I'm a Creamware user since the fantastic TripleBoard 2 (I use it as realtime spectrum analyzer) and I have a PC with Luna II. I use this with Nuendo 2 but recently, updating my pc system, I have noted this strange (?) situation :
Using LunaII (44.1 Khz + 25ms latency), my "Nuendo project X" has 60% CPU usage. The same Nuendo project, opened using the mobo Integrated audio has 35% CPU usage!
Using the integrated audio I should double (!!) the number of the effects in my Nuendo project. Clearly I don't want use the integrated audio but someone can explane me this situation considering that my default Luna project consists in a simple routing config? (STM 1632 + Analog I/O + ASIO 24 I/O + 24 bit WAVE I/O)
Thanx,
Antonio
This is my setup :
Win XP pro SP1
Nuendo 2.2
Luna II (PCI slot 4 - IRQ 7 - not shared)
ASUS A8V Delux
Athlon 64 3500+
2 Gb DDR 400 ram
3 HD Ultra2 SCSI (Controller SCSI : Adaptec AHA-2940U2W)
1 HD IDE Maxtor 120 Gb
1 HD SATA Raptor 74 Gb
NVIDIA GeForce4 MX 4000
Luna 2, Nuendo and CPU usage
First off, what version of the Scope software are you using (and what drivers?) Also, what latency are you using with the software ASIO layers you've tried? Hopefully you're comparing 25ms to 25ms...
Second, you seem to be using a mixture of parts that I suspect isn't giving you top performance:
Now the Adaptec 2940U2W SCSI controller has me a bit concerned. With only 3 dsps its likely you're not pushing enough PCI traffic yet to cause pci overflows, but I'm willing to bet you have a lot of latency on your PCI bus nonetheless. Turning your SCSI drives down below 40mb/s and/or playing with the PCI Latency timer of each individual device in your system (AGP cards etc) may help yield more performance.
While PCI latency is typically only visible as poor PCI performance (PCI overflows with a dsp card) too much latency on your peripheral devices can eat up cpu time dramatically as the cpu essentially sits idle that long waiting for each device to return its state or do any i/o needed (to main system ram).
Beyond any of this it might be worthwhile to share what exactly you have going on in that Nuendo project....
Second, you seem to be using a mixture of parts that I suspect isn't giving you top performance:
The ASUS A8V Deluxe uses the VIA K8T800Pro chipset and the VT8237 (southbridge) peripheral controller. Luckily that xbit article on the K8T800 and this article on the KT600 (which the VT8237 initially debuted on) both seem to indicate that VIA is at least not sharing the PCI slots with the SATA/PATA IDE ports. Notice I'm not focusing directly on the VIA chipset itself here, although its common knowledge that VIA hasn't been the best option for heavy PCI traffic in the past.On 2005-03-03 05:48, logicube wrote:
This is my setup :
Win XP pro SP1
Nuendo 2.2
Luna II (PCI slot 4 - IRQ 7 - not shared)
ASUS A8V Delux
Athlon 64 3500+
2 Gb DDR 400 ram
3 HD Ultra2 SCSI (Controller SCSI : Adaptec AHA-2940U2W)
1 HD IDE Maxtor 120 Gb
1 HD SATA Raptor 74 Gb
NVIDIA GeForce4 MX 4000
Now the Adaptec 2940U2W SCSI controller has me a bit concerned. With only 3 dsps its likely you're not pushing enough PCI traffic yet to cause pci overflows, but I'm willing to bet you have a lot of latency on your PCI bus nonetheless. Turning your SCSI drives down below 40mb/s and/or playing with the PCI Latency timer of each individual device in your system (AGP cards etc) may help yield more performance.
While PCI latency is typically only visible as poor PCI performance (PCI overflows with a dsp card) too much latency on your peripheral devices can eat up cpu time dramatically as the cpu essentially sits idle that long waiting for each device to return its state or do any i/o needed (to main system ram).
Beyond any of this it might be worthwhile to share what exactly you have going on in that Nuendo project....
Really thanks for your answer.
I tested SFP 3.1c and SFP 4 and the results are the same. I compared the my "favourite" LUNA latency (25ms), that guarantees the best performances in terms of CPU usage (number of vst/dx effects) with the mobo audio that has 185ms latency. This was my error and this is the cause of this performance dissension. I tryed turn my SCSI drives down below 40mb/s and I set my graphic aperture size to 32Mb, but the result is the same
---------------
My latency test :
Scope setting : 44.1 Khz / 25 ms latency
Stop/Play Mode = Performance VST : CPU 60 % / Disk = 0 %
Scope setting : 44.1 Khz / 13 ms latency
Stop/Play Mode = Performance VST : CPU 64 % / Disk = 0 %
Scope setting : 44.1 Khz / 7 ms latency
Stop/Play Mode = Performance VST : CPU 68 % / Disk = 0 %
Scope setting : 44.1 Khz / 4 ms latency
Stop/play Mode = Performance VST : CPU 77 % / Disk = 0 %
Scope setting : 44.1 Khz / 3 ms latency
Stop Mode = Performance VST : CPU 95 % / Disk = 0 %
Play Mode = Performance VST : CPU 100 % / Disk = 50/100 % (Nuendo doesn't start)
Do you think this values are excessive or you think they are regular?
-------------
Clearly the cause of this differences between the "two" compared boards is the latency but now I have a doubt : How many is useful a dsp card in a mixdown project considering that I don't use the DSP effects?
In my mixdown projects I use VST/DX effects only and during this phase my priority is the number of simultaneus effects I can use. I don't like multiple audio bounces to help the CPU. Please revise me if I'm mistaken but at this point how is really important a short latency in a mixdown project? Is possible that I'm mistaken, in effect I never used other PCI soundcard in my life in exception of TripleBoard 2 and Luna (I have a RME hammerfall DSP multiface on my laptop)...pls let me know
P.S.
I have a ASUS A8V deluxe, can you suggest me the best mobo with 939 socket??
I tested SFP 3.1c and SFP 4 and the results are the same. I compared the my "favourite" LUNA latency (25ms), that guarantees the best performances in terms of CPU usage (number of vst/dx effects) with the mobo audio that has 185ms latency. This was my error and this is the cause of this performance dissension. I tryed turn my SCSI drives down below 40mb/s and I set my graphic aperture size to 32Mb, but the result is the same
---------------
My latency test :
Scope setting : 44.1 Khz / 25 ms latency
Stop/Play Mode = Performance VST : CPU 60 % / Disk = 0 %
Scope setting : 44.1 Khz / 13 ms latency
Stop/Play Mode = Performance VST : CPU 64 % / Disk = 0 %
Scope setting : 44.1 Khz / 7 ms latency
Stop/Play Mode = Performance VST : CPU 68 % / Disk = 0 %
Scope setting : 44.1 Khz / 4 ms latency
Stop/play Mode = Performance VST : CPU 77 % / Disk = 0 %
Scope setting : 44.1 Khz / 3 ms latency
Stop Mode = Performance VST : CPU 95 % / Disk = 0 %
Play Mode = Performance VST : CPU 100 % / Disk = 50/100 % (Nuendo doesn't start)
Do you think this values are excessive or you think they are regular?
-------------
Clearly the cause of this differences between the "two" compared boards is the latency but now I have a doubt : How many is useful a dsp card in a mixdown project considering that I don't use the DSP effects?
In my mixdown projects I use VST/DX effects only and during this phase my priority is the number of simultaneus effects I can use. I don't like multiple audio bounces to help the CPU. Please revise me if I'm mistaken but at this point how is really important a short latency in a mixdown project? Is possible that I'm mistaken, in effect I never used other PCI soundcard in my life in exception of TripleBoard 2 and Luna (I have a RME hammerfall DSP multiface on my laptop)...pls let me know
P.S.
I have a ASUS A8V deluxe, can you suggest me the best mobo with 939 socket??
The CPU usage is way too high. 
It's draining somewhere eh.
Are you using only stereo ASIO on Luna? Extra ASIO channels add to the CPU usage. But still...
Are you using ASIO Scope drivers, and not directx or so?
It's draining somewhere eh.Are you using only stereo ASIO on Luna? Extra ASIO channels add to the CPU usage. But still...
Are you using ASIO Scope drivers, and not directx or so?
more has been done with less
https://soundcloud.com/at0m-studio
https://soundcloud.com/at0m-studio
I wouldn't axe the scsi (or your luna) yet, but the 'integrated soundcard' latency of 185ms does explain the discrepancy in cpu load. I'm still curious what you have going on in your Nuendo project? I'm assuming its a fairly loaded project and its not empty or just a softsynth or two.
And yes, as you lower the latency of your soundcard the cpu usage can rise dramatically, but 23ms or so should normally be ok for doing mixing (many people will switch latencies for recording and then make it higher for mixing when you use more effects/cpu load).
Oh, how many ASIO tracks are you using in & out?
<font size=-1>[ This Message was edited by: valis on 2005-03-04 06:53 ]</font>
And yes, as you lower the latency of your soundcard the cpu usage can rise dramatically, but 23ms or so should normally be ok for doing mixing (many people will switch latencies for recording and then make it higher for mixing when you use more effects/cpu load).
Oh, how many ASIO tracks are you using in & out?
<font size=-1>[ This Message was edited by: valis on 2005-03-04 06:53 ]</font>
My Nuendo project consists in :
10 Stereo channels + 4 mono channels + 4 stereo group tracks + 1 FX track + Stereo Bus
Audio tracks 24 bit
9 Nomad Factory compressors
7 Nomad Factory EQ
6 Steinberg Qmetric EQ
6 Waves Rverb
2 Waves Enigma
3 Nomad factory Gate/Limiter
2 Waves Denoiser
1 Big Tick Exaline
1 Nomad Factory Brickwall
1 iZotope Ozone 3
2 PSP mixtreble
2 Voice designer
1 Nomad Factory Chorus
3 Limiters (Waves L2)
2 ASIO24 IN + 2 ASIO24 OUT
10 Stereo channels + 4 mono channels + 4 stereo group tracks + 1 FX track + Stereo Bus
Audio tracks 24 bit
9 Nomad Factory compressors
7 Nomad Factory EQ
6 Steinberg Qmetric EQ
6 Waves Rverb
2 Waves Enigma
3 Nomad factory Gate/Limiter
2 Waves Denoiser
1 Big Tick Exaline
1 Nomad Factory Brickwall
1 iZotope Ozone 3
2 PSP mixtreble
2 Voice designer
1 Nomad Factory Chorus
3 Limiters (Waves L2)
2 ASIO24 IN + 2 ASIO24 OUT
I agree, with those kind of plugin loads your performance definately seems normal.
Steinberg's Qmetric eq is oversampled interally (2x) and with 7 bands it packs a bit of a cpu hit, something I'd use sparingly on 1 or perhaps 2 parts that really need it.
Waves denoiser uses a complex array of filters to suppress the noise print you feed it (dynamically even)...not something I typically run realtime.
6 Waves Rverb is a bit heavy too...suggest setting up some fx busses to use as 'sends' into your LunaII so u can take advantage of 1 or 2 Scope verbs there. (6 verbs might be a bit heavy too, but there are others here that insist its necessary too so I'll digress on that point).
Waves L2 is *definately* something I'd not run in a project if I could help it (and I can). Not only is it consuming a lot more cpu than you need to for individual channel dynamics, personally I'd rather turn my monitors up and leave the really extreme dynamics squashing until my mix was finished and I was mastering the final stereo mix.
iZotope Ozone 3 is something I'd again recommend using offline. There's so many 'mastering quality' algorithms packed into that thing that it easily eats 20-30% cpu on a p4 ~2Ghz under normal usage.
Perhaps you can optimize your workflow a bit? Don't be afraid to commit changes to a bounce or 'freeze' as you can easily create a workflow so that the old 'originals' are lying around in the event you do something that u can't undo when bouncing...
Steinberg's Qmetric eq is oversampled interally (2x) and with 7 bands it packs a bit of a cpu hit, something I'd use sparingly on 1 or perhaps 2 parts that really need it.
Waves denoiser uses a complex array of filters to suppress the noise print you feed it (dynamically even)...not something I typically run realtime.
6 Waves Rverb is a bit heavy too...suggest setting up some fx busses to use as 'sends' into your LunaII so u can take advantage of 1 or 2 Scope verbs there. (6 verbs might be a bit heavy too, but there are others here that insist its necessary too so I'll digress on that point).
Waves L2 is *definately* something I'd not run in a project if I could help it (and I can). Not only is it consuming a lot more cpu than you need to for individual channel dynamics, personally I'd rather turn my monitors up and leave the really extreme dynamics squashing until my mix was finished and I was mastering the final stereo mix.
iZotope Ozone 3 is something I'd again recommend using offline. There's so many 'mastering quality' algorithms packed into that thing that it easily eats 20-30% cpu on a p4 ~2Ghz under normal usage.
Perhaps you can optimize your workflow a bit? Don't be afraid to commit changes to a bounce or 'freeze' as you can easily create a workflow so that the old 'originals' are lying around in the event you do something that u can't undo when bouncing...
Thanks for your suggestions even if it is not a mixing forum 
Anyway, I use the limiters to limit the punching sounds and not to punch them, it's a "subtle" difference. I use Waves L2 for a chirurgical limiting and Nomad Factory limiter for its tube simulation, it depends by the kind of sound I'm treating. I think that Waves L2 has a very very low CPU usage and I think that to obtain a good "analog mixdown sensation" the best way is encrease the CPU usage to => apply all the needed effects in one single project, without multiple bouncing..I'm in according with you for all the other things
My first problem was the difference in the CPU usage between the Luna and the mobo audio and I undestood that the real problem was the latency. After this problem (solved), my new problem is : during the mixdown phase how is useful a DSP board if I don't use the FXs integrated in the DSP board?
Antonio
Anyway, I use the limiters to limit the punching sounds and not to punch them, it's a "subtle" difference. I use Waves L2 for a chirurgical limiting and Nomad Factory limiter for its tube simulation, it depends by the kind of sound I'm treating. I think that Waves L2 has a very very low CPU usage and I think that to obtain a good "analog mixdown sensation" the best way is encrease the CPU usage to => apply all the needed effects in one single project, without multiple bouncing..I'm in according with you for all the other thingsMy first problem was the difference in the CPU usage between the Luna and the mobo audio and I undestood that the real problem was the latency. After this problem (solved), my new problem is : during the mixdown phase how is useful a DSP board if I don't use the FXs integrated in the DSP board?
Antonio
I know creamware since 1998 and I know all its products, XTC included. I know also other DSP boards as powercore or UAD-1, but this is not the question. The question was clear in my previous post..If I decide to use/buy professional software bundles for my effect collection how is useful a DSP board...etc etc
Clearly I'm speaking about the mixdown phase, not realtime processing or other DSP board exclusive features
Antonio
Clearly I'm speaking about the mixdown phase, not realtime processing or other DSP board exclusive features
Antonio
If you don't use the DSP effects, the DSP is useless (other than reliability, low latency, and routing, which might be useful if you track live instruments) as you could simply get a straight simple audio interface instead. I'm not sure why you would want to ask that question over here, as it strictly depends on the use you make of it. No one is going to hurt you if you decide to sell your DSP board to get something that fits better your needs/workflow.
I posted my second question as conseguence of the first "problem" and I posted over here cause I'm a Creamware user/fan and cause this forum is open and very active.
As I mentioned I use every day Luna for its routing, for my recording and for its fantastic Modular 2 modules that can bring out unequalled drum sounds. I use also a Tripleboard (with a dedicated PC) for its real-time spectum analyzer that I use in warp mode and for its denoiser (what is Waves denoiser?). I have never used other audio cards (in exception of my RME hammerfall dsp multiface that I use on my laptop for my lives and that is on PCMCI bus) so you can imagine my reaction when I see that a stupid modo audio had half CPU usage..
Anyway I think this tread should be stopped here, I will ask you some info about Creamware Bundles and XTC with another post..
Thanx
Antonio
As I mentioned I use every day Luna for its routing, for my recording and for its fantastic Modular 2 modules that can bring out unequalled drum sounds. I use also a Tripleboard (with a dedicated PC) for its real-time spectum analyzer that I use in warp mode and for its denoiser (what is Waves denoiser?). I have never used other audio cards (in exception of my RME hammerfall dsp multiface that I use on my laptop for my lives and that is on PCMCI bus) so you can imagine my reaction when I see that a stupid modo audio had half CPU usage..
Anyway I think this tread should be stopped here, I will ask you some info about Creamware Bundles and XTC with another post..
Thanx
Antonio
Hey man, really thank you for your suggestion! I will use the STM 1632 mixer and some MasterVerb to mix the Nuendo group channels..I think I have to frequent often this forum..On 2005-03-05 13:09, symbiote wrote:
On the other hand, several people have reported (and I've gotten to the same conclusion with my own tests =P) that mixing down with SFP's mixers sounds better than using Logic's/Cubase's mixing engine. This is something you need to try for yourself tho, see which one you prefer.
Antonio