Ableton Forum

hi, soundcard manifacuters don't give information about latency performances. The main reason is because lacency depends from the relationship between the soundcard and the computer.
So a survey with information about souncards latency in relation to the computer wher you use it can produce an useful database for guys interested about the latency performance of a soundcad, and also about the DAW capacity of a computer. Compare data is the key to built an idea about it.

so, what is your:

COMPUTER
INTERFACE
LATENCY (specify samples, frequency, in&0ut or in+out)

- computer: macbook pro I7 2.66 (2010) 8gb ram
- interface: motu 8pre
- latency: (128 samples, 44.1khz) 7.14ms (in+out)

anto wrote:- computer: macbook pro I7 2.66 (2010) 8gb ram
- interface: motu 8pre
- latency: (128 samples, 44.1khz) 7.14ms (in+out)

I recall reading from somewhere (may have been RME forum) that firewire and usb cards introduce additional card specific buffer, and therefore the amount of latency that is shown by Live is not accurate. Can someone who knows better verify or correct this. I'd really like to know.

And an additional question to OP: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

LeifonMars wrote: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

yes

anto wrote:

LeifonMars wrote: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

yes

Huh?

LeifonMars wrote:

anto wrote:

LeifonMars wrote: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

yes

Huh?

sorry:

the lowest possible without cracks and pops

in other words the lowest usable latency with your setup, at 80%, and then verify it with a real project

LeifonMars wrote:

anto wrote:- computer: macbook pro I7 2.66 (2010) 8gb ram
- interface: motu 8pre
- latency: (128 samples, 44.1khz) 7.14ms (in+out)

I recall reading from somewhere (may have been RME forum) that firewire and usb cards introduce additional card specific buffer, and therefore the amount of latency that is shown by Live is not accurate. Can someone who knows better verify or correct this. I'd really like to know.

And an additional question to OP: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

Live's reported latency is not always correct, because there is an error discrepancy in soundcard driver reports to live. That is why live includes an adjustment value for the difference in the reported value and the actual (see "driver error compensation").

As a result many people report latency values that are far lower than the actual value for a round trip through the system. One of the very misleading drivers out there is ASIO4ALL. It is slower than almost every native ASIO driver and yet so many people report great latencies. I personally had this experience until I measured the driver error. It was significant and my actual latencies were much slower than my native ASIO driver could give.

There are several ways to measure this. One is go through Live's documentation on how to set the driver error compensation and use the value determined summed with the in and out latency values to give the overall latency. There is free software that does this as well but both methods require a physically wired loop to make the measurements.

Any measurements here are dependent on the input buffer value and the selection is completely dependent on workload. A complex set usually requires a lot more buffer value.

That being said don't expect great latencies if you are using any consumer grade products. RME is one of the best latency setups that you are likely to come across for Live use. The rest of the usual product field is quite a bit slower.

My 2 cents

^^ + my 2 cents

- computer: Desktop pc. Gigabyte UD5 P67, 2600k @ 4.5 ghz. 16 gigs of ram.
- interface: Focusrite Scarlett 8i6 (usb 2.0)
- latency: (192 samples, 96khz). 5.33ms in, 7.33ms out. (set to 2ms in driver software).

Quite amazed at the performance of the Scarlett. Well tested at these settings. Never thought a usb 2.0 card would run this well at 96khz and have such low latency. Its picky about which usb port I use though.

- computer: macbook pro i7 2Ghz 4gb ram
- interface: motu ultralite mk3
- latency: (120 samples, 44.1khz) 7.41ms (in+out)

dancerchris wrote:

LeifonMars wrote:

anto wrote:- computer: macbook pro I7 2.66 (2010) 8gb ram
- interface: motu 8pre
- latency: (128 samples, 44.1khz) 7.14ms (in+out)

I recall reading from somewhere (may have been RME forum) that firewire and usb cards introduce additional card specific buffer, and therefore the amount of latency that is shown by Live is not accurate. Can someone who knows better verify or correct this. I'd really like to know.

And an additional question to OP: do you mean one should measure here the lowest possible buffer setting one can get, or the lowest possible without cracks and pops, and if so, what kind of project should be able to run pop free under that buffer setting?

Live's reported latency is not always correct, because there is an error discrepancy in soundcard driver reports to live. That is why live includes an adjustment value for the difference in the reported value and the actual (see "driver error compensation").

As a result many people report latency values that are far lower than the actual value for a round trip through the system. One of the very misleading drivers out there is ASIO4ALL. It is slower than almost every native ASIO driver and yet so many people report great latencies. I personally had this experience until I measured the driver error. It was significant and my actual latencies were much slower than my native ASIO driver could give.

There are several ways to measure this. One is go through Live's documentation on how to set the driver error compensation and use the value determined summed with the in and out latency values to give the overall latency. There is free software that does this as well but both methods require a physically wired loop to make the measurements.

Any measurements here are dependent on the input buffer value and the selection is completely dependent on workload. A complex set usually requires a lot more buffer value.

That being said don't expect great latencies if you are using any consumer grade products. RME is one of the best latency setups that you are likely to come across for Live use. The rest of the usual product field is quite a bit slower.

My 2 cents

Yah the latency reported is rarely completely accurate. Espcially on pc with ASIO4all

On my mac the reported latency is usualy very close to real deal. Though sometimes I change sample rate and it will report bogus numbers(EG at same buffer. at 44.1 ilget 10.3 then at 96 it 4,75, then il check 88200 and il get 5.3 then back to 44,1 and il get 14,8... if change it too often in a row my whole computer whill freeze, its really wierd)

Anywhow my mac latency seems to report the right values usualy.

I have friend on pc with ASIO4all and his computer reported 56ms
but when we listen to the sound his guitar round trip it sounded like 1ike less then 10ms
barely audible latency... we can all agree that 56ms is VERY audible
What ableton reported was completly off!!

if you don't have a 'test' project then this is pointless..
you could have a 2 track setup ..
or a 64 track setup !

post a test project as a reference..
there's already a long running thread by "Tarekith" :
http://forum.ableton.com/viewtopic.php? ... mance+test
..so you can use the project from there..

I dont have to run a test project intended for Live 6 or 7 to know this works out fine, and I dont believe there would be any point of posting non-functional resulst. For reference I'm using a project with 16 channels, 2 instances of alchemy, 1 instance of Tassman4, 2 instances of audjoo helix, 2 FM8, 7 audio channels and 2 channels with drum racks. On most channels there are various effects; Multiple instances of Izotope nectar & alloy, redline reverb and several others.

Cpu peaks at around 40% +/- 5, depending on what part of the song it is, running at 96khz and 2ms set in driver (and it sounds very tight when recording, noticeably better than when 5ms is set). Like indicated I was quite surprised as my previous Esi Esp 1010 pci card started messing up at under 5 ms at even 44.1khz.

studio :
processor : Intel Q9650 2.8GHz
RAM : 8GB
win7x64
Audio :
MOTU 24i/o PCIx-424 (8 i/o utilized currently)

Launchpad
RemoteSL25
BCR2000

Live 8.2.5/m4l

template includes :
5x Drum Racks
Evolver I input
Evolver II input
4x Operators+Absynth with arps
2x Reaktor Synths
5x m4l monome emulation tracks
2x Reaktor SpaceMasterII
2xGuitaRig4
Reaktor Limiter.

it gets laid with a lot of processing real quick.

for audio buffer
I start at 128(3.2 ms i/o overall 6.4).. until I hit (+-60%)
go to 256 (6.30 ms i/o overall 12.6)...I might go to 512/1024 later when mixing
instrument buffer stays at 256..

Laptop :
(can't bother to hook it up right now)
Intel Core2Duo 1.7GHz
8GB RAM
RME MultiFaceII
winXP

all the same controllers

64 samples
instrument 128samples
mainly used as an fx unit for bass/vox (guitarig4 hq mode)
and trigering audio clips with fx
or as 1 instrument
in the studio will carry 2 instance of ACE/up to 8xAbsynth/Reaktor
with processing.

in both cases +-60% equals 100% really

dancerchris wrote:

LeifonMars wrote:

anto wrote: That being said don't expect great latencies if you are using any consumer grade products. RME is one of the best latency setups that you are likely to come across for Live use. The rest of the usual product field is quite a bit slower.
My 2 cents

Don't remember the precise figures but I tested latency with my original motu 828 and a rme fireface and got better results with the motu with the sme buffer sizes.