SAC SYSTEM BUILD QUESTIONS

[John Ludlow]

That's true. Still - a terabyte of storage is sufficient for over 15 hours of 32 track recording at 96/24 - twice that at 48/24. And I think that most motherboards that do m.2 have capacity for 2 of them. That would be 60, 32 track, hours at 48/24. That feels like a lot to me - but maybe it isn't if you do it professionally. And, as previously pointed out, it is total overkill from a throughput standpoint if there is any advantage in doing it more conventionally.

Oh - one more potential con: the 970 EVOPlus uses NVMe technology which is new-ish. As a result, even though it fits in the standard m.2 slot, it may not work at all with some older motherboards that have m.2 slots - or, if it boots, it might cause crashes. In other cases you can re-flash the ROM on the MB so that it knows about NVMe (I had to do this...). It was a hassle to run down, and flashing ROM is still nerve-wracking for me, even after all this time. But NVMe is quite a bit faster than the previous generation of m.2's - so it was still worth it in the long run. You won't have this issue on MB that were created after NVMe became a standard.

But one more potential pro too: it turns out that Samsung makes a 2TB m.2 version of the same drive. So, using both m.2 slots, that would be a total of 4TB installed - or over 120 hours of 32 stereo track recording at 48/24, in case that seems as if it's more than you'd ever need. The price break is not as good on the 2TB though - more than double the 1TB.

[John Ludlow]

I couldn't get the card to work for anything. The motherboard has the correct header connector and the bios controls, but the card simply wouldn't initialize.

Thunderbolt support wasn't explicitly listed as supported on this particular motherboard, but I figured it was worth a try since it had the external card header.

About Interfaces - I recently made a switch from the Berhinger ADA8200s to MOTU 8Ms. In addition to ADAT they also support AVB, USB and Thunderbolt.

I wanted to do a comparison between the RME rayDAT card and Thunderbolt connectivity and see what sort of latency differences there might be.

From MOTU's specs I should be able to reduce the Input-to-Output latency by about 2ms using Thunderbolt. Current roundtrip latency with the existing RME/ADAT configuration is 6.5ms

Gotcha. You seem to do really careful work, Phillip (which I admire). Please keep us appraised of your result.

In my case, I would really like both the convenience of being able to use the converters with more than one PC (like a laptop for situations in which higher track count wasn't an issue) and also the freedom to not have an expansion card using up "u's" in my rack. Plus, Sweetwater stocks it and having a month to try before returning it if I don't like it takes some anxiety out of multi-thousand dollar decisions. But, I'm also very wary of data transport standards that weren't made for music - particularly Apple ones that may have been 'sorta-ported' to PC.

That said, the UFX+ does USB 3 too. That throughput is so much bigger than the amount of data that has to be transferred for my usage that it should be frictionless. I can't imagine why Thunderbolt would lower latency compared to it. But...

[cgrafx]

Gotcha. You seem to do really careful work, Phillip (which I admire). Please keep us appraised of your result.

In my case, I would really like both the convenience of being able to use the converters with more than one PC (like a laptop for situations in which higher track count wasn't an issue) and also the freedom to not have an expansion card using up "u's" in my rack. Plus, Sweetwater stocks it and having a month to try before returning it if I don't like it takes some anxiety out of multi-thousand dollar decisions. But, I'm also very wary of data transport standards that weren't made for music - particularly Apple ones that may have been 'sorta-ported' to PC.

That said, the UFX+ does USB 3 too. That throughput is so much bigger than the amount of data that has to be transferred for my usage that it should be frictionless. I can't imagine why Thunderbolt would lower latency compared to it. But...

Motu specs their Thunderbolt interface at .5ms faster than the USB interface and it can handle 64 channels of audio vs 32 for USB.

I don't believe you will have any issues with USB.

The biggest difference I would expect running Thunderbolt vs USB is potential reliability from a connection standpoint. Thunderbolt is effectively PCIe and should behave a lot like a physical interface card. USB is a transiant interface that expects things to be connected and disconnected randomly, so can be affected by both the order that devices are connected as well as what devices are connected.

If you plug a usb 1.1 device into the same USB channel that your interface is connected to, it will drag the speed of the entire chain down to the speed of the 1.1 connection. USB also seems to sometimes have difficultly remembering what devices are assigned to ports.

So with USB you have to pay more attention to how/when you plug in your interface.

In my case I'm really building purpose built configurations. They don't ever get disconnected and they never get used for anything other than what they were configured for.

[John Ludlow]

In my case I'm really building purpose built configurations. They don't ever get disconnected and they never get used for anything other than what they were configured for.

That is mostly my intent also. I would plug a thumb drive into it occasionally for data transfer elsewhere and VST/VSTi additions. And I do run a Fishman Tripleplay (guitar-to-MIDI converter) sometimes which uses USB to serve a receiver. But, in general, it's just for audio. I presume that I can arrange for the USB 3 circuit that connects to the UFX+ to be dedicated to it - and the Fishman, and one-off copies, could use a different one.

But this is just the sort of vague fiddliness that I'm trying to avoid in what I intend to be a rock-solid build. It sure sounds as if Thunderbolt is the way to go vs. USB 3. But, on a PC, if you're not particular in your manufacturers compatibility, Thunderbolt is apparently fiddly too. I have a request into Sweetwater to check with their RME rep to recommend a PC-based, iron-clad, Thunderbolt hardware solution for the UFX+. I guess I just need to wait and see what he comes up with (and if he can...). If it doesn't work out then I can always go the hard-wired expansion card route - and I would expect for that to be solid. But it's not my first choice and it would also cost me a lot more to go that way (at least as I currently have it figured). AND - I would lose the benefit of Sweetwater's 30 day refund deal on their stocked items, which is a major reason to shop with them rather than somewhere else.

The preferred plan is currently a UFX+ with my old Fireface 800 and a Grace Designs m108 hanging off it via ADAT - giving me variously configurable line-level/instrument/mic preamp/MIDI connections. The UFX+ would give me MADI connectivity for expansion later if 26/16 isn't sufficient anymore. And I'd fill a 500 series lunchbox with more exotic preamps for special situations. The whole thing would roll up in a couple of, smaller, Gator-esque airport-suitcase-style, impact-plastic-covered, racks with the two skateboard wheels and the extendable handle - plus a monitor, keyboard, and mouse. And, when it wasn't traveling (the usual...) it would live in my home studio.

The alternative is probably an HDSP MADI card with MADI-compatible converters and dedicated pres. I feel pretty certain the second solution would work flawlessly (presuming SAC and SAW do...). But, it's not as configurable, and it isn't no-questions-asked returnable - and so I don't prefer it. And - it would be a lot more expensive. So I'm rooting for Sweetwater to figure flawless PC Thunderbolt out.

[John Ludlow]

Another potential pitfall of USB 3 with the RME Fireface UFX+ - USB 3 connections made from a mb header to an outside access point using a cable.

This is verbatim from the RME site for it, but I put the relevant bit in red:

<<<
Notes on USB 3.0 Operation

The special features and possible problems of using USB 3 are described in more detail in the manual.Current Compatibility Information:

• Fully compatible to Intel's USB 3 implementation, which - on current Windows and Mac computers - is part of the chipset. USB 3 sockets that are connected via an internal cable (not directly soldered onto the motherboard) can cause transmission errors. These are shown in the Settings dialog.
• Fully compatible to AMD's USB 3 implementation. USB 3 sockets that are connected via an internal cable (not directly soldered onto the motherboard) can cause transmission errors. These are shown in the Settings dialog.
• Fully compatible to Fresco USB 3. This rarely used chip can be found for example on the Sonnet Tango 3.0 PCIe card.
• Compatible to NEC / Renesas USB 3. Real-world performance and error-free operation depend on the firmware version, driver version and the PCB layout of the respective extension card/motherboard.

• ASMedia - not compatible.
• Etron EJ168A - not compatible.
• Texas TUSB7340 - not compatible.
• Via VL800/805 - not compatible.

>>>

They don't even mention USB 3 achieved via expansion cards...

Edit: Whoops! It was late last night - actually it does. The approved card is now highlighted in blue.

I wonder whether Thunderbolt from a wire attached to a header, rather than directly from a soldered connection on the motherboard, could also be your issue, Phillip?

[John Ludlow]

That's true. Still - a terrabyte of storage is sufficient for over 15 hours of 32 track recording at 96/24 - twice that at 48/24. And I think that most motherboards that do m.2 have capacity for 2 of them. That would be 60, 32 track, hours at 48/24. That feels like a lot to me - but maybe it isn't if you do it professionally. And, as previously pointed out, it is total overkill from a throughput standpoint if there is any advantage in doing it more conventionally.

I've thought of a reason not to use m.2 drives if you are also relying on connectivity to your audio hardware via Thunderbolt. Both m.2 and Thunderbolt make use of PCIe lanes. They may ultimately compete with each other for controller time when you're tracking as data comes from the converters through Thunderbolt and is then stored on the m.2 chip. Basically, the data stream has to pass through both systems virtually simultaneously while they both make use of PCIe - thus possibly competing. Moving storage to a different subsystem could resolve that and allow both to move as fast as they want. And, again, even SATA is total overkill, as far as the speed needed to record and playback, so nothing is really functionally lost by moving storage to a slower technology.

[cgrafx]

I've thought of a reason not to use m.2 drives if you are also relying on connectivity to your audio hardware via Thunderbolt. Both m.2 and Thunderbolt make use of PCIe lanes. They may ultimately compete with each other for controller time when you're tracking as data comes from the converters through Thunderbolt and is then stored on the m.2 chip. Basically, the data stream has to pass through both systems virtually simultaneously while they both make use of PCIe - thus possibly competing. Moving storage to a different subsystem could resolve that and allow both to move as fast as they want. And, again, even SATA is total overkill, as far as the speed needed to record and playback, so nothing is really functionally lost by moving storage to a slower technology.

It will all depend on the particular chipset being used and how many PCI lanes are available. If I'm not mistaken the SATA buss still uses the PCI buss to move data, so your still moving data over the same functional subsystem.

I'm still with you on using separate storage for your project files. Use the M2 drive where it will provide the most performance which in this case will be for the Operating system and applications.

You don't need the performance of the M2 drives for storing Audio, and its just generally a good practice to separate your project files from the OS and Apps.

[John Ludlow]

My Sweetwater salesman has convinced me that Thunderbolt is as reliable as PCIe by referring me to the RME owners manual for the UFX+, which is pretty bold. I quote one piece of it as relayed to me by Sweetwater:


<<<
Unlike USB 2 and 3, Thunderbolt is under a very strict quality control and certification process. This rigid approach guarantees highest compatibility and functionality. Connecting a UFX+ to any Thunderbolt port will just work - period.
>>>


That's confidence at a level I've never heard anyone use in relation to USB or Firewire. So... I've decided to pull the motherboard I bought specifically for this upgrade, and return the Firewire card that I could never get to work with Windows 10-64 LTSC. Instead, I'm buying a MB that has Thunderbolt 3 soldered into it and I'll step it down to Thunderbolt 2 using the device that RME recommends (Star-Tech TBT3TBTADAP). Odd that you can't plug a TB 2 device into a TB 3 socket. Oh, wait - Apple.

[mr_es335]

John;

Oh, wait - Apple.

... Maybe it does not matter, but

THUNDERBOLT
Thunderbolt is the brand name of a hardware interface developed by Intel (in collaboration with Apple) that allows the connection of external peripherals to a computer. Thunderbolt 1 and 2 use the same connector as Mini DisplayPort (MDP), whereas Thunderbolt 3 re-uses the USB-C connector from USB. It was initially developed and marketed under the name Light Peak, and first sold as part of a consumer product on 24 February 2011.
Wikipedia

PS: So , Intel owns Thunderbolt - if I understand things correctly.

[John Ludlow]

PS: So , Intel owns Thunderbolt - if I understand things correctly.

Oh - good sleuthing Dell. I guess it's back to a cloying mystery then... .