processor.
snooks is offline   Reply With Quote
Old 04-04-2017, 04:26 AM   #5
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default

well im getting around 30-40% RT CPU before it craps out at 32 samples 44.1Khz. my specs are listed here

https://au.pcpartpicker.com/user/mat...s/saved/gfvRBm

that really doesnt seem reasonable given the relatively good quality of my rack PC, as well as how little the system is being taxed. RAM is often only 2-3GB max (out of 16GB)

also my DPC latency is usually between 5-20 microseconds, with occasional spikes around 40, so i dont see why that would be impacting real time performance. i do see a lot of page faults in LatencyMon but I have no idea how to investigate the cause or fix that. also i notice that the Nvidia, USB and SATA drivers cause the highest reported DPC execution times, sometimes these are listed at around 300-400 microseconds, but the actual DPC latency max is never usually much more than 40
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Old 04-04-2017, 04:59 AM   #6
cyrano
Human being with feelings
 
cyrano's Avatar
 
Join Date: Jun 2011
Location: Belgium
Posts: 5,246
Default

Have you tried posting to the RME forum?

https://www.forum.rme-audio.de/index.php

Maybe you just have one of those motherboards that don't seem to work for audio? Several people on the RME forum have gone through the same kind of ordeal.
__________________
In a time of deceit telling the truth is a revolutionary act.
George Orwell
cyrano is offline   Reply With Quote
Old 04-04-2017, 05:09 AM   #7
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

That's just about as good as it gets with DPC latency. Well, apart from the spikes which are the limiting factor.

Last edited by snooks; 04-04-2017 at 05:28 AM. Reason: edit: yup, LatencyMon
snooks is offline   Reply With Quote
Old 04-04-2017, 07:01 AM   #8
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

I think the RT CPU isn't giving the best indication of what's going on.

eg here, playing a project with a 64 sample buffer at 48KHz, I see the RT longest block is 0.2ms out of 1.33ms. That's 15%, but the RT CPU is around 2.5%. I then rec arm a bunch of tracks to increase the load (live multi processing off) so that there is 0.8ms out of 1.33ms, which is 60% of the possible time in the buffer but the RT CPU is 19%.
snooks is offline   Reply With Quote
Old 04-04-2017, 07:09 AM   #9
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
to run as many instances of Guitar Rig 5 at 32 samples buffer size as possible,
How many is that and why do you need that number (whatever it is) running at that buffer size, are you playing a live gig or something? As an FYI scaling up does not provide a linear result.
__________________
Music is what feelings sound like.

Last edited by karbomusic; 04-04-2017 at 07:21 AM.
karbomusic is offline   Reply With Quote
Old 04-04-2017, 07:16 AM   #10
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by snooks View Post
I think the RT CPU isn't giving the best indication of what's going on.
It's only real use is to determine if it is maxed out while servicing the audio driver thread, if it isn't maxed, it has little value. Use Process Explorer then double click Reaper.exe to get to the threads tab and see per thread CPU. The idea of near unlimited GR5 instances @ 32 samples is a ceiling just asking to be hit though so we need to know how many it fails at and the reason for needing 32 samples for all of them since there is no need for this when mixing unless there is some need for super-tight fx automation during playback.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-04-2017, 03:54 PM   #11
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

Yes, I never do anything at uber low latencies but testing I see that a full 30 track mix with around 80 FX is doable at 48KHz/24 samples on my laptop/'pute. Also 32 tracks (duplicating 2 ins * 16 so not too non-ridiculous) with full routing through 2 FX per channel through to FX and output busses with all live monitoring is possible click free. But I can't run more than 2 Guitar Rig 5s on the "Almost Clean" preset at 32 samples without clicks. After the 2 live monitored Guitar Rigs I can insert dozens of additional plugins though.

So like with VIs, CPU heavy VSTs and live monitoring at uber low latencies is not something you expect great results with. Acceptable results, yes. But when you can insert around 120 ReaComps for every Guitar Rig 5 it's comparing apples and oranges if we pay attention to the guy who can run hundreds of plugins at the lowest buffer settings.

TL/DR; 2 Guitar Rig 5s on an i7-4720HQ pegged at 3.4GHz running at 48KHz/32 samples.
snooks is offline   Reply With Quote
Old 04-04-2017, 07:41 PM   #12
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Quote:
Originally Posted by matthewjumpsoffbuildings View Post
Oh Ive gone apeshit over the last few months with Reaper, Windows and BIOS settings... including but not limited to:

Disabling/Enabling Hyperthreading
Disabling Speedstep, C-States etc
Disabling ALL non-essential devices (LAN, Wifi, Bluetooth, onboard audio)
Disabling almost every Windows Service
Turning on/off HPET
All kinds of Reaper settings (multithreading, thread priority, MMCSS, buffering options, bla bla)
Updating drivers
The list goes on...

the issue is that there are SO many variables here and im by no means an expert, so im just constantly guessing, stabbing in the dark hoping that some magic combination will be the answer, but itll take me a million years cause i have no idea what im doing

As for the speed of my interface, the thing is its just a digital interface, I still have to use an external AD/DA converter, which adds another 50 or so samples. the RME running at anything above a buffer of 64 samples buffer isnt acceptable for my needs.
OK, but did you actually do what he suggested?

Anyway, the first two questions I have are what is your overall system cpu load in task manager when you hear the glitching, and have you actually done a round trip latency test (oblique rtl utility is good). What latency are you getting at 32, and 64 samples buffer?

Anyway, more detail would certainly be helpful. Many isn't a number, and you don't state how much overall cpu load there is.

Last edited by drumphil; 04-04-2017 at 09:02 PM.
drumphil is offline   Reply With Quote
Old 04-04-2017, 07:43 PM   #13
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by snooks View Post

So like with VIs, CPU heavy VSTs and live monitoring at uber low latencies is not something you expect great results with. Acceptable results, yes. But when you can insert around 120 ReaComps for every Guitar Rig 5 it's comparing apples and oranges if we pay attention to the guy who can run hundreds of plugins at the lowest buffer settings.
I agree and the reason I mentioned was because it was GR5 and not some lighter VST(s). If Reaper and Native (in this) case offered public PDBs we could run Windows Performance Analyzer (WPA) and see the exact CPU timings for every single method that executes but based on what I saw concerning the setup, GR5 * lots of them @ 32 samples... feels like there wouldn't be a remarkable finding. I can't say it couldn't be better but it doesn't necessarily smell funny to me either.

Also, once the CPU churn is close enough to the hardware, the ratio/distance between kernel time and user mode time should get smaller aka more CPU time spent in kernel than normal. If all the CPU is up in user mode, then hardware/interrupts etc. are likely not involved.
__________________
Music is what feelings sound like.

Last edited by karbomusic; 04-04-2017 at 07:49 PM.
karbomusic is offline   Reply With Quote
Old 04-04-2017, 11:59 PM   #14
heda
Human being with feelings
 
heda's Avatar
 
Join Date: Jun 2012
Location: Spain
Posts: 7,239
Default

Are those instances of Guitar Rig armed for recording? Does improve when changing the "Allow live FX multiprocessing" in Preferences Buffering ?
heda is offline   Reply With Quote
Old 04-05-2017, 11:01 AM   #15
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

For me, with "Allow live FX multiprocessing" off I can only load 1 live monitored instance of GR5 ("Almost Clean" preset) without crackles. With it set to 4, I can load 2. Audio processing threads set to 8.
snooks is offline   Reply With Quote
Old 04-05-2017, 11:44 AM   #16
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

I have GR5 and an RME (UFX), I'm happy to test later if I have time but my interface only goes down to 64 samples - I restrained myself from asking why 32 is required from a practical perspective - the difference between 32/64 should be completely irrelevant as a player unless it is in addition to some other unknown or unaccounted for latency essentially taking it over the edge.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-05-2017, 04:34 PM   #17
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

I'd be interested in your results. I'll see how many live monitored GR5s I can spin up at 64 samples too. I agree that an extra 1.3ms going from 32 to 64 samples isn't something to worry about.
snooks is offline   Reply With Quote
Old 04-05-2017, 06:06 PM   #18
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Excuse the haste, I'm trying to test a couple mics that I just had to go retrieve from the shipper. I stopped at 9 GR5s, recording live. I'm pretty sure it would start popping/clicking soon after. I didn't expect that many to work though. I saved the project so when I get some more time I might be able to see where it truly breaks.

Being in a hurry, I literally dropped in GR5, chose some random preset and duplicated till there were 9 tracks and hit record @ 64 samples/48k...



__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 01:45 AM   #19
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

Cool, please let us know what preset you are using if you get time to produce crackles. When using the lower CPU "Clean Plex" instead of "Almost Clean" for example I can run 8 instances at 32 samples. I'm not using HQ mode in GR5 for any of this.
snooks is offline   Reply With Quote
Old 04-06-2017, 06:37 AM   #20
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by snooks View Post
Cool, please let us know what preset you are using if you get time to produce crackles. When using the lower CPU "Clean Plex" instead of "Almost Clean" for example I can run 8 instances at 32 samples. I'm not using HQ mode in GR5 for any of this.
My limit is right about 16 tracks, 64 samples/48k using the preset called "80s solo" in non-HQ mode. That's with my current setup/config, stock windows 10 with no tweaks, wifi on etc. Most of the reaper settings as far as buffering and audio are at their defaults. I think the only major tweak I have done is setting thread behavior to Automatic (Experimental). There are probably a couple others but I've used this config for years. I know I've never heavily tweaked Reaper and haven't tweaked the OS at all. It's also my dev box so there is an awful lot of stuff installed and running in the bg unrelated to audio FWIW.
__________________
Music is what feelings sound like.

Last edited by karbomusic; 04-06-2017 at 07:32 AM.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 01:04 PM   #21
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

I've disabled some stuff on the laptop in Device Manager. ACPI, Bluetooth, webcam, Ethernet and NVidia audio. With Wi-Fi on I get 16 live monitored tracks/instances recording completely crackle free at 48KHz/64 samples with "80s Solo" preset. With Wi-Fi disabled in Device Manager it's 24.

Oh yeah, that's with ThrottleStop pegging the processor at slightly less than 3.6GHz, not 3.4 like I said earlier. So a bit of laptop ointment has been applied.

Out of interest, what processor is that you have?
snooks is offline   Reply With Quote
Old 04-06-2017, 01:20 PM   #22
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by snooks View Post
Out of interest, what processor is that you have?
Crap, I'll have to look when I get home but I got the machine from Jim Roseberry if you remember him from the forum? (purrfect audio). It's running overclocked @ 4.5 Ghz, no throttling. Come to think about it, I should refer the OP to him because about the only thing he does is find hardware combinations that are guaranteed to work at ~n samples latency.

I should probably confirm if I have throttling disabled, been too long since I checked but it looks like the OP has some issues to potentially sus out - if only we knew how many instances his started crackling at.
__________________
Music is what feelings sound like.

Last edited by karbomusic; 04-06-2017 at 01:27 PM.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 03:34 PM   #23
snooks
Banned
 
Join Date: Sep 2015
Posts: 1,650
Default

Good stuff, hopefully this will give matthewjumpsoffbuildings something more concrete to base his judgement of his 'pute on.
snooks is offline   Reply With Quote
Old 04-06-2017, 04:55 PM   #24
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Some throttling appears to be on (dropped to 1600 MHz a time or few), I'll leave as is for now since it's not causing me any problems. Here's the basic CPU info either way...

__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 05:36 PM   #25
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Hey karbomusic, what cpu load do you reach when the crackles start?
drumphil is offline   Reply With Quote
Old 04-06-2017, 05:48 PM   #26
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by drumphil View Post
Hey karbomusic, what cpu load do you reach when the crackles start?
23% or so but the pops/clicks are due to the audio driver thread being maxed out (in this case record armed) which can only live on one core. Remember when I was saying up ^there somewhere about it being closer to hardware etc. when kernel time is high? See the screenshot below and notice the first core is showing that about 90% of that core's CPU time is spent down in kernel servicing hardware etc. (kernel = red). When CPU isn't near the hardware (up in user mode), it will appear more like the fifth core does....



Meaning, what I'm seeing is expected (by me anyway). Once I just play back and remove rec-arm the kernel time drops enough that the crackles go away, again expected. Once we moved beyond single CPU machines, total CPU % is fairly useless for judging pops and clicks.
__________________
Music is what feelings sound like.

Last edited by karbomusic; 04-06-2017 at 06:29 PM.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 06:31 PM   #27
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by snooks View Post
Good stuff, hopefully this will give matthewjumpsoffbuildings something more concrete to base his judgement of his 'pute on.
Looking back at his post, he should begin with a good baseline, aka a fresh portable install + nothing else except GR5 with all default settings (including any non-standard windows changes).

It's incredibly easy to back one's self into a performance corner when we go out and just start making every tweak know to man, then the next, then the next. It isn't organized and the unknowns cause more harm than good. As evidenced by my near 100% stock install running fairly well in comparison on a heavily utilized development machine (I'm showing > 350 programs installed not including updates) w/zero services disabled.

I'm really curious as to how much testing was done before any changes were made or if the changes were assumed needed and made beforehand.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-06-2017, 08:34 PM   #28
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Quote:
Originally Posted by karbomusic View Post
23% or so but the pops/clicks are due to the audio driver thread being maxed out (in this case record armed) which can only live on one core. Remember when I was saying up ^there somewhere about it being closer to hardware etc. when kernel time is high? See the screenshot below and notice the first core is showing that about 90% of that core's CPU time is spent down in kernel servicing hardware etc. (kernel = red). When CPU isn't near the hardware (up in user mode), it will appear more like the fifth core does....



Meaning, what I'm seeing is expected (by me anyway). Once I just play back and remove rec-arm the kernel time drops enough that the crackles go away, again expected. Once we moved beyond single CPU machines, total CPU % is fairly useless for judging pops and clicks.
Yeah, I get all that, I just like to have all the variables in my head when I try and understand a specific situation.
drumphil is offline   Reply With Quote
Old 04-07-2017, 07:25 AM   #29
shoyoninja
Human being with feelings
 
Join Date: Mar 2010
Posts: 431
Default

Hi.

Saw the post and the part list. I notice you are using a GT710 as the graphics device.


Please go the device manager and set windows to work with its "default" driver, uninstall the graphics card and let the Standard VGA adapter, or whatever its called, to run it.

I had a LOT of frustration on my previous audio PC due to a low end GFX card from Nvidia (quite similar to yours). To work with audio I could not use its drivers, and DPC did not report any significant issues.
shoyoninja is offline   Reply With Quote
Old 04-07-2017, 07:30 AM   #30
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by drumphil View Post
Yeah, I get all that, I just like to have all the variables in my head when I try and understand a specific situation.
np, I was mostly typing all that for those who may not be aware.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-07-2017, 05:08 PM   #31
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Uh, where did the op go?
drumphil is offline   Reply With Quote
Old 04-07-2017, 05:13 PM   #32
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by drumphil View Post
Uh, where did the op go?
Hopefully he didn't jump of that building.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-07-2017, 05:32 PM   #33
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default

Hey guys im back, Ive been testing some stuff out. I certainly feel like jumping of a building sometimes trying to sort this out :P

Uninstalling my nvidia drivers completely results in my LatencyMon DPC tests going WAAAY down. The problem is Im using the GT 710 for dual monitors, and now only one of the monitor outputs is working :/

Do you think upgrading to a better GPU (was thinking maybe a GTX 1050), would help with this? Or are the nvidia drivers always an issue, making changing hardware pointless

Also im testing out Reapers anticipative FX feature to take the load off realtime CPU. obviously this only works with pre-rendered tracks, so its not a solution for live monitoring through VSTS. But it definitely helps when mixing large existing projects. I do however notice that when Im using anticipative FX, if I have a large-ish project, and I seek/start/stop, there is often quite a noticeable crackle, sometimes its around a second of crackles in quick succession that sort of sounds like buzzing, sometimes more of a stutter/repeat.

@karbomusic - a lot of the changes were made based on advice from forums (gearslutz, RME forums etc). I could reset the BIOS settings, restore all windows services etc, and then go through the changes one by one, testing with LatencyMon each time.

Its just really hard to be sure, because the order I test things in might make a huge difference. For example - what if a certain BIOS setting has HUGE benefits but ONLY if a specific windows service is disabled? If I test the windows service first, I wont see any performance difference because the BIOS setting hasnt been changed, so I will re-enable the service thinking it makes no difference. Then Ill test the BIOS setting and it wont make a difference either, since the windows service is enabled. See what I mean? Its a testing nightmare. The chain is basically

BIOS > Drivers > Windows > DAW

But each of those steps could affect/be affected by anything in the previous or next steps
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Old 04-07-2017, 05:41 PM   #34
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
@karbomusic - a lot of the changes were made based on advice from forums (gearslutz, RME forums etc). I could reset the BIOS settings, restore all windows services etc, and then go through the changes one by one, testing with LatencyMon each time.
Just to be clear before you go drastic. I always like to see how things work before doing tweaks; it provides a known baseline (where did we start) and so on. If you tried it non-tweaked first and had these problems then it's a little less important to revert. Sometimes when nothing seems to help, even default settings, it can be a hardware related of some type like the video card test just showed.

However, it is very easy to just drop a vanilla portable install of Reaper to a folder and test that, that will be easy and worthwhile. I do hate to hear you are having problems, there have been problems I have had in the distant past so I do empathize but I'm confident you'll eventually figure it out if you stay steady and methodical.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-07-2017, 10:26 PM   #35
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Exactly how many instance are you able to run? How much load is on the other cores when you get crackling?

Regarding the graphics card issue. I have a system with a 710 and a system with a 1050. The 1050 does better on DPC latency overall, but both will produce a spike if I start a program like firefox that engages the GPU. Not a problem if you don't want to start firefox in the middie of a recording.

Your onboard video does support multiple monitors. You can get an cheap passive cable or adapter to convert HDMI to DVI or vice versa if your monitors are missing one or the other.

It's also important to note that you shouldn't run the DPC latency monitoring program while you test your low latency performance. It will cause pops and clicks and crackles that wouldn't otherwise occur at a given system load.

Also, the intel network adapters have been known to cause dpc latency problems at times, so it is definitely worth disabling that to see if it makes any difference.

None of this may matter if you are just expecting too much from the system which is why I'm curious to see your per core with kernel times CPU usage as reported in task manager, and know how many instances you are able to run successfully.

Do the problems go away if you use a 64 sample buffer? Does it run any better at all? Or is the problem the same.

Also, is it possible that the plugins you are using only output certain sample rates, and that your project is at a different sample rate requiring that it be resampled on the fly? What sample rate are you using?

Last edited by drumphil; 04-07-2017 at 10:50 PM.
drumphil is offline   Reply With Quote
Old 04-08-2017, 02:33 PM   #36
shoyoninja
Human being with feelings
 
Join Date: Mar 2010
Posts: 431
Default

Quote:
Originally Posted by matthewjumpsoffbuildings View Post
Uninstalling my nvidia drivers completely results in my LatencyMon DPC tests going WAAAY down. The problem is Im using the GT 710 for dual monitors, and now only one of the monitor outputs is working :/
In my case it did the trick, but I changed over to Radeon cards to avoid surprises.
shoyoninja is offline   Reply With Quote
Old 04-12-2017, 04:53 AM   #37
drumphil
Human being with feelings
 
drumphil's Avatar
 
Join Date: Jun 2006
Location: Australia
Posts: 3,738
Default

Well, this is going to be the slowest troubleshooting session ever.
drumphil is offline   Reply With Quote
Old 04-12-2017, 07:24 AM   #38
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default

Quote:
Originally Posted by drumphil View Post
Exactly how many instance are you able to run? How much load is on the other cores when you get crackling?
I can run around 25 instances of Guitar Rig 5 in low quality mode using the "Billy Dual Grange" preset. When I get to around 30 the crackles start. This seems like a lot but in many projects I will often end up with 20 instances of it, plus a bunch of other FX, so Guitar Rig is a good one to test performance with. When doing this, Reapers performance monitor reports around 50% Real Time CPU, and around 25% Total CPU.

Quote:
Originally Posted by drumphil View Post
Regarding the graphics card issue. I have a system with a 710 and a system with a 1050. The 1050 does better on DPC latency overall, but both will produce a spike if I start a program like firefox that engages the GPU. Not a problem if you don't want to start firefox in the middie of a recording.
Thats good to know, this is a rack PC with no networking, its only used for audio and Ill never be opening any other programs other than Reaper on it

Quote:
Originally Posted by drumphil View Post
It's also important to note that you shouldn't run the DPC latency monitoring program while you test your low latency performance. It will cause pops and clicks and crackles that wouldn't otherwise occur at a given system load.
Yes Im aware of this, I only run LatencyMon by itself or Reaper by itself when testing

Quote:
Originally Posted by drumphil View Post
Also, the intel network adapters have been known to cause dpc latency problems at times, so it is definitely worth disabling that to see if it makes any difference.
Yes, I have already disabled a lot of stuff - Onboard Audio, LAN/Network controllers, Intel Graphics, etc, as well as C-States, Speedstep etc in the BIOS

Quote:
Originally Posted by drumphil View Post
None of this may matter if you are just expecting too much from the system which is why I'm curious to see your per core with kernel times CPU usage as reported in task manager, and know how many instances you are able to run successfully.
Ye the CPU is not taxed that hard

Quote:
Originally Posted by drumphil View Post
Do the problems go away if you use a 64 sample buffer? Does it run any better at all? Or is the problem the same.
It gets a little better at 64, but its not completely stable at 64 either.

Quote:
Originally Posted by drumphil View Post
Also, is it possible that the plugins you are using only output certain sample rates, and that your project is at a different sample rate requiring that it be resampled on the fly? What sample rate are you using?
Guitar Rig 5 definitely works at 44.1Khz, which is my standard sample rate
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Old 04-12-2017, 07:39 AM   #39
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Quote:
Originally Posted by matthewjumpsoffbuildings View Post
I can run around 25 instances of Guitar Rig 5 in low quality mode using the "Billy Dual Grange" preset. When I get to around 30 the crackles start. This seems like a lot but in many projects I will often end up with 20 instances of it, plus a bunch of other FX, so Guitar Rig is a good one to test performance with. When doing this, Reapers performance monitor reports around 50% Real Time CPU, and around 25% Total CPU.
When record armed or during playback with no tracks record armed? Either way I find that pretty damn decent performance, now why do you need 30 GR5s going at once? That definitely falls into the somewhat extreme category. There will always be a ceiling you can eventually hit and these days scaling up is reaching it's limits where it becomes a diminishing returns scenario.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 04-12-2017, 07:43 AM   #40
matthewjumpsoffbuildings
Human being with feelings
 
Join Date: Mar 2013
Posts: 463
Default

This is not record armed tracks. I dont have anticipative effects turned on though, as I find this gives lots of crackling on seeking/stopping playback, especially with reverb tails

I dont need 30 GRs at once, though as I said its not unlikely I will have 20 in some projects. Its just that I use GR often and its a good VST to test performance with. Its definitely not unlikely that I will have 30 plugins that use around as much CPU as GR running in a project. And as I said my total CPU is only around 25%, with Real Time CPU coming in around 50%. Surely my system can do better than start to fall apart at that low utilisation level?
__________________
[sun sprites] spotify facebook
matthewjumpsoffbuildings is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 08:30 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.