|
|
Post by aremos on Jan 6, 2020 15:15:51 GMT -6
With a 2012 Cheesegrater (12 core / 5,1 / 3.46 / 64) & PT Ultimate 2019.10 ... If you had to track directly from the converter, which of the 2 scenarios would you choose (latency?):
1) going in USB (like Dangerous AD+)
2) going in Ethernet [Dante - DVS] (like Apogee Symphony II SE 2x6)
|
|
|
|
Post by kcatthedog on Jan 6, 2020 16:25:13 GMT -6
The Dangerous is just AD but the 2x6se is AD and DA. Do you need the mastering DA of the 2x6se or the other features of the Dangerous?
|
|
|
|
Post by aremos on Jan 6, 2020 22:02:03 GMT -6
It's really a question about latency when tracking.
Without PT hardware, would it be better going into USB or Ethernet (Dante)?
|
|
|
|
Post by Blackdawg on Jan 6, 2020 23:15:37 GMT -6
It's really a question about latency when tracking. Without PT hardware, would it be better going into USB or Ethernet (Dante)? Niether really provides a clear advantage. Getting an HDX card would be the best way for latency. Though Dante could mean you could have a monitor mixer setup separate for the band since dante can go everywhere. Would just require the equipment for that. |
|
|
|
Post by Guitar on Jan 7, 2020 7:57:48 GMT -6
Every interface has different latency performance. You need to specifically learn about the exact interfaces in question.
No protocol is inherently much "faster or slower" than the other, it's more about drivers and implementation.
There is a mega thread on GS on the topic of RTL performance, but they haven't tested everything under the sun. There's a lot though.
|
|
|
|
Post by Blackdawg on Jan 7, 2020 9:21:51 GMT -6
Every interface has different latency performance. You need to specifically learn about the exact interfaces in question. No protocol is inherently much "faster or slower" than the other, it's more about drivers and implementation. There is a mega thread on GS on the topic of RTL performance, but they haven't tested everything under the sun. There's a lot though. Except HDX. And masscore Ravenna net. |
|
|
|
Post by Guitar on Jan 7, 2020 10:07:37 GMT -6
Every interface has different latency performance. You need to specifically learn about the exact interfaces in question. No protocol is inherently much "faster or slower" than the other, it's more about drivers and implementation. There is a mega thread on GS on the topic of RTL performance, but they haven't tested everything under the sun. There's a lot though. Except HDX. And masscore Ravenna net. HDX is apparently a PCIe card. Yeah, I guess PCI cards have historically been way faster than everything else. Even the old Echo's, Deltas, and whatnot. I've never heard of the second one you mentioned! EDIT: just looked it up. Apparently Ravenna is ethernet based, right into your LAN port I guess. When I mentioned protocols, I'm talking about the connection between the interface and the computer, not necessarily just brand names. For example, USB, Firewire, Thunderbolt, PCI, Ethernet, and so on. You can't just say something like "Thunderbolt is faster than USB" without talking about specific interfaces, because there are examples that defy that kind of statement. That is what I was trying to say to the OP. Although I think you are right that PCI and Ethernet seem to usually come up on the fast side of things. But so do USB and Firewire when implemented well. |
|
|
|
Post by aremos on Jan 7, 2020 10:12:05 GMT -6
Ravenna is Merging's (Horus, Hapi) ethernet protocol. Metric Halo has theirs also. And then there's Dante ...
|
|
|
|
Post by Guitar on Jan 7, 2020 10:19:41 GMT -6
Ravenna is Merging's (Horus, Hapi) ethernet protocol. Metric Halo has theirs also. And then there's Dante ... Don't forget REDNET from Focusrite! |
|
|
|
Post by Blackdawg on Jan 7, 2020 21:25:22 GMT -6
Except HDX. And masscore Ravenna net. HDX is apparently a PCIe card. Yeah, I guess PCI cards have historically been way faster than everything else. Even the old Echo's, Deltas, and whatnot. I've never heard of the second one you mentioned! EDIT: just looked it up. Apparently Ravenna is ethernet based, right into your LAN port I guess. When I mentioned protocols, I'm talking about the connection between the interface and the computer, not necessarily just brand names. For example, USB, Firewire, Thunderbolt, PCI, Ethernet, and so on. You can't just say something like "Thunderbolt is faster than USB" without talking about specific interfaces, because there are examples that defy that kind of statement. That is what I was trying to say to the OP. Although I think you are right that PCI and Ethernet seem to usually come up on the fast side of things. But so do USB and Firewire when implemented well. Even more complex than just the interface connection type. Thunderbolt 3 is 40Gbps. per buss(Not port, most computers do two ports per bus) USB 3.1 is 10Gbps(per bus) USB 3 is 5 Gbps Most ethernet adapters in almost all equipment is 1Gbps, 10Gbps is up and coming though but still expensive for most things PCI 3.0 x16 is 16Gbps PCI 4.0 x16 is 32Gbps USB, Thunderbolt, and Ethernet had PLENTY of bandwidth for ultra low latency audio. However, it is how the processing is down and the protocols. HDX uses DSP with a PCI card and with the software and PCI card nets less than a millisecond of delay. This is why tracking though PT Ultimate with HDX is so easy. Thunderbolt is plenty fast enough to do that it's just the software and in some ways the hardware won't allow it. Yet. Im sure it'll over take it at some point but not at this time. Ravenna is the same way, it is just eithernet but has ultra low latency(in a Masscore system) due to how the software treats it. So you can't just compare a connector/bus type. Its a loaded question. Which was why when he was asking what would be the lowest latency and he uses ProTool Ultimate software, he should just use HDX hardware. Dante or thunderbolt or USB won't be any different really. |
|
|
|
Post by Guitar on Jan 8, 2020 7:01:05 GMT -6
"You can't just compare a connector/bus type." etc... That is how I was using the word "protocol" in my earlier posts. The original post says something like "Going in USB or going in Ethernet" and as we know "it's not that simple."
That's what I was saying though. Apparently we are saying the same thing now. It all comes down to the exact interfaces in question.
Bandwidth is more of a factor for how many channels and how high of a sample rate you can use, not latency speed. But no, bandwidth does not equal speed. It equals "amount of data."
One thing I remember about latency performance is latency happens in little chunks and then adds up to a total number. For example if you have a 7 ms round trip latency, maybe 1 ms of that could be happening with DSP inside the interface, for example.
Presonus Quantum is so fast partly because they eliminated DSP entirely from the interface, so it is faster than most other interfaces that use software mixers, etc, because they got rid of one of the "chunks" or buffers or whatever the term is.
I am still learning the nuts and bolts of digital audio, so don't take this as argumentation, I am just trying to learn and be able to talk about it.
|
|
|
|
Post by cyrano on Jan 8, 2020 7:57:02 GMT -6
The one thing that let me understand latency was an animation of a factory supply chain, during an IBM course. Unfortunately, it's no longer on the net. The purpose of the animation was to show that speeding up the chain didn't reduce latency. Very unintuitive. It might even result in a higher latency. IBM has an AI software to analyse the problem and show where real the bottleneck is.
That's why the latency "measurements" aren't really useful. The same interface might show different numbers on a different system.
USB chipsets differ a lot. There's even one budget set on the market that doesn't work at all for audio over 4 I/O. The same goes for ethernet chipsets. There's a list of good sets for AVB and most decent computers seem to have one of these. Still, different latency with different chipsets and different computers. Firewire chipsets have been weeded out. That's why they seem more reliable to some. They also have the advantage to support DMA and that's closer to PCI than USB.
Of course, PCI still is the closest to the CPU, as it is inside the computer. Harder to design, if you keep noise in mind. Thunderbolt is the next stage, offering similar performance, but audio over ethernet is quite close if tuned right.
These days it's easy to send back an interface if it doesn't work for you. And that's what you still have to do: test with YOUR system.
|
|
|
|
Post by adamjbrass on Jan 8, 2020 8:35:09 GMT -6
Don't forget REDNET from Focusrite! That's really the best way to run Dante I/O boxes, if you have concerns about Latency. Rednet PCIe card. Works in a Cheese-grader Mac, no problem. |
|
|
|
Post by Blackdawg on Jan 8, 2020 14:44:04 GMT -6
"You can't just compare a connector/bus type." etc... That is how I was using the word "protocol" in my earlier posts. The original post says something like "Going in USB or going in Ethernet" and as we know "it's not that simple." That's what I was saying though. Apparently we are saying the same thing now. It all comes down to the exact interfaces in question. Bandwidth is more of a factor for how many channels and how high of a sample rate you can use, not latency speed. But no, bandwidth does not equal speed. It equals "amount of data." One thing I remember about latency performance is latency happens in little chunks and then adds up to a total number. For example if you have a 7 ms round trip latency, maybe 1 ms of that could be happening with DSP inside the interface, for example. Presonus Quantum is so fast partly because they eliminated DSP entirely from the interface, so it is faster than most other interfaces that use software mixers, etc, because they got rid of one of the "chunks" or buffers or whatever the term is. I am still learning the nuts and bolts of digital audio, so don't take this as argumentation, I am just trying to learn and be able to talk about it. Yeah on the same page now. I'd still say that HDX is the only one that is reliably always going to be the fastest option. That is the whole idea behind it. You eliminate a lot of the "what ifs" and random question marks by using a proprietary system/connection and make third party users conform to that. Thus it is the highest bandwidth/lowest latency setup. |
|
|
|
Post by cyrano on Jan 10, 2020 15:42:10 GMT -6
A recorder has no latency...
|
|
