VCV Rack, a real Modular and latency

I watched a lot of videos with combined real modular and VCV Rack systems, some even with wild in- and out-routing through the systems. I tried a lot of things to get this running but all of my efforts ended in having not much fun because of latency issues. Maybe it’s an audio interface related problem but before investing in an es-8 I wanted to ask if it is worth it regarding this issue or if there are other possibilities.
I am currently working with those little roland aira modular boxes as dc coupled interfaces (demora, bitrazer, torcido and scooper) and while I am able to run vcv rack in 96kHz with 128 (little crackles) or 256 samples buffer on an I7 lenovo laptop respectivly an ryzen 1700x desktop machine the latency is still horribly noticeble. No rhythmicly precise things are possible.
Am i missing the obvious, is it the audio interface or are there other solutions?
I have the same issues with softube modular via reaper and with reason 10. I would be so happy to mix these worlds without having to use micro-delays on the real modular site… :slight_smile:
The goal would be to have everything master-clocked through a pamelas new workout going to several real and vcv modules and have everything right in time.

Thanks in advance!

Post the audio driver and device (listed in the Core AUDIO module) you use.

I know that this is 10 months old, but I have had to abandon my Aira FX boxes for VCV due to latency issues. The Coreaudio driver is Roland’s and proprietary for these boxes. They will only operate at 96Khz sample rate. Since these are DC coupled (and cheap on the second hand market), I thought these would be a good match, but unfortunately, the latency makes them less than ideal. I don’t think that this is a fault with VCV… Roland has long abandoned these units, so I don’t expect the Coreaudio drivers to be updated anytime soon.

Roland released a new driver for the Aira FX for Catalina. I installed it. I am getting around 2ms latency with 256 samples buffer and the audio is very clean. This is fine for most of my purposes… just wanted to provide a quick update on this.