Ableton Forum

I went through the audio setup & driver error compensation lessons in ableton live 8.1.1. The values came out to

Buffer size: 270 samples
Input latency: 9.16ms
Output latency: 9.16ms
Driver error compensation: 1ms
Overall latency: 19.3ms

That seems high from what I've read around the net, for example this article which says "I was able to get latency settings in Ableton Live of 2ms or 32 samples with no pops or clicks at a sample rate of 44.1kHz"

Pretty sure my hardware is fine. It's a MBP running OS X 10.6.2, 2.53 GHz Intel Core 2 Duo, with 8 GB of 1067MHz DDR3 ram. What's up? Should I be able to get better latency with the firebox + this machine, or is 20ms the best I can do?

i am on OSX 10.6.2 (core audio driver) and use mainly 256 sometimes 128 samples and get i think around 11ms latency with the 256 samples setting @ 48khz, sounds like something is not right. (how old is the card? did you run the latest firmware update?)

11ms each way or total?

pergesu wrote:11ms each way or total?

total

I play with 20ms, doing lots of live instruments and looping, it don't bother me. Do you hear anything wrong? If not, what's the problem ?

chapelier fou wrote:I play with 20ms, doing lots of live instruments and looping, it don't bother me. Do you hear anything wrong? If not, what's the problem ?

also why don't you use the latency compensation?

As far as I know, this is useless when monitoring the input.

I am using a firepod but have had similar results. I get the same basic levels when switching drivers to asio4all. I can get a single track to be okay at much lower ratings but when I adjust for heavy loading/multi tracks I get the following settings:

44100 hz
176 samples (buffer)
5.62 ms (input)
9.50 ms (output)
87 samples (error)
17.1 ms (overall)

Note that many are using 96 khz or higher to get low latencies. Subfunk for example has a larger buffer size than I do and yet his overall time is about half. I surmize he is using 96k. However the drawback is using twice the resource requirements. Running with 96k vs 44.1 can saturate your bandwidth with large recordigs. Sometimes more memory can alleviate this problem.

I imagine that some others are reporting only input latency, however some have reported amazingly low round trip values with PCI cards rather then external boxes like the PreSonus stuff.

I get about 10-11ms on my Firebox.

I play E-Drum in live with 20ms total latency and it's really good!

My motu is set to 192 samples @ 48k which is about 9 ms total. I can actually get around 16 total on
my xstation. Even on a nice FW or card imterface, it's always a trade off, higher sample rates can get lower latency at a cost of cpu. That's why I always think those cpu threads are daft, they omit too many details.

dancerchris wrote:I am using a firepod but have had similar results. I get the same basic levels when switching drivers to asio4all. I can get a single track to be okay at much lower ratings but when I adjust for heavy loading/multi tracks I get the following settings:

44100 hz
176 samples (buffer)
5.62 ms (input)
9.50 ms (output)
87 samples (error)
17.1 ms (overall)

Note that many are using 96 khz or higher to get low latencies. Subfunk for example has a larger buffer size than I do and yet his overall time is about half. I surmize he is using 96k. However the drawback is using twice the resource requirements. Running with 96k vs 44.1 can saturate your bandwidth with large recordigs. Sometimes more memory can alleviate this problem.

I imagine that some others are reporting only input latency, however some have reported amazingly low round trip values with PCI cards rather then external boxes like the PreSonus stuff.

i am on 48khz

p.s. firebox, firmware update, done or not done?... don't know if that can be the issue, just asking.

SubFunk wrote:
i am on 48khz

I just did the calcs and with 176/41000 samples in buffer you should get a latency (just buffer) of slightly under 4 ms. The rest of that is round trip junk that the driver adds for internal hardware latency. Asio4all lets you handle all the sizes a little better but the presonus driver has a CPU setting and an input latency value that seems independent of the buffer value. All this adds up to the values that I obtained. Maybe I should retry running AISO4all?

SubFunk: Are you using a PreSonus product? MAC? Firewire? Are you using the PreSonus Driver?

i use a presonus firebox, firewire (well it is only available as a firewire device) on a MBP 2.16ghz, 4gb ram, OSX 10.6.2, core audio driver. @ 48khz.
running mainly Live8, Logic8 and bias peak6. in my home studio setup since about 3-4 years.

Some things to consider

1) the macworld article was from 2005 when Macbook/Macbook Pro were using the Texas Instruments firewire chipset which is vaslty superior to the Agere chipset used in the past 2 years. If you look at Presonus's website, they only endorse TI and VIA chipsets but most manufacturers have at least tried to work with the Agere chipset given that it's only one hardware to write for the entire mac book pro range.

2) Ableton Live's audio engine needs more buffering and processing than something say, Logic or Cubase because of it's on the fly, drop and drag without cutting out coolness. I have Cubase 5 and can run it 1-2 buffer settings lower than Live, so if I can run Live at 256 samples comfortably, I could run Cubase 5 at 128 or 192 samples comfortably. I can also run Renoise with tons of tracks (if not using VSTs, or at least not freezing them) at 128 samples. I can get even better results using the internal audio card with asio4all. However, for most purposes, 20 ms is a manageable latency - perhaps not if you're tracking live drums, but most other things should be manageable.

3) The guy in the article had a Mac Pro g5 dual core tower, not a laptop.

4) Various other factors can affect performance - background programs running, power consumption, graphics card drivers, wifi, etc, etc.