As I mentioned in my review (see PDF in my link) the HDSPe MADI FX has a special driver which saves system resources.
I/O ports are organized in groups of 8 channels, where the driver doesn't allocate resources if no channel in such a group of 8 is being used to transfer audio.
If you look at my diagram below (snapshot from Excel) from another blog article about the UFX+, there you can see the RTL (Round trip latency) times of different RME products compared to the UFX+: Input-, output latency and the sum of both which is the RTL. So the latency of the whole signal chain, e.g.: A/D conversion at analog in, time for transfer to DAW and back and D/A conversion towards connected speakers.
Lets have a look at the RTL values at 64 samples as reported by the ASIO driver to the DAW.
Note: for digital cards like HDSPe MADI FX and RayDAT you have to add the converter latency of an AD/DA converter.
You can take these values from the excellent RME manuals, which even contain these values.
In the case of the
- HDSPe MADI FX with Octamic XTC connected via MADI: additional 0,91ms for AD/DA converter latency
- RayDAT with UFX connected via ADAT in standalone mode: addtitional 0,9ms for AD/DA converter latency
Everything at 44.1 kHz sample rate, with higher sample rate (double speed) the values are even smaller.
HDSP MADI FX 3,605+0,91 (for AD/DA converter latency of XTC) = 4,5ms
RayDAT 3,742+0,9 (for AD/DA converter latency of UFX) = 4,6ms
UFX+ via TB 4,0ms
UFX+ via USB3 4,5ms
UFX via USB 5,1ms
UFX via Fw 5,6ms
As you can see
- even for the old USB driver of UFX the difference in RTL is close to the performance of PCIe (RayDAT): 5.1ms vs 4.5ms
- the new USB driver of the UFX+ is on par with the HDSPe MADI FX, both 4.5ms
- the best values you get with UFX+ via TB which is 4.0ms
But I think the main reason for this is that the UFX+ is the newer product with faster converters:
So lets compare the converter latency of Octamic XTC and UFX+ according to manual:
Octamic XTC: 0,28 + 0,63ms = 0,91ms
UFX+ : 0,28 + 0,16ms = 0,44ms which is 0,47ms less compared to the converter latency of the XTC.
There you have the around 0.5ms difference, the UFX+ is simply the newer product with faster converters.
But 0.5ms, well this is a low difference.
I would like to sum it up this way:
1. all RME driver deliver very good RTL values
2. RTL times of USB and FW drivers are on par with the performance of PCIe based products where TB (Thunderbolt) is "external PCIe".
3. times for digital transport like ADAT or MADI are not significant in the calculation
4. RayDAT and HDSPe MADI FX card would have slightly better values when connecting devices via ADAT/MADI that have more modern converters with lower D/A conversion times. So if you would choose one of the newer products of RME then you might save here maybe around 0.5ms I would guess.
Therefore I recommended you to look more at the products feature set:
- number of MADI busses
- MADI redundancy features of the HDSPe MADI FX
- driver optimizations of the HDSPe MADI FX
- whether the devices have a FX chip
- number of optical MADI channels (with the optional card you can have 3x optical MADI busses with the HDSPe MADI FX which delivers galvanic isolation
- UFX+ has DUREC and can be easily connected to devices via USB and TB and has a lot of different ports on board
X10SRi-F, E5-1650v4, Win10Pro20H2, Cub11Pro, UFX+ (v0.9735), XTC, 12Mic, ADI-2 Pro FS R BE