Hi Sam.

Yes, I have also read or heard something like that in some reviews (Youtube..).

I think this is a good decision, because you will get the much better overall package at a reasonable, fair price with excellent support. I would even consider an UFX+, because it is not that much more expensive compared to an UFX II, but it offers to you the advantage:

• to have two different connection types towards the PC, USB3 and Thunderbolt (even USB2 then you can use the interface as 30ch interface like the UFX II in cases where needed for whatever reason). Should you have any issues, you have a plan B and there is nothing better than Thunderbolt (external PCIe) when it comes to lowest possible RTL (round trip latencies between recording interface and PC)

• not having a decreasing number of ADAT channels available when using higher sample rates (double speed and higher) if you use higher sample rates you do not have the typical degradation in the number of channels over ADAT by channel multiplexing for double (and quad) speed (sample rate).

• to use MADI channels for e.g. loopback recording instead of using "precious" ADAT and AES channels for it, giving you more options / flexibility

• to expand your setup using MADI including very useful features like "MIDI over MADI" which saves MIDI cabling and supports much longer distances.

• to have no length restriction with MADI compared to ADAT (10m max for TOSLINK cable according to standard), MADI cabling can be ring structure with up to 2km long multimode fibers between each of the - usually up to - 8 devices on a MADI bus

Here one example from my blog articles / reviews, how nicely you can expand with MADI:
https://www.tonstudio-forum.de/blog/ind … XTC-DE-EN/

MADI also offers the convenient possibility to remotely control certain RME preamps via TotalMix FX using MIDI over MADI, allowing you to set, store and recall gain, phase, phantom power, etc. via TM FX. This very pleasant comfort is currently possible with the Octamic XTC and RME has promised to implement it in the 12Mic as well, but this will take some time.

My recommendation is to get a turnkey system from a company that specializes in building reliable audio workstations with components that are well matched and best suited for the intended use. This would include clarifying what would be most appropriate for your applications. AMD or Intel. More cores or higher single thread performance.
For CPU hungry VST/VSTi or in cases where people use a lot of inserts per track a higher CPU cloak / single thread performance can be an advantage. My personal gut feeling regarding this is not more than 12-16 cores and a base clock of around 3.5 GHz, but this is only a rough estimation and needs to be validated by real application tests using the DAW / applications that you intend to use.

The whole thing can get arbitrarily complicated and should better be done by a company that knows about it and has already tested many HW and driver combinations and found out a good tuning.

Important topics:

• identify the best CPU for the job

• the choice of CPU / CPU manufacturer and the number of cores also to optimize for the desired DAW, as many (most?) DAWs do not scale well over a high number of CPU cores

• to be clarified whether the AVX512 instruction set is being used by certain DAWs which mighr speed up things

• find a good motherboard / BIOS / chipset combination

• find components with drivers that cause the lowest possible DPC latencies

As Windows and MacOS are no real-time operating systems it's very important to use well written drivers, as such low level routines can't be interrupted by the process scheduler. Only the driver itself can detach itself from a CPU core. If it's not programmed well, then the core becomes utilized too long for this driver. If there is an audio related process waiting for this core to be freed, then it can happen that you get audio loss, especially at lower ASIO buffersizes.

Therefore the hardware / driver selection is very important for a DAW with near-realtime application demands.

Last recently so much new hardware and new chipsets showed up, so that its not easy to give good/proven tips.

Lerian, do you know how DPC latencies are on different AMD setups using Ryzen 3 ?

I would wait for the next generation of "full" AMD chipsets with passive cooling if possible. The B550 is an alternative to the X570, but it has some limitations compared to the X570. If I remember right the B550 itself has no PCIe 4.0 lanes (you get only those from CPU) and limited amount of USB ports in the chipset.

The only 4 boards with passive cooled X570 have astronomical prices between €361 - 899
https://geizhals.de/?cat=mbam4&xf=3 … e+L%FCfter

IMHO something is wrong with the prices / features on these consumer/gamer boards. For an excellent Supermicro Server Mainboard for Intel Xeon I payed 5y ago €280. I notice since years that these gamer products have a ridiculously high price.

EDIT: and I fully agree with you, thunderbolt should be in the chipset.

The question is whether a DAW can make use of that many cores. Currently I am sitting it out and are waiting for all the nice features to arrive (Thunderbolt 4, USB4, DDR5, ECC, 10 Gbit, ...) on one board and then I hope its not an overpriced "bling bling" product.

Not talking about CPU architecture/performance , I know that the Ryzen 3 have an excellent Single Thread performance.
Driver quality is also very important for low DPC latencies.
I know from thread of 2y ago, that there were issues in Cubase to scale with CPUs which have more than ~12 cores if I remember right
Another issue was the limitation of MMCSS in Windows 10. Microsoft deceided that only Windows 10 Server should get again an unlimited number of threads possible that can be priotized by MMCSS.
Steinberg had then contact with Microsoft under NDA .. and if you were a customer which has an Intel or AMD CPU with more than (I dont remember exactly) 12 or 14 or 16 cores, then certain (secret/under NDS) Registry settings were required for Windows 10.
Later they told they would reprogram something in Cubase to make this "hack" / finetuning under M$ NDA not required anymore.

For me this sounded like Bullshit Bingo per excellance ...

So ... maybe with Reaper you have the luck that it scales well.
Logic (Apple) - as far as I heard - puts some higher load on the 1st CPU thread and this can limit performance. Not sure whether this is still the case.
Cubase .. tbh .. didn't track this since months ..

This is what I mean and what I am personally kind of worried about.

Therefore I personally will contact a company which has (hopefully) more concrete knowledge in that area, to get an advanced system where all components perfectly fit together.

Cool thanks, will have a look later.

That article was written in feb. 2020, windows didn't had written the code for all the cores to be used - as also stated in the article - you need windows pro for workstations to be able to use more than 64 threads, which the tester didn't used. Reaper devs are the fastest in the space to address these issues, but as long as the OS can't access the cores properly, reaper can't either.

https://www.anandtech.com/show/15483/am … x-review/3

About the DPC latency, that was a problem in zen2 cpus, since zen3 that is no longer a problem due to different cpu architecture.

Wow, i did not expected so much information. Thank you guys. I already extracted some really interesting stuff from your advices, it helped me understand some things, and other i need to figure out.
You are both right, in my quest of low latency, i can't skip the thunderbolt option. So, yes i will choose a soundcard with thunderbolt, yes i will keep up to date and maybe wait a bit to see the next steps in technology (thunderbolt 4, ddr5...)
And your links Lerian, reminded me that i already new the brand "pcaudiolabs", and indeed Ramses i will ask them advice and quote, that the best thing to do.
Thanks again to both of you, i'll educate myself with the links you posted!

I come to other conclusions, Lerian, look:
Example Calculations for RTL (Round Trip Latency)
- for a single recording interface:
  RTL = ASIO input latency + output latency
- for PCIe based interface + preamp through ADAT:
  RTL = ASIO input latency + output latency + converter latency for A/D and D/A
Please note: according to RME the latency for the routing through recording interface or ADAT is not significant in this calculation.

Case 1: UFX+ (Thunderbolt):  RTL = 1,297 + 1,315 = 2,612 ms
Case 2: UFX+ (USB3):             RTL = 1,043 + 2,018 = 3,061 ms
Case 3: RayDAT--ADAT--XTC:   RTL = 0,771 + 1,519 + 0,28 + 0,63 = 3,2ms
Case 4: RayDAT--ADAT--UC:    RTL = 0,771 + 1,519 + 0,98 + 0,63 = 3,9ms

Result: as you can see the RTL with an UFX+ - no matter whether using Thunderbolt or USB3 - is lower compared to a PCIe based recording interface and interface or preamp connected via ADAT.

Additionally you have the advanced features and possibilities of the flagship interface UFX+ and the possibility to expand through MADI:  USB2/3 / TB, FX Chip, Advanced DURec with RTC, Autoset, MADI, MIDI over MADI, WC, quality of the analog section incl. the preamps and instrument inputs. The bottom line is that this is the better unit with increased flexibility, availability and lower latency.

Another advantage of the UFX+ is, that the number of analog ports is constant, no matter which sample rate you use.
With an ADAT based solution you can only route 8 channels of the UC through ADAT.
And with double (and quad) you loose many channels by port multiplexing (8 -> 4 -> 2).

With the UFX+ you have a constant number of excellent preamps and analog ports (overhauled analog section compared to UFX+) and can use the digital ports like ADAT1, ADAT2, AES for other special purposes and still expand the installation by making use of MADI. Or you use free MADI ports for special routing tasks in the area of loopback recording.

The clear winner is here the RME UFX+, cost and feature wise
UFX+                                                              = €2198
RayDAT + WCM + UC = €585 + €109 + €845 = €1539

With the UFX+ you can use all of its ports at any sample rate. With the combination of RayDAT and UC you can only use the 8 channels routed through ADAT. And as the UC has only 1 ADAT port, thats more or less all (ok plus coaxial SPDIF, but this is not that much more)

USB3 have no improvement over USB2 regarding latency. The only thing it can bo better is more bandwidth. If you only do 1 or two guitar recordings plus 2 mics at the time and don't need the extra i/o then you can get away with the babyface pro fs. I can have quite big projects with 512 samples on a 5yo computer (i5 6600k). On recording it can go as low as 2.99 ms at 96khz (3.15 at 48khz), i don't think you can get better than this.

https://www.youtube.com/watch?v=KlkTpZL4kcg

Just tested 24/48 at 64samples i have 3.9ms RTL with the older babyface pro (not fs). No drops or glitches with a comp, eq, limiter and soothe2 on the master. The daw i use is reaper. Even UFX+ on TB, according to ramses, have 4ms at 64samples - probably tested on a similar computer as mine (2015).

https://www.tonstudio-forum.de/blog/ind … 8-RME-UFX/