You are not logged in. Please login or register.
RME User Forum → AVB/Dante - Network based devices → Digiface vs AOX
Question: If I only need 32 channels madi <> Dante, Will the Digiface then perform the same as the AoX? I’m contemplating to swap my Windows Workstation to a Mac Mini M4
Will latency be the same? Or will the AoX perform better, even when only using 32 channels?
What kind of performance are you referring to?
Latency round trip, cpu overhead etc. If PCie performa better as usb3 that the same amount of channels that could influence my buying decission.
I can only speak for my Windows 10 setup with a system/server mainboard (Supermicro X10SRi-F) from 2014 that has been upgraded regularly over the years. It now runs an 8-core CPU (Intel Xeon E5-1680 v4 for €160 instead of €1750 from eBay), 10-Gbit LAN (Intel X710-DA2), 64 GB RAM, and an nVidia RTX 4070.
I have been using USB 2, USB 3, and PCIe-based RME products for over 10 years: UFX, RayDAT and UFX as preamp, UFX+, UFX III, and now the HDSPe MADI FX. All products deliver excellent performance and stability.
Based on what I can observe via Windows gadgets visualizing the CPU load of all 16 cores (main and hyperthread), I cannot see any increased CPU utilization with USB-based products compared to RayDAT or HDSPe MADI FX.
Regarding RTL: with RME’s ASIO drivers, I do not see significant differences in RTL; the differences are small.
An interesting aspect of the HDSPe FX driver in my current setup (with HDSPe MADI FX and UFX III now for DURec backup recording) is that it works in a very resource-optimized way.
More information on these topics can be found in my blog:
1. https://www.tonstudio-forum.de/blog/ent … tup-en-de/
Here you’ll also find an RTL comparison between the UFX III analog ports and the HDSPe MADI FX with the M-1620 Pro via MADI.
Also information about the resource optimizing HDSPe FX driver.
2. https://www.tonstudio-forum.de/blog/ent … cts-en-de/
This article lists the different RTLs of various RME products and product combinations, showing that with RME’s ASIO drivers, the difference between USB and PCIe is small.
Regarding USB 3 and system load: I remember that someone claimed a while ago that USB-based interfaces would be worse than PCIe-based ones, causing higher CPU load and less stability under load, and that it would be impossible to work stably with a CPU utilization of more than 70%.
I think this is a wrong claim, maybe only true on poor systems with high DPC latencies, where it is clear upfront that they would fail with small buffer sizes and on heavy load due to blocking drivers and other issues (https://forum.rme-audio.de/viewtopic.php?pid=208109).
I ran a quick check and came to wholly different results; see here in my answer:
https://forum.rme-audio.de/viewtopic.ph … 92#p210092
The "artificial" load test was playback of an artificial DAW load project with 400 tracks, 803 (!) VSTs, over 26 GB of memory utilization, and a CPU-Z CPU stress test running in parallel, causing an instant CPU load of 100% (!!!).
You will agree that you would never have such a high load and system penetration under normal work conditions.
Result: no audio loss at an ASIO buffer size of 64–128 samples.
So, under more typical working conditions (not an artificial monster project plus a running stress test), it should be clear that RME drivers are simply excellent, and you do not need to worry whether you are using a USB 3 or a PCIe-based interface.
The prerequisite, of course, is that the computer system itself is solid and has no performance issues such as high DPC latencies caused by bad drivers, poor USB3 controllers, or other issues.
Regarding Apple, a lot is changing right now. It may be difficult to predict which approach will work better in the long run.
And it might also depend on the Apple model that you buy.
Thank you Ramses, I currently run a MadiFX (rev1) with an M32 AD Pro (AVB)
My system is a z690 based 14900k system with 32gb ddr5 7000
This experience latency wise is fine (currently WL12 and Cubase 15) but with my SSL nucleus 2, Burl Bomber ADC and DAC all being Dante, I need intergration at driver level. Currently the hardware signal path connection is there, but no software control via one console. The Mac Mini is an option as my experience with M4 is very good so far. But I also have an Intel Nuc with 13700 and 64gb mem laying around with would make a nice match with the madiface Dante. I like the small footprint and getting rid of the big “midi” tower of my current workstation. So decission is a mixture between performance, which you explained, flexibility (external USB 3 can go on any platform) and price AoX still needs the madi daughterboard, so that’s a 2k€ total price to 1.289€ for the Madiface Dante. In any scenario my MadiFX will go up for sale.
I've personally had a much better experience with PCIe vs USB 3.0 on both Windows & macOS.
Many others are having issues with the Digiface Dante via USB 3.0 on macOS, though the Class Compliant driver/mode seems to be a good solution. Even then, I personally experience minor glitches or irregularities in CC mode that are not present on my HDSPe MADI FX in a OWC Mercury Helios 3S Thunderbolt enclosure. The Helios is significantly more quiet than Sonnet's enclosure BTW.
I can achieve half the round trip latency via my MADI FX than I can with a Digiface Dante in CC mode, and with higher reliability. Everyone's experiences will be different, but that has been mine. It's worth noting that the RTL on the Digiface Dante in CC mode is still quite good (6ms at 64 samples on my M3 Pro).
One last thing to mention, the Digiface Dante can operate as a router even when not connected to a host, the AoX must be actively connected to a host. This could be make or break for a setup.
Personally I will be moving from the Digiface Dante to the AoX at some point in the future.
RME User Forum → AVB/Dante - Network based devices → Digiface vs AOX
Powered by PunBB, supported by Informer Technologies, Inc.