Important pieces of information are missing in the report.

1. Was RME ASIO driver in use and what buffer size was in use
2. Was ASIO4ALL configured with a comparable buffer size in “samples” or in a different unit like “ms”?
3. Has the same sample rate been in use?
4. How did you measure the latency?

Keep in mind that the latency values in a DAW are being shown from what is compiled in into the ASIO driver of a vendor.
And this is only valid for local analog ports where the converter latency is known, not for ports on external devices behind ADAT.

(side note: RME calculates the values properly, so their driver usually report very precise. In other cases you need to plug a loop and measure with external tools like https://oblique-audio.com/rtl-utility.php).

ASIO4ALL can't do this for obvious reasons. It doesn't know the hardware/converters, has unlike a real ASIO driver no direct access to the audio HW and doesn't know the latency through the Windows audio subsystem.
Depending on which Windows driver ASIO4ALL bases his work, it can use any Windows audio driver and some Windows drivers (all except WASAPI exclusive mode are not even lossless).

So, sorry, but your report is much too vague and leaves many questions open.

Thanks - yes, I'm definitely using the RME Babyface drivers for the Babyface.  And latency is based on what I measured using the RTL utility along with what's shown in Cubase 12.  The values are pretty close in both, accounting for the fact that Cubase separates input/output and the RTL utility combines both.

The latency on the BF is at 1024 buffer (~24 ms output) and the latency on the Realtek is at 512 buffer (~21 ms output).  I get occasional audio dropouts on both at those values but it's OK.  If I drop either to lower latency I get a lot more audio dropouts with the project I'm using as a benchmark.  The 21 ms value for the Realtek is consistent with what I've experienced with onboard audio on a number of other laptops, so that value makes sense.

Regarding the i9 desktop, I also have an i9 desktop with a Fireface 802 and it runs the same project at ~3 ms latency.  Laptops always have higher latency (in my experience) so the difference is expected.

Regarding feeling the latency, I've found that 20 ms is my threshold where I start to notice it.  So they're both workable.  I've been doing ~20 ms with onboard audio on laptops for years and find that it's workable.

So they're both "OK".  The part that confuses me here is that the BF isn't providing lower latency than the Realtek on the same laptop.

If anyone has any other thoughts, please let me know.  I'll keep digging through some configuration options and see what I can come up with.

Thanks,

rgames

> The RME driver will work more efficient compared to the Windows Realtek driver and ASIO4all.
I'm not sure what metric you're using to define efficiency but the measured latency for the Realtek/ASIO4all setup is about the same as the Babyface.  I'm not concerned about efficiency, I'm concerned about minimum latency. If one driver is less efficient but provides lower latency then I'll use that one.

I agree the stated buffer sizes don't matter.  That's why I measured the latency.  And there's basically no difference.  If I try to go below 20-25 ms with either setup I get a lot of dropouts.

That's the part that's confusing.  I assumed I would get lower latency with the Babyface.  I do not.

rgames

Yes, understood.  That is consistent with what I measured and what shows up in Cubase.

To be clear, I cannot run that project at 512 samples on the BF.  I can run at 512 samples on the Realtek but, as you pointed out, there are other buffers involved so the real buffer size is unknown but certainly larger than 512.  But it doesn't matter because the buffer size is not the metric I care about.  That's why my original post mentions the latencies, not the buffer sizes.

Anyway, given that RME has a reputation for lower latency than onboard audio, if anyone has any ideas on why I can't get it then please let me know.

Thanks,

rgames