Latency, 24 bit, CPU load

[arjen_1]

Hi,

I would like to know what kind of latency/audio performance is typical for OS9.
In my case I am using an ESI Waverterminal U24 usb audio interface with the Ploytec driver.
I've set it to 24bit, fast system setting, with a latency of 20ms. I can get 7ms but then the CPU load is massive while using VSTi's.

When I go down to 16 bit, cpu load drops and latency stays the same. However I like to work on 24 bit with VSTi's. Just sounds better imho.

So what kind of latency / audio performance do you guys get?

Greetz,
Arjen

[IIO]

that the bit depth doesnt change the latency is the expected result, latency is (software-side) always in samples, so only a higher samplingrate will shorten the latency.

your host is cubase, right? in opposite to some other hosts, in cubase and nuendo a shorter buffer will not have that much of an effect on CPU hunger of effects, in logic it is far worse, (not talking about maxmsp, where you dont want to work using a IO vectorsize of 32, because that means CPU x8 for effects.)

[arjen_1]

that the bit depth doesnt change the latency is the expected result, latency is (software-side) always in samples, so only a higher samplingrate will shorten the latency.

I understand. However bit depth does increase the CPU load.

your host is cubase, right?

Yep

[MacTron]

I understand. However bit depth does increase the CPU load.

Not really, as our systems are 32 bits since System 7 era (more or less) LOL. In some cases, if the software is well programed can take some speed advantage of using 16 bits (16+16=32), but it is unusual.
If your CPU load are increasing for using 24 bits, may be that implies some extra process like converting 16 bits to 24

[arjen_1]

If your CPU load are increasing for using 24 bits, may be that implies some extra process like converting 16 bits to 24

That is interesting. I tend to use a lot of VSTi's. So are Imposcar, Atmosphere, etc 16 or 24 bit? If 16 bit that could explain the extra converting, CPU load.

Would a PCI card decrease CPU load compared to USB cards? And in terms of latency...what is considered 'normal' for our machines?

[IIO]

Not really, as our systems are 32 bits since System 7 era (more or less) LOL.

in fact it should not make a difference as the plug-in interfaces and everything else inside a program like cubase is 32 bit audio, no matter hat the files are. but it might be that he sees the result of the host program using more CPU for itself when running many tracks.
it then look like a synth would need more than before.

[IIO]

That is interesting. I tend to use a lot of VSTi's. So are Imposcar, Atmosphere, etc 16 or 24 bit? If 16 bit that could explain the extra converting, CPU load.

imposcar (and some others, for example the synyths from muon or dacota) are internally 64 bits.

but their inputs and outputs are always 32 bits.

in other words, the plug-ins – as well as the cubase mixer – are doing exactly the same, no matter of you use for example cubase 4.1 and then switch to cubase 4.1/24 with 24 bit audio files.

the only thing what needs more CPU than before is reading and writing the audio material from RAM.

and the only exception, for the sake of completeness, is protools,which was using integer audio signals of 24 bit or dual 24 bits for transporting and summing, (at least until version 8, which now has a 64 bit driver for summing)

[arjen_1]

Thanks for your responses!

Someone offers me the RME 96/8 PCI card. Would this PCI card decrease CPU load compared to USB audio I am using right now?

Greetz,
Arjen


 

[Jakl]

Thanks for your responses!

Someone offers me the RME 96/8 PCI card. Would this PCI card decrease CPU load compared to USB audio I am using right now?

Greetz,
Arjen

Great card there arjen_1 used one for years a RME DigiPad - rock solid card in OS9 and OSX.
I think that it does help with CPU load not sure by how much though?
I think the drivers were better for windows with digicheck etc.
Never really had CPU overload using the card with ADAT - Analog - SPDIF at all ever.

[supernova777]

arjen i sure wouldnt pass up an opportunity to try an RME interface if the price is right!
they are supposedly some of the most rock solid interfaces
considering you are using a ESI usb with a 3rd party driver
its quite possible

[arjen_1]

considering you are using a ESI usb with a 3rd party driver
its quite possible

Not entirely true. Ploytec is (or was) the developper of all ESI drivers. The 'universal' driver I use is far better than the orginal v1 'ESI' driver. Sonically I am a big fan of my U24. Stereo image and clearness is perfect. In fact the DAC's are the same as the RME's. However I am trying to reduce CPU load to run multiple not-so-well-programmed- (a.k.a. Vanguard)  VSTi's. Therefore the RME could be a good option. I will try to get it for a decent price. Otherwise it's mixdown time.   

If I get it I will let you know my experience with the card. Thanks for your fasssst responses. 

[arjen_1]

Since the seller didn't send me a reply I tried some other things to reduce CPU load. Somewhere I read that connecting USB audio to a PCI USB card instead of the built in USB ports gives you a better bandwith with less CPU load. And...I can confirm it works! Around 20% less CPU load.  Cool trick!

[supernova777]

Since the seller didn't send me a reply I tried some other things to reduce CPU load. Somewhere I read that connecting USB audio to a PCI USB card instead of the built in USB ports gives you a better bandwith with less CPU load. And...I can confirm it works! Around 20% less CPU load.  Cool trick!

this may only be relevant to quicksilver machines... i think i remember readng that some mdd's have had their usb ports 'unlocked' to 2.0??? mactron? can u confirm this??
or am i dreaming?

also keep in mind the differences in architecture of the mdd, multiple pci bus etc. that mactron has spoken of before..

but congrats on your improved success Arjen;)

[IIO]

latency is often announced somewhere in the manual/tech sheets/advertisements, but the difference between the cards is not so extreme that this should be the only thing to look at.

the hammerfall PCI solutions are, however, one of the best OS9-solutions when it comes to conversion quality.

[arjen_1]

this may only be relevant to quicksilver machines... i think i remember readng that some mdd's have had their usb ports 'unlocked' to 2.0???

If only that could be achieved.... I am using a USB 2.0 PCI card for audio now. I still is only recognized as 1.1 so the ESI is on 1.1 too. (It's a USB 2.0 interface)

In theorie this 'trick' is relevant for all machines. The performance gain can will probably be different. I've been working for a couple of hours with Vanguard (still a very sluggish but bad ass sounding VSTi) and audio/cpu load is so much better. I can load an extra VSTi now. 

[supernova777]

honestly its probably due to the cpu load caused by the 3rd party driver
i realize they were the ones to code the driver for this ESI company from the start..
but realize that other drivers for solutions would be very much different ie: PCI, Firewire
this is probably the core problem with why so many mac users were skeptical about usb interfaces back then
usb audio interfaces were looked at as a joke for a good long time as they were introduced..
perhaps you have stumbled upon that very reason behind that old belief.
this is totally not the case for windows/macosx probably due to improved usb performance
and it may not be the usb speed itself but rather that the system is using some cpu to facilitate the usb
thats why firewire was always looked upon as better then usb on the mac  AND the pc because
usb requires cpu to regulate the function of it.. whereas firewire + pci do not require this extra cpu to manage the datathruput

http://www.diffen.com/difference/FireWire_vs_USB

Typical USB PC-hosts rarely exceed sustained transfers of 280 Mbit/s, with 240 Mbit/s being more typical. This is due to USB's reliance on the host-processor to manage low-level USB protocol, whereas FireWire delegates the same tasks to the interface hardware (requiring less or no CPU usage).

this quote is referencing a pc.. but its the same on mac, usb requires cpu in its function, firewire does not.

the improvement u are experiencing is likely because the pci card maybe has some chips on it that are regulating the usb on the card itself? im not sure.. its a guess.

[MacTron]

... i think i remember readng that some mdd's have had their usb ports 'unlocked' to 2.0??? mactron? can u confirm this?

Yes, this is true. But only on the FW800 models, and only USB 1.1 speed once booted on Mac Os 9 as any other USB 2 on Mac Os 9.

connecting USB audio to a PCI USB card instead of the built in USB ports gives you a better bandwith with less CPU load...

Yes this is true, a PCI interface (133 or 266 MB/s MAX theoretical throughput) for a USB 1.1 (1.5 MB/s MAX theoretical throughput) must be enough. LOL

Any way, USB is one of those technologies "only half performance than advertised" as DDR and many others...

[IIO]

sometimes being creative and brave can open new doors, sometimes it just causes problems and risky situations.

in my opinion an USB 2.0 PCI card has nothing to do in a computer which should boot into OS9. and an USB audio interface has nothing to do at an USB 1.1. interface.

[supernova777]

too bad a bunch of dual processor 1.0 ghz cpu's couldnt come up for sale on ebay right Arjen?
best cpu u can get for quicksilver.. if only we could fine one.. or 2. or 3;)

http://macos9lives.com/smforum/index.php?topic=578.0
jumping from 550, to 854 on the benchmark score

[arjen_1]

too bad a bunch of dual processor 1.0 ghz cpu's couldnt come up for sale on ebay right Arjen?
best cpu u can get for quicksilver.. if only we could fine one.. or 2. or 3;)

Yep! But if you know how long I searched for my 933mhz.....You don't hear me complaining. I love my machine; mint condition. In fact it's has an interesting history. Mine was used by it's previous owner for creating some of the world most reknown star maps. Now it has the joy of being used for music. 

A dual processor won't give you that much advantage on OS9 does it?

in my opinion an USB 2.0 PCI card has nothing to do in a computer which should boot into OS9. and an USB audio interface has nothing to do at an USB 1.1. interface.

Nothing wrong with backward compatability imho. At the time all USB audio interfaces had 2.0 specifications. I had an Imac G3 so no choise to use PCI. I am happy with it.