7950x in hand

[Jim Roseberry]

1 hour ago, Teegarden said:

@Jim Roseberry Is it still worth while slightly overclocking RAM with the latest Ryzen series? I read an article a few years back that you could gain up to 25% performance without risking instability on Ryzen systems and even our Microsoft forum member Pete Brown advices to think about overclocking memory: Tip: Memory performance can make a real difference, especially if you work with large sample libraries. Slightly overclocking your memory is an often overlooked source of performance.

And one question that kept sticking to me: 

• "effective latency" vs "total round trip latency cakewalk" - which matters when checking latency and why
  (the explanations I found so far are not clear to me)

When using the word "overclocking", I want to clarify:

I'd recommend running RAM at (or near) it's maximum listed clock-speed.  This assumes the motherboard supports the higher RAM clock-speed.

I would absolutely not recommend trying to achieve RAM clock-speed above what the RAM is rated.  That's almost guaranteed to cause instability.

 

If you're in Cakewalk's preference settings:

"Effective Latency" shows that it's listing the audio interface driver's Input latency.

This can get confusing... I'll try to explain:

 

When working with a DAW, you're dealing with two types of latency.

1. Playback (or one-way) - when playing virtual instruments
2. Round-trip (or two-way) - when playing/monitoring in realtime thru software based effects/processing

 

Round-trip latency is the sum of the following:

• ASIO input buffer
• ASIO output buffer
• The driver's (often hidden) "safety" or "streaming" buffer 
• Latency of the A/D and D/A converters

Let's say you have your audio interfaces set to 44.1k with a 64-sample ASIO buffer size.

• ASIO input buffer = 1.5ms
• ASIO output buffer = 1.5ms

So you're already at 3ms... without counting the latency of the driver's safety/streaming buffer and latency of the A/D D/A converters.

Total round-trip latency is going to be 4-5ms (or much higher if the interface uses a large safety/streaming buffer size).

 

To complicated matters, some interfaces don't report their latency accurately.

Ideally, you want to use a loop-back utility like RTL Utility... to measure the actual round-trip latency.

If an audio interface reports its ASIO input/output latency accurately, Cakewalk will list it accurately under ASIO Reported Latency.

 

If an audio interface driver doesn't report its ASIO input/output latency accurately, recorded audio maybe slightly offset (out of time).

In this case, you can take a short transient and loop-back record it to a second track... and zoom-in and see if there's any difference in time (between them).

If so, you can take the time difference between them (in samples), and enter it in Manual Offset - which is under Record Latency Adjustment (samples).

Hope that helps clarify!

 

Edited November 16, 2022 by Jim Roseberry

[Teegarden]

Thanks, it's almost clear to me🤩: 

Well, my RME AIO reports 2,7 ms effective latency at 96kHz, 256 samples in the CbB mixing latency tab. I always thought RME is quite reliable and accurate.

However, below that ASIO reported latencies:

• Input: 3,1 ms (296 samples) 
• Output: 3,7 (353 samples)
• Total Roundtrip: 6,8 ms (649 samples)

The number that is important for playing virtual instruments and recording vocals without noticeable delay is 3,1 (which is related more ore less equal to/the same as "effective latency"?) ?
I use an external reverb, so I don't need CbB to add that to a vocal when recording. I do that because I thought that way I could avoid the round trip latency, which is more than double and probably getting noticeable at 6,8 ms.

[Jim Roseberry]

2 hours ago, Teegarden said:

The number that is important for playing virtual instruments and recording vocals without noticeable delay is 3,1 (which is related more ore less equal to/the same as "effective latency"?) ?

It's actually the Output that you're most concerned with when playing a virtual-instrument (3.7ms in your case).

2 hours ago, Teegarden said:

I use an external reverb, so I don't need CbB to add that to a vocal when recording. I do that because I thought that way I could avoid the round trip latency, which is more than double and probably getting noticeable at 6,8 ms.

You can actually combine monitoring via hardware (dry vocal) and monitoring via software (just for the reverb).

• Setup a track with your preferred vocal reverb... and set the plugin to 100% processed/wet signal
• Enable input monitoring on this "Vocal Reverb" track.  You can adjust the level of the reverb by adjusting this track's level.
• Monitor the dry vocal directly off the audio interface (or hardware mixer)

This way, the dry signal is near zero latency... and you get to hear your choice of reverb.

The 6.7ms round-trip latency won't be distracting on the 100% wet reverb signal.  It'll sound like 6.7ms of Pre-Delay.

 

You can use this some process if you have vintage keyboards (that don't have onboard effects).

It's nice to be able to monitor with chorus, delay, reverb... while still being able to record the dry signal.

 

[Teegarden]

Thank you very much! I always highly appreciate your knowledge, info and help.

You are one of the reasons this forum is so valuable!