The difference between UFX and RayDAT is not that significant.

But I could think of, that the communication over PCIe Bus is in general more robust under load.
But this also depends on your PCs design / chipset.

The minimum ASIO Buffer Sizes which you can set differ:
UFX USB/FW Driver: 48
RayDAT: 32

Here the different Input / Output Latencies in Cubase Pro 8.5, UFX using USB (Firewire not much different):
32:  RayDAT:  0,771ms  /  1,519ms
        UFX:           n/a             n/a
48:  RayDAT:     n/a             n/a
        UFX:        1,701ms  / 2,630ms
64:  RayDAT:  1,497ms  / 2,245ms
        UFX:        2,063ms  / 2,993ms
96:  RayDAT:     n/a             n/a
        UFX:        2,789ms  / 3,719ms
128: RayDAT:  2,948ms  / 3,696ms
        UFX:        3,515ms  / 4,444ms
256: RayDAT:  5,850ms  / 6,599ms
        UFX:        6,417ms  / 7,347ms
512: RayDAT: 11,655ms / 12,404ms
        UFX:       12,222ms / 13,152ms

Thanks Matthias for your remarks.

I hope I understood you right telling, that the RayDAT Latency values from Cubase are worse compared to the pure digital performance which this card should have, as there comes some add-on delay from AD/DA conversion of the UFX ...

To exclude any influence of the UFX I could remove all cables between UFX and RayDAT, but sorry, too much work.

What I did now is to remove all busses in this Cubase project (Inputs, Outputs, Monitors).
So when selecting the RayDAT ASIO driver I would expect now, that Cubase simply measures ("somehow", how exactly I don't know) the pure latency between the Application and the RayDAT card using its ASIO driver.

Result: the values shown in Cubase are exactly the same for the RayDAT ASIO driver, no matter whether I have ADAT channels defined in the project in the Input and Output busses or not.

Or what you mean exactly ? Sorry either we misunderstand each other a little or I am missing background how Cubase measures.

Do you have a proposal to better compare measure both interfaces ? I had the idea, that the Cubase values are "just right" to simply display the performance (in terms of Latency) from application perspective.

Thanks Matthias, understood now what you mean.

I found in the UFX Handbook the information that A/D and D/A requires 2ms.
Is it right to assume that this includes the mentioned 3 samples ?

Then I think I need to add 2ms to RayDAT Total (Input/Output) Latency to compare it to UFX values, correct ?

So when I look at the new calculation, see below, then UFX and the "combination of UFX and RayDAT" are really close.

Only when using an ASIO buffer of 32 with the RayDAT the latency is slightly lower.

TBH, I hoped for more benefits "latency wise" when buying the RayDAT and adding it to the setup as I hoped that the latency via PCIe is significantly lower compared to the way over Chipset and USB/Firewire.

Can you tell me whether the  RayDAT has some further advantages which I didnt see now ?

One think we discussed already, that its maybe more robust under load, as the communication over PCIe is better compared to use the route via Chipset and USB/Fw driver.

Is maybe the CPU consumption lower ?

Wondering whether it was really worth getting the RayDAT.

It still makes sense to have by this a backup for the UFX, as my monitors support AES.
And buying 2 UFX would have been significanly more expensive.

Latency-wise times have changed and USB is even faster than PCIe (or Thunderbolt). But the difference is a few samples with no meaning for real-world usage.

Definitely.

For the same reason. A PCIe based system will feel more 'snappy' and have less CPU load which will help when maxing out the CPU on bigger projects.