7.5 Bug: VST performance peak?

Everything else...

Re: 7.5 Bug: VST performance peak?

Postby Fabio Bartolini » Thu Feb 06, 2014 10:33 am

bugishcompany wrote:FABIO:

Do you know if Steinberg is trying to implement a "bridge" in ASIO-Guard for audio processing on CUDA enabled graphic cards?

This could explain why graphics performance seems so linked to plugin performance in version 7.


Hello,

no, I'm not aware of CUDA implementations being worked on.

The graphic performance appears to be linked to the latest drivers' implementation, with some of the improvements made for gaming and 3D graphics causing real-time spikes with particular plug-ins.
These plug-ins all share common characteristics - they over-ride the default behaviour of the pointer when hovering on the controls, so it might be related to a particular GUI framework.
I'm no developer, so I forwarded the outcome of my tests to the proper people in-house.

They were actually already on it and the GUI handling is being revised - it is quite unlikely that the drivers will be written the "old way" in the future, so it must adapt to the way drivers are working now and will probably work in the near future.

By the way, during my tests the behaviour of Cubase 7/7.5 and Cubase 6/6.5 was the same concerning this.
But Cubase 6/6.5 only shows the average VST Performance, so the spikes and fluctuations might look like a normal fluctuation due to the load, especially with a lot of plug-in instances or working at low latencies.
When looking at the Cubase 7/7.5 real-time meter, it might trick into thinking that the performance is much different, but the only way to directly compare the two versions is by comparing the average meter only.

Hope this helps.
Fabio Bartolini, Steinberg Tech Support
Steinberg Media Technologies GmbH
Hamburg, Germany

Check out Steinberg on YouTube, Twitter and Facebook!
User avatar
Fabio Bartolini
Moderator
 
Posts: 285
Joined: Tue Oct 02, 2012 11:55 am
Has thanked: 20 times
Been thanked: 72 times

Re: 7.5 Bug: VST performance peak?

Postby mroekalea » Thu Feb 06, 2014 10:49 am

Doing some research with Google shows otherwise, try:https://www.google.nl/#q=aero+support+dropped+in+windows+8
Menoj
Win8.1, 3930K@3.2Ghz, 32GB with Cubase 7.5, wavelab 8 elements, HSO, MR816X+Focusrite pro 26 IO, Adam P22a + sub8, CC121, Cubase IC pro, Halion 5, VG2, Arturia AF, TRacks CS grand + all custom elements, BFD3, Amplitube 3 (fender, orange, SVX collections, slash), HSO library, Fender Jazz basses, MB octaver, fender bassman 100, Orange AD200B, MB TTE500, TECAMP, Fender & FMC 12 cabs!
mroekalea
Member
 
Posts: 508
Joined: Thu Dec 16, 2010 11:12 am
Location: Netherlands
Has thanked: 3 times
Been thanked: 19 times

Re: 7.5 Bug: VST performance peak?

Postby Freddie H » Thu Feb 06, 2014 10:51 am

mroekalea wrote:Doing some research with Google shows otherwise, try:https://www.google.nl/#q=aero+support+dropped+in+windows+8



Glass looks dropped (unfortunately) Not AERO itself!
-Highly developed spirits often encounter resistance from mediocre minds. -It really matters!
User avatar
Freddie H
Member
 
Posts: 426
Joined: Fri Apr 08, 2011 7:59 am
Has thanked: 52 times
Been thanked: 70 times

Re: 7.5 Bug: VST performance peak?

Postby mroekalea » Thu Feb 06, 2014 10:53 am

Ok point taken. Good that you're sharp today! ;-)
Menoj
Win8.1, 3930K@3.2Ghz, 32GB with Cubase 7.5, wavelab 8 elements, HSO, MR816X+Focusrite pro 26 IO, Adam P22a + sub8, CC121, Cubase IC pro, Halion 5, VG2, Arturia AF, TRacks CS grand + all custom elements, BFD3, Amplitube 3 (fender, orange, SVX collections, slash), HSO library, Fender Jazz basses, MB octaver, fender bassman 100, Orange AD200B, MB TTE500, TECAMP, Fender & FMC 12 cabs!
mroekalea
Member
 
Posts: 508
Joined: Thu Dec 16, 2010 11:12 am
Location: Netherlands
Has thanked: 3 times
Been thanked: 19 times

Re: 7.5 Bug: VST performance peak?

Postby Freddie H » Thu Feb 06, 2014 11:01 am

mroekalea wrote:Ok point taken. Good that you're sharp today! ;-)



I just try to help! ;)

By the way.. Next wrong TWEAK I can hint you about.

"Program scheduler"
shall be set to ------> PROGRAM on Windows 7 and Windows 8.
Not "Background Task" like on XP32. ASIO driver is not a "background task".

I can tell you the story about why and so forth just take my word for it. Have it set to wrong setting, "Background Task" on Win 7 and Win 8 will cause problems and will make drop outs on Audio too.


Best Regards
Freddie
-Highly developed spirits often encounter resistance from mediocre minds. -It really matters!
User avatar
Freddie H
Member
 
Posts: 426
Joined: Fri Apr 08, 2011 7:59 am
Has thanked: 52 times
Been thanked: 70 times

Re: 7.5 Bug: VST performance peak?

Postby mroekalea » Thu Feb 06, 2014 11:19 am

Thanks, I have this setting indeed set to background services, will try (although I do not have drop-outs), just curious in what happens.

Greetings,
Menoj
Menoj
Win8.1, 3930K@3.2Ghz, 32GB with Cubase 7.5, wavelab 8 elements, HSO, MR816X+Focusrite pro 26 IO, Adam P22a + sub8, CC121, Cubase IC pro, Halion 5, VG2, Arturia AF, TRacks CS grand + all custom elements, BFD3, Amplitube 3 (fender, orange, SVX collections, slash), HSO library, Fender Jazz basses, MB octaver, fender bassman 100, Orange AD200B, MB TTE500, TECAMP, Fender & FMC 12 cabs!
mroekalea
Member
 
Posts: 508
Joined: Thu Dec 16, 2010 11:12 am
Location: Netherlands
Has thanked: 3 times
Been thanked: 19 times

Re: 7.5 Bug: VST performance peak?

Postby charlesvk » Thu Feb 06, 2014 12:58 pm

@Freddie: Can you tell me how I can get there to change this setting please? So I can try this.

And overall: There is a lot of talk about mouse movements related to performance respons, but I started the thread while I had difficulties using vsti´s (like Omnisphere) with a midicontroller, no mouse. I´m not saying there´s no relationship, but I feel the thread is going into another direction now. Correct me if I´m wrong.
Dell XPS Studio 8100 i7 860@2.80GHz, 16GB RAM, W7 Ultimate SP1 64-bit, Motu 828mk3, Cubase 7.5 64-bit
charlesvk
Junior Member
 
Posts: 98
Joined: Fri Aug 19, 2011 2:23 am
Location: The Netherlands
Has thanked: 24 times
Been thanked: 4 times

Re: 7.5 Bug: VST performance peak?

Postby Dave » Thu Feb 06, 2014 2:23 pm

Hi charlesvk

To change to Programs from Background on Win7

Control Panel
System and Security
System
Advanced System Setttings
Performance Settings Button
Advanced Tag
Select Programs

Hope this helps

Cheers
Dave
Cubase 7.5.10, CC121, MR816CSX, Windows 7 32&64 iMac OSX 10.9 Maverick.
Wavelab 8.03, Groove Agent 3, HSO, Halion 5, Grand 3, Halion Sonic 2, Cubasis for iPad, ICPro other plugins.

Keep smiling its later than you think.
Dave
Junior Member
 
Posts: 132
Joined: Sun Dec 19, 2010 3:39 pm
Has thanked: 4 times
Been thanked: 0 time

Re: 7.5 Bug: VST performance peak?

Postby Freddie H » Thu Feb 06, 2014 3:09 pm

charlesvk wrote:@Freddie: Can you tell me how I can get there to change this setting please? So I can try this.

And overall: There is a lot of talk about mouse movements related to performance respons, but I started the thread while I had difficulties using vsti´s (like Omnisphere) with a midicontroller, no mouse. I´m not saying there´s no relationship, but I feel the thread is going into another direction now. Correct me if I´m wrong.



Here you go!

http://www.eightforums.com/tutorials/24 ... ndows.html


Processor scheduling on Vista, Windows 7, or Windows 8 manage processor resources for the best performance.
It shall be on "Programs" because ASIO-driver is not background services on these systems.


ASIO-driver on XP32 is another story. The "Processor scheduling" should be set to "Background services" ON XP32 ONLY instead. But that is only on XP32 only.
It were RME Mattias Carstens @ RME that found this tweak on XP32. It was meant to solve driver issue with specific RME cards in 90ths with XP32.

If you ask Mattias he also will say exactly as I do on Windows 7 and Windows 8. It shall be set to "Programs" because ASIO-drivers is not background services on these new systems.

Still people keep doing all these old XP tweaks on their new systems with out having any clue "why" or "what" they continue to do so? Then they come to all kinds of FORUMS, moan that every s-ucks...and I don't know what... PC and Cubase sucks too. It actually comes down to that they have actually f-u-cked up the system them self with all XP tweaks that they shouldn't had done in the first place.


Anyway, I hope now that at least some of you use the correct information I gave you.
If you do everything will work better! ;)


Best Regards
Freddie
-Highly developed spirits often encounter resistance from mediocre minds. -It really matters!
User avatar
Freddie H
Member
 
Posts: 426
Joined: Fri Apr 08, 2011 7:59 am
Has thanked: 52 times
Been thanked: 70 times

Got it sorted!

Postby meta-redundant » Thu Feb 06, 2014 11:44 pm

Earlier I posted how spikes and occasional clicks were definitely related to graphics for me, and how I was able to prove this by removing the AMD and Intel drivers entirely. Problem is, this is not workable in the long term because it disables just about every graphic option (including Aero).

I decided to try Fabio's advice and install a much older AMD driver - 12.1. And guess what - not only did the CPU peaks disappear, but the average load also decreased as well!! I am so happy to get to the bottom of this - it wasn't completely unworkable before, just annoying...and now it is like a new lease on life! For me the red peak indicator never lights up now, not even momentarily. Here's what I did if you have a Radeon (AMD) card of any kind and want to give it a try:

    --Download driver package 12.1. You can do this from the AMD site by choosing your hardware and OS, but then when presented with the most recent driver, instead click on "Previous Drivers and Software" in the bar on the right. I chose the oldest one, v12.1 built on 1/25/2012.

    --Completely uninstall all current AMD drivers and software. On windows, do this through Control Panel:
    http://support.amd.com/en-us/kb-article ... ivers.aspx

    --Reboot.

    --Install the package you downloaded in the first step. I did NOT install Catalyst Control Center NOR any of the video transcoding options.

    --Reboot that bad dog!

I mentioned earlier that moving the slider to disable mouse cursor enhancement also worked, and coupled with Fabio's comments about plugins "overriding the the default behavior of the pointer", I feel that this is definitely a contributing problem for many people who are getting CPU spikes (at least on Windows). Important note: I did not actually need to have any plugin windows open to get the spikes - the plugins only needed to be instantiated somewhere in the project.

THANKS FABIO for the suggestion! 8-)
Cubase 7.5 x64 | Intel Xeon E3-1275 3.50GHz (x4) w/ HT | Win7 Sp1 | RME Fireface UCX (via USB) | 32Gb Memory | UAD Quad | ATI Radeon HD 5800 | Intel RAID SSD array | iPad3 w/ Lemur MIDI control
User avatar
meta-redundant
Junior Member
 
Posts: 92
Joined: Wed Jun 29, 2011 9:20 pm
Has thanked: 3 times
Been thanked: 5 times

Re: Got it sorted!

Postby Patanjali » Fri Feb 07, 2014 1:27 pm

meta-redundant wrote:I did NOT install Catalyst Control Center

When I had AMD cards in my system, I needed CCC to correct the 15% overscan defaulted to by the driver on the three Dell ST2340Ts connected by HDMI cables. This does not happen with nVidia cards and drivers.

I had to remove the two AMD2460 quad video cards because they together introduced too many video driver 'lost video' temporary blackouts (none since putting each in separate systems). However, I am still getting 1ms DPC latencies with my three nVidia cards (though it doesn't stop me getting 1.3ms record latency at 192k), so I will try old drivers to see if I can get down to the 2-8us residual DPC latency the new system is capable of.
Patanjali

Half of the folk music duo, DevaKnighT. Music available from CD Baby (MP3/FLAC) and the usual culprits. All recorded and processed on Cubase 7.x at 192k.

Comp: i7-4930K : Asus P9X79-E WS : 32GB : GV-N750OC-2GI : SSDx4+2 : UAD-2Q : Dell S2340Tx2 & Samsung UA55HU9000W 4K.
HW: Nmn U87 Ai : JA251x2 : YRG : Korg padKontrol+uKEY2 : RME FF400+800 : M-Patch-2 : Tannoy Reveal Active+TS-12.
SW: Win 8.1 Pro 64 : Cubase 7.5.20 64 : iC Pro : RX3Adv : Ozone5Adv : Goliath/SO PP/Pianos/Gypsy : Sup.Drum 2.0 : AT7.
User avatar
Patanjali
Member
 
Posts: 669
Joined: Mon Dec 20, 2010 10:37 am
Location: Sydney, Australia
Has thanked: 5 times
Been thanked: 35 times

Re: Got it sorted!

Postby Patanjali » Wed Mar 05, 2014 9:08 am

Patanjali wrote:... I am still getting 1ms DPC latencies with my three nVidia cards ...

I just read on the DPC Latency Checker website that the utility does not read accurately under Win 8.x because it has changed the way Windows handles DPC/ISR things. Under 8.1, the checker can just show a lot of 1ms peaks if even a moderate amount of activity is going on. The reality is that the peaks are not all 1ms. They say they are intending to update the utility for 8.x.

A method of really finding out what is going on with DPCs is outlined at http://www.sysnative.com/forums/windows ... 7-8-a.html, where it also covers using the DPC Latency Checker, but also the Win 8.x options. The MS ADK provides a way of really drilling down to particular excessive DPC instances.
Patanjali

Half of the folk music duo, DevaKnighT. Music available from CD Baby (MP3/FLAC) and the usual culprits. All recorded and processed on Cubase 7.x at 192k.

Comp: i7-4930K : Asus P9X79-E WS : 32GB : GV-N750OC-2GI : SSDx4+2 : UAD-2Q : Dell S2340Tx2 & Samsung UA55HU9000W 4K.
HW: Nmn U87 Ai : JA251x2 : YRG : Korg padKontrol+uKEY2 : RME FF400+800 : M-Patch-2 : Tannoy Reveal Active+TS-12.
SW: Win 8.1 Pro 64 : Cubase 7.5.20 64 : iC Pro : RX3Adv : Ozone5Adv : Goliath/SO PP/Pianos/Gypsy : Sup.Drum 2.0 : AT7.
User avatar
Patanjali
Member
 
Posts: 669
Joined: Mon Dec 20, 2010 10:37 am
Location: Sydney, Australia
Has thanked: 5 times
Been thanked: 35 times

Re: 7.5 Bug: VST performance peak?

Postby meta-redundant » Wed Mar 05, 2014 8:47 pm

FYI, LatencyMon supports Win 8.x:

http://www.resplendence.com/latencymon

This is what I used.
Cubase 7.5 x64 | Intel Xeon E3-1275 3.50GHz (x4) w/ HT | Win7 Sp1 | RME Fireface UCX (via USB) | 32Gb Memory | UAD Quad | ATI Radeon HD 5800 | Intel RAID SSD array | iPad3 w/ Lemur MIDI control
User avatar
meta-redundant
Junior Member
 
Posts: 92
Joined: Wed Jun 29, 2011 9:20 pm
Has thanked: 3 times
Been thanked: 5 times

Re: 7.5 Bug: VST performance peak?

Postby Patanjali » Wed Mar 05, 2014 11:56 pm

meta-redundant wrote:http://www.resplendence.com/latencymon

My results were:

Old system: i7-920, GA-X58-UD4P, 12GB
Lowest interrupt to process: 10us
Highest interrupt to process: 179us
Highest ISR execution time: ~18us
Highest DPC execution time: 179us
Highest hard pagefault resolution time: 116ms

New system: i7-4930X, P9X79-E WS, 32GB
Lowest interrupt to process: 18us
Highest interrupt to process: 87us
Highest ISR execution time: ~17us
Highest DPC execution time: 155us
Highest hard pagefault resolution time: <3ms

I don't know why the older system has such a huge pagefault time, nor why either have any pagefaults as both have excess RAM for their purpose, so I don't run any pagefiles on them.


My results for the Tight Loop Latency Tests (TLLT) were:

Old system: i7-920, GA-X58-UD4P, 12GB
HIGH_LEVEL: ~3us
CLOCK_LEVEL: ~4us
DISPATCH_LEVEL: ~25us

New system: i7-4930X, P9X79-E WS, 32GB
HIGH_LEVEL:~18us
CLOCK_LEVEL: ~20us
DISPATCH_LEVEL: CPU0:~82us, others <20us

While both systems are well within acceptable DAW operational levels, it is interesting to note my older system has very good latency levels, whereas, the later has significantly higher levels, especially on CPU0 for the DISPATCH_LEVEL test. What it all really means, I don't know. It would be interesting to see what sort of systems generate TLLT values that would compromise a DAW. I must try it out on my Acer tablet.

What did you get for your results?


However, if you really want to isolate what is creating problems, the process described at http://www.sysnative.com/forums/windows ... 7-8-a.html really enables drilling down to exactly what was going on at a particular time. LatencyMon really only shows the current instantaneous level and the worst case summary.
Patanjali

Half of the folk music duo, DevaKnighT. Music available from CD Baby (MP3/FLAC) and the usual culprits. All recorded and processed on Cubase 7.x at 192k.

Comp: i7-4930K : Asus P9X79-E WS : 32GB : GV-N750OC-2GI : SSDx4+2 : UAD-2Q : Dell S2340Tx2 & Samsung UA55HU9000W 4K.
HW: Nmn U87 Ai : JA251x2 : YRG : Korg padKontrol+uKEY2 : RME FF400+800 : M-Patch-2 : Tannoy Reveal Active+TS-12.
SW: Win 8.1 Pro 64 : Cubase 7.5.20 64 : iC Pro : RX3Adv : Ozone5Adv : Goliath/SO PP/Pianos/Gypsy : Sup.Drum 2.0 : AT7.
User avatar
Patanjali
Member
 
Posts: 669
Joined: Mon Dec 20, 2010 10:37 am
Location: Sydney, Australia
Has thanked: 5 times
Been thanked: 35 times

Previous

Return to Miscellaneous

Who is online

Users browsing this forum: No registered users and 3 guests