Old 03-25-2019, 04:09 PM   #1
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default Thunderbolt, vs Thunderbolt cards, vs PCIe ?

I'll be upgrading both my main PC and my audio system soon.

I'll be using a Lynx Aurora(n), but I'm a bit puzzled by current connection options. I know that Dante is ridiculously expensive (due to licensing issues?) plus it currently has latency issues. So my two choices are Thunderbolt, or staying with PCIe

With audio, I see that PCIe cards are pretty much going the way of the Flintstones, with Thunderbolt-equipped mobos starting to be standard. therefore, I'll probably go that way, but I'd REALLY like to understand more. I've read countless articles, but I have yet to find one that fills me with confidence.
I trust some of the computer gurus here more than on most other forums.

SO, MY QUESTIONS:
----------------------------------------

1: Does Thunderbolt have the potential to offer lower DAW latency than even a PCIe card?

I read something, rather badly explained, about how TB uses DMA so that the audio goes straight to & from your ram, thus creating no cpu overhead. That's sounds tasty to me, but is that also the case with PCIe cards, or not?


2: If using a PCIe Thunderbolt adapter card, with my current non-TB motherboard, should I expect LESSER performance, vs a mobo with built-in TB-3?


3: Do all TB3-equipped motherboards offer the exact same TB performance?

I ask because I read something (again, very badly explained) about how some TB implementations use some kind of controller chip, whereas the "real thing" has a more direct path to the promised land.

---------------------------------

### LAST: Do any gains that might happen exist because the system will allow for a smaller hardware buffer, or are those gains created elsewhere? (With the buffer size still determined by cpu speed, driver design, etc)

Last edited by Cableaddict; 03-25-2019 at 05:43 PM.
Cableaddict is offline   Reply With Quote
Old 03-25-2019, 11:22 PM   #2
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

I understand that Thunderbold is more or less "PCI in a cable".

OTOH a PCI can have multiple "lanes" (AFAIK, either of 1 2 or 4) working in parallel, so "PIC" by itself is not a very relevant description.

In the German "Sound & Recording" magazine about a year ago, I read a very enthusiastic report on a Windows PC that comes pre-insatalled with the OS tweaked for DAW use and with an especially selected Thunderbold card.

I suppose unless you are very experienced with "embedded" PC tweaking, you should spend some additional $$ for a PC coming from a company with decent experience on "audio" PCs.

-Michael
__________________
www.boa-sorte.de

Last edited by mschnell; 03-25-2019 at 11:28 PM.
mschnell is online now   Reply With Quote
Old 03-26-2019, 10:44 AM   #3
toleolu
Human being with feelings
 
Join Date: Apr 2014
Location: Sunshine and palm trees.
Posts: 556
Default

Quote:
Originally Posted by Cableaddict View Post
2: If using a PCIe Thunderbolt adapter card, with my current non-TB motherboard, should I expect LESSER performance, vs a mobo with built-in TB-3?
That depends on whether the built in TB on the motherboard has it's own dedicated microprocessor, or if it uses the CPU.

It used to be that expansion cards were always a bit better performance wise because they contained their own microprocessors, whereas embedded devices on the motherboards used the CPU for processing thereby adding some additional overhead to the CPU.

I don't know if that's changed at all with the newer motherboards, but hopefully there's something in the specs that might indicate one way or the other. The other thing is whether the motherboard manufacturers are still putting expansion card slots on their motherboards. If the manufacturers are doing away with slots in favor of embedded controllers, then the decision has pretty much been made for you.

One of the biggest advantages to using expansion cards was is if the card died, you just replace the card, if the embedded device dies, you have to replace the motherboard. Given the choice, I would rather plug a new card in a slot rather than replacing the entire motherboard.

Last edited by toleolu; 03-26-2019 at 10:54 AM.
toleolu is offline   Reply With Quote
Old 03-26-2019, 11:04 AM   #4
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by toleolu View Post
That depends on whether the built in TB on the motherboard has it's own dedicated microprocessor, or if it uses the CPU.

It used to be that expansion cards were always a bit better performance wise because they contained their own microprocessors, whereas embedded devices on the motherboards used the CPU for processing thereby adding some additional overhead to the CPU. .


Toleolu, YES, that's exactly what I was looking for, thanks.

Again, I know that theoretically, TB uses DMA (direct memory access) so there's no cpu overload. But I also know that this s not always implemented, hence my question. I wanted to ask the mobo manufacturers about this, but didn't know the correct words until your post.

Followup: Again, if the mobo DOES have a microprocessor, using DMA, is this theoretically less cpu overhead than using a PCIe card, or do ALL good PCIe cards also use DMA?

(Hmm, now I have a good question for Lynx as well.)
Cableaddict is offline   Reply With Quote
Old 03-26-2019, 11:07 AM   #5
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by mschnell View Post

In the German "Sound & Recording" magazine about a year ago, I read a very enthusiastic report on a Windows PC that comes pre-insatalled with the OS tweaked for DAW use and with an especially selected Thunderbold card.
Thanks, Michael.
I'm very interested in in that "especially selected Thunderbolt card" since I'll probably have to update my Aurora first, using my current PC. I guess maybe it had the "best" controller chip.

Well, this is yet more specific stuff I can ask the manufacturers.
Cableaddict is offline   Reply With Quote
Old 03-26-2019, 12:15 PM   #6
toleolu
Human being with feelings
 
Join Date: Apr 2014
Location: Sunshine and palm trees.
Posts: 556
Default

CPU overhead, in this regard, isn't about DMA. Look at it this way.

Any controller, TB, network, audio, video, etc. that's installed in a computer has to run a program of some type in order to do what it does. And, like any program, a processor of some type is required to execute the code associated with that program. If the controller has it's own dedicated microprocessor executing that code, then the CPU doesn't pay any attention to that controller until it's needed for something. On the other hand, if that controller does not have a dedicated microprocessor, then that code has to be executed somewhere else, in most cases, that's the CPU. That's where the overhead comes in because in addition to everything else it's doing, the CPU has to spend cycles executing that code.

Don't get too hung up on hardware specs, they can be very misleading. Hardware these days is pretty rock solid, the main thing is just making sure that everything is compatible and plays nice with each other. Your focus should be on what you are trying to accomplish with the computer and what all you'll need in order to get there. That usually comes down to software, and the software will generally determine the hardware specs you need. Oh, and most importantly, keep it simple. These things are complex enough without adding layers of complexity for things that you may, or may not, need.

Good Luck.
toleolu is offline   Reply With Quote
Old 03-26-2019, 05:21 PM   #7
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

No, these things are actually CRITICAL for what I'm trying to do.

My Windows rigs are used for live-music mixers (and Keyboard VSTi's, and amp simulators, and..... )

Massively complex setups that also require the absolute lowest latency possible. I'm currently at 5.8 ms, true i-o (not including any midi) and while that's "serviceable," it's really NOT fast enough to truly feel like analog response.

There is no easy solution, and all sorts of things factor-in. There are even several areas / parameters that affect latency. It's not just what HW buffer you set. - But even with THAT, there are several factors that affect how low the buffer can go.

Sadly, none of the standard online benchmarks ever apply, and almost no one has a true handle on how to maximize all areas. (Though good strides have been made in this area, in the last year or so.)

It's a never-ending battle.

----------------------------

Oh, and - (I'm not aiming this at YOU, of course)
Anyone who thinks super-low latency doesn't matter is a lousy musician, or an engineer with putty in his ears. End of argument.
Rant over!

Last edited by Cableaddict; 03-26-2019 at 05:47 PM.
Cableaddict is offline   Reply With Quote
Old 03-26-2019, 10:42 PM   #8
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

5.8 mSec "true i-o " is equivalent to 2 m free field distance from the loudspeaker.

To me this seems like decently "analogue".

-Michael
__________________
www.boa-sorte.de
mschnell is online now   Reply With Quote
Old 03-26-2019, 11:04 PM   #9
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by mschnell View Post
5.8 mSec "true i-o " is equivalent to 2 m free field distance from the loudspeaker.

To me this seems like decently "analogue".

-Michael
5.8 ms is workable, but not good. At 8 ms, you find yourself pushing a little with vocals, which sounds bad & makes you tire quickly. You will also be playing your instrument too hard, sometimes even (subconsciously) playing ahead of the pocket, to compensate. This has been documented many times, by many good musicians & engineers. This is really not subject to discussion, even though the vast majority of "musicians" don't get it.

The same would happen if, with analog, your floor monitor is 8' away from your ears.

IMO, 6 ms, is JUST BARELY workable, but far from ideal.

If monitoring "through the DAW" while recording, the goal should be 5 ms or less, IMO.
Cableaddict is offline   Reply With Quote
Old 03-27-2019, 12:12 AM   #10
toleolu
Human being with feelings
 
Join Date: Apr 2014
Location: Sunshine and palm trees.
Posts: 556
Default

FYI, I'm not a musician, just a hobbyist who spent 45 years in IT managing systems infinately more complex than anything you're involved with. Doesn't matter if you're recording a band or managing thousands of users, computers all pretty much work the same way.

Good luck.
toleolu is offline   Reply With Quote
Old 03-27-2019, 08:39 AM   #11
Sweetfoot
Human being with feelings
 
Join Date: Feb 2017
Posts: 6
Default

Quote:
Originally Posted by mschnell View Post
5.8 mSec "true i-o " is equivalent to 2 m free field distance from the loudspeaker.

To me this seems like decently "analogue".

-Michael
If your monitor is already 2m away, that extra 5ms means you now have 10.8 ms of latency, which is pretty noticeable
Sweetfoot is offline   Reply With Quote
Old 03-27-2019, 08:42 AM   #12
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

Quote:
Originally Posted by Sweetfoot View Post
If your monitor is already 2m away, that extra 5ms means you now have 10.8 ms of latency, which is pretty noticeable
Yep. but most of the latency would be seen with "analogue" gear as well.

-Michael
__________________
www.boa-sorte.de
mschnell is online now   Reply With Quote
Old 03-28-2019, 02:36 PM   #13
poetnprophet
Human being with feelings
 
poetnprophet's Avatar
 
Join Date: Jan 2018
Posts: 664
Default

What's your hardware buffer? Just curious, i track with a 64mb buffer and my latency round trip real shit is 3.1 ms.

Oh, and this is with USB 2.0.
__________________
https://www.kdubbproductions.com/
https://www.youtube.com/channel/UCpC...2dGA3qUWBKrXQQ
i7 8700k,4.9Ghz,Win10,Reaper 5,Motu 828es,MJE Hulk 990,GAP Pre73/EQ81
poetnprophet is offline   Reply With Quote
Old 03-28-2019, 02:43 PM   #14
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Once you get down into the sub 5 ms area, it becomes a very strange area of investigation.

For instance:

With drums, the pocket is mostly created with the hats & ride cymbal, so how far away are the OH mics? Maybe 3-4 ms worth?

With digital piano: A real grand piano has it's own "latency, since the hammer strikes the string quite a bit later than the key bottoms out. So, at what point would a fast pc computer rig actually be TOO fast?

What makes some live bands play in the pocket and not others, besides basic talent? Could it be how far each player is from each other?

All these things really DO matter, just like moving a recorded snare backwards/ forwards by 2 ms can make a huge difference in the feel of the track.

The bottom line is, we should strive for the absolute lowest latency, even if we then choose to dial some back in, for certain situations. Also, again, for vocals, you will always want to get as close to zero as possible.
Cableaddict is offline   Reply With Quote
Old 03-28-2019, 02:52 PM   #15
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by poetnprophet View Post
What's your hardware buffer? Just curious, i track with a 64mb buffer and my latency round trip real shit is 3.1 ms.

Oh, and this is with USB 2.0.
Are you saying you get 3.1ms, in-out, with a 64 buffer @ 44.1K, via USB 2.0?

Forgive me if I don't believe you! (Not even with RME drivers.) Unless maybe you're running just a couple of tracks, with minimal plugins. Even then .....
-----------------

How exactly are you measuring that? You have to record to a second computer, there's no other absolutely accurate way:

Send I mic direct into computer B, and one mic routed through Reaper, running on computer A, into a second track on computer B.
Record in computer B, line the 2 mics up, and snap your fingers.
Then zoom way in on the 2 recorded tracks of computer B, and measure the actual time difference.
Cableaddict is offline   Reply With Quote
Old 03-28-2019, 03:24 PM   #16
poetnprophet
Human being with feelings
 
poetnprophet's Avatar
 
Join Date: Jan 2018
Posts: 664
Default

Quote:
Originally Posted by Cableaddict View Post
Are you saying you get 3.1ms, in-out, with a 64 buffer @ 44.1K, via USB 2.0?

Forgive me if I don't believe you! (Not even with RME drivers.) Unless maybe you're running just a couple of tracks, with minimal plugins. Even then .....
-----------------

How exactly are you measuring that? You have to record to a second computer, there's no other absolutely accurate way:

Send I mic direct into computer B, and one mic routed through Reaper, running on computer A, into a second track on computer B.
Record in computer B, line the 2 mics up, and snap your fingers.
Then zoom way in on the 2 recorded tracks of computer B, and measure the actual time difference.
I'm at 96k, not 44....sorry I didn't include that bit. Motu interface

That's what Reaper is reporting: 2.0 in/1.1 out. I don't have time to try that other stuff, sorry.

When I ran it with cakewalk, it only reported the 2.0 (sometimes 2.1) in but not the out. But yes it is up there near RME and Lynx

My project template starts at 12 tracks, there are some plugs maybe 1-2 each avg, I'm only recording 1 track at a time (vocals)
__________________
https://www.kdubbproductions.com/
https://www.youtube.com/channel/UCpC...2dGA3qUWBKrXQQ
i7 8700k,4.9Ghz,Win10,Reaper 5,Motu 828es,MJE Hulk 990,GAP Pre73/EQ81

Last edited by poetnprophet; 03-28-2019 at 04:16 PM.
poetnprophet is offline   Reply With Quote
Old 03-28-2019, 03:28 PM   #17
poetnprophet
Human being with feelings
 
poetnprophet's Avatar
 
Join Date: Jan 2018
Posts: 664
Default

Quote:
Originally Posted by Cableaddict View Post
Send I mic direct into computer B, and one mic routed through Reaper, running on computer A, into a second track on computer B.
Record in computer B, line the 2 mics up, and snap your fingers.
Then zoom way in on the 2 recorded tracks of computer B, and measure the actual time difference.
Aren't you then also including the latency of the other computer/interface? That doesn't seem right.
__________________
https://www.kdubbproductions.com/
https://www.youtube.com/channel/UCpC...2dGA3qUWBKrXQQ
i7 8700k,4.9Ghz,Win10,Reaper 5,Motu 828es,MJE Hulk 990,GAP Pre73/EQ81
poetnprophet is offline   Reply With Quote
Old 03-28-2019, 06:06 PM   #18
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by poetnprophet View Post
Aren't you then also including the latency of the other computer/interface? That doesn't seem right.
No.

You're recording 2 channels on computer B at the same time.

Whatever input latency computer B has, is the same for both tracks. You're only looking at the time difference between those 2 tracks.
===================


FWIW, I just tested my Lynx PCIe card / Aurora system, which is just about as fast as it gets, short of a few modern Thunderbolt setups:

A 64 buffer AT 44.1K gives a true i-o latency of 5.7 ms.
===========================

FWIW, (and it IS interesting) here are more numbers, using that same system, with a stripped down session:

DRIVER 23j / FW 14.1

44.1K / 32 buffer: 1.65 ms
44.1K / 64 buffer: 3.1 ms
44.1K / 128 buffer: 6.2 ms

88.2K / 64 buffer: 1.35
88.2K / 128 buffer: 3.1 ms
88.2K / 256 buffer: 5.25 ms

Last edited by Cableaddict; 03-29-2019 at 07:08 PM.
Cableaddict is offline   Reply With Quote
Old 03-28-2019, 06:41 PM   #19
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

And of course, you have to factor-in that the propagation-delay of any high-latency plugins will be half the value, at 88.2K, not to mention that certain plugins (such as amp simulators) tend to sound better at higher SR's.

So "higher sample rates at twice the buffer" wins the prize pretty much always, if it's stable.

Sadly, "higher sample rates at twice the buffer" isn't always possible. Sometimes you have to go one step higher with the buffer, which dumps the whole thing into the crapper. That's where really good audio driver design becomes very important.

I'm also DYING to see some verifiable data on whether Thunderbolt, using true direct-memory-access, can shave a little off of these numbers. I'll know for sure, sometime later this year.

Last edited by Cableaddict; 03-29-2019 at 07:07 PM.
Cableaddict is offline   Reply With Quote
Old 03-28-2019, 06:58 PM   #20
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by poetnprophet View Post
I'm at 96k, not 44....sorry I didn't include that bit. Motu interface

That's what Reaper is reporting: 2.0 in/1.1 out. I don't have time to try that other stuff, sorry.

OK, @ 96K MAYBE. (That's still amazing for USB.)

- But never trust the reported numbers inside a DAW. They are rarely accurate.
Still, I do find Reaper to be closer than other systems I've checked. With the numbers I posted, above, Reaper was correct with 2 of them. Over on 2, and under on 2. So, you're probably pretty close. Again, damned impressive for USB.

- And ESPECIALLY never trust what the ASIO panel says!
============================================

Oh, and for your reference, and my nagging brain:
(To quote that guy in the street that Dirty Harry shot, "Hey man, I GOTS ta' know!" )

My system, set to your 96K / 64 buffer, does a true 1.3 ms, with Reaper reporting 1.5 ms. - Of course, I can't actually use that setting in the real world.



(I'm hoping that my next rig will be able to run this fast, even with a full session. I think that with an i9 cpu of at least 8 cores, running at least 4.5 GHz, a Thunderbolt interface, and careful plugin distribution, this should be possible.)

Last edited by Cableaddict; 03-29-2019 at 01:13 AM.
Cableaddict is offline   Reply With Quote
Old 03-28-2019, 11:04 PM   #21
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

Quote:
Originally Posted by Cableaddict View Post
All these things really DO matter, just like moving a recorded snare backwards/ forwards by 2 ms can make a huge difference in the feel of the track.
Yeah ! We need to increase the buffer count to get a greater latency for a better feel

But of course in a "downmix" you can do offline-rendering with no CPU performance problems at any sample rate with automatic latency compensation and add latency ("delay") at any point by a plugin.

-Michael
__________________
www.boa-sorte.de

Last edited by mschnell; 03-31-2019 at 09:38 PM.
mschnell is online now   Reply With Quote
Old 03-29-2019, 12:24 AM   #22
poetnprophet
Human being with feelings
 
poetnprophet's Avatar
 
Join Date: Jan 2018
Posts: 664
Default

Yes I am very happy with the performance, win10 i7 @ 4.9ghz and a motu usb.

and sorry, i reported wrong. it's 1.3 in and 2.0 out so roundtrip of 3.3 based on reaper.
__________________
https://www.kdubbproductions.com/
https://www.youtube.com/channel/UCpC...2dGA3qUWBKrXQQ
i7 8700k,4.9Ghz,Win10,Reaper 5,Motu 828es,MJE Hulk 990,GAP Pre73/EQ81
poetnprophet is offline   Reply With Quote
Old 03-30-2019, 11:25 AM   #23
toleolu
Human being with feelings
 
Join Date: Apr 2014
Location: Sunshine and palm trees.
Posts: 556
Default

Quote:
Originally Posted by Cableaddict View Post
(I'm hoping that my next rig will be able to run this fast, even with a full session. I think that with an i9 cpu of at least 8 cores, running at least 4.5 GHz, a Thunderbolt interface, and careful plugin distribution, this should be possible.)
And if Reaper isn't written to utilize all those CPU cores, what then? "If you build it, he will come." doesn't really apply here.

Personally I have no idea how Reaper utilizes multi-core CPU's, but I would think that that's something you'd want to nail down from Justin or someone who does know before you go spending money on things you don't need.
toleolu is offline   Reply With Quote
Old 03-31-2019, 01:08 AM   #24
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

Quote:
Originally Posted by toleolu View Post
And if Reaper isn't written to utilize all those CPU cores, what then? "If you build it, he will come." doesn't really apply here.

Personally I have no idea how Reaper utilizes multi-core CPU's, but I would think that that's something you'd want to nail down from Justin or someone who does know before you go spending money on things you don't need.
It's already a known thing. Reaper's multi-core utilization is near-perfect, and has been for years.
Cableaddict is offline   Reply With Quote
Old 03-31-2019, 10:14 AM   #25
toleolu
Human being with feelings
 
Join Date: Apr 2014
Location: Sunshine and palm trees.
Posts: 556
Default

Once again, you seem to be missing the point, and please understand, I'm just trying to help. I'll admit that there's a lot about audio I don't know, but after 40 years in IT I've got a pretty good handle on what works, and what doesn't work, when it comes to computers.

If money isn't an issue for you and you want to build the biggest beast of a computer you can, that's fine, more power to you brother. But understand that you reach a point of diminished returns. Take multi cores for instance.

The concept behind multi core processors is not to make a given program perform better, it's to make multiple programs running at the same time perform better. If, while you're recording your band, you need to be checking your email, surfing the internet, playing Minecraft, or whatever, then yes, load up with as many cores as you can. If, on the other hand, you're just running Reaper when you're recording, then 8 cores is probably way more than you what you need. If you're thinking is that my performance with 4 cores is X, then by going with 8 cores my performance will be X times 2, then I think you're going to be a little disappointed. It's that old, "If a little is good, a lot must be better" thing.

But don't take my word for it, check out what the guys are doing with video editing and gaming. Those two applications are considerably more CPU intensive than audio editing and if you look at what they're doing, you'll find that the consensus is that anything above 4 to 6 cores isn't really providing much of a benefit. That is unless while you're gaming or editing your video you need to be doing other stuff like I mentioned before.

A computer is just a bunch of switches getting turned off and on a bunch of times per second. It's not a matter of how many switches you have, or how fast you're switching them on and off, it's a matter of how efficiently those switches are being turned off and on in relationship to what you're trying to do. And that my friend, is what software is all about. Not just in that the software is designed to produce the results you're looking for, but that you're using that software in the way it was designed to be used.

Have you ever considered posting something with details about what you're trying to do, how you're going about trying to do it, and the results you're expecting from doing things that way? There's a lot of really knowledgeable people on this forum and chances are, they may be able to suggest things that will get you a lot closer to the results you're looking for than any amount of hardware will. At the very least, you'll get confirmation that your processes and methodology are fine, and then at the point, you can do whatever you please with respect to hardware.

--Added to clarify things ----

I wanted to add this so as not to offend. I have no doubt, based on how long you've been on this forum and your number of posts that you've got a pretty good handle on things when it comes to Reaper and audio, that was not what I meant when I suggested getting input from others on the forum about what you're doing. Just coming from the point of view that it never hurts to get second opinions on things. If I offended you with that suggestion then I apologize.

Last edited by toleolu; 03-31-2019 at 10:58 AM. Reason: Point clarification
toleolu is offline   Reply With Quote
Old 04-02-2019, 03:23 PM   #26
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

^ I appreciate the effort, (really) but I'm way past that point, plus what you write is not completely accurate. When talking about low-latency DAWs, # of cores can matter a LOT, in conjunction with track layout & plugin distribution. (and this is an area that no one, including me, seems to have a complete handle on yet.)

Additionally, any info or benchmarks concerning both gaming and video are basically irrelevant. Sadly, we audio people inhabit a very tiny and fairly unique niche of the computer world.

Anyway, it's completely off track for the question I am trying to get an answer to, in this thread.

==============================================



Can anyone answer my original question, regarding thunderbolt, PCIe cards, and DMA?
Cableaddict is offline   Reply With Quote
Old 04-02-2019, 10:02 PM   #27
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

Quote:
Originally Posted by Cableaddict View Post
Sadly, we audio people inhabit a very tiny and fairly unique niche of the computer world.
Nonetheless there are companies specialized in exactly this issue - including the thunderbold issue in your initial post.

E.g: "Digital AudionetWorX" -> https://www.da-x.de/

Here, they provide links to (among others) the report in the "Sound & Recording" magazine I mentioned above -> https://www.da-x.de/de/audio-compute...c-systeme.html

Moreover dedicated Music Stores do offer especially crafted Audio PCs featuring Thgunderbold 3 interface. e.g. -> https://www.musicstore.de/de_DE/EUR/...4AAAFow5KP7DyZ , and (hopefully) provide appropriate competence to their customers.

-Michael
__________________
www.boa-sorte.de

Last edited by mschnell; 04-02-2019 at 10:09 PM.
mschnell is online now   Reply With Quote
Old 04-03-2019, 06:45 PM   #28
Cableaddict
Human being with feelings
 
Join Date: Apr 2008
Posts: 1,883
Default

I don't want to buy a pre-made PC.

I want specific information, so as to better optimize my own builds.



Obviously, I'm wasting my time here.
Cableaddict is offline   Reply With Quote
Old 04-03-2019, 07:29 PM   #29
analogexplosions
Human being with feelings
 
analogexplosions's Avatar
 
Join Date: May 2011
Location: Brooklyn
Posts: 315
Default

What is this about latency issues on Dante?

I'm getting ~6m round-trip with Dante Virtual Soundcard, which in itself is kinda made to be a poor-man's method to get audio in and out from the Dante network.

If using a Dante PCIe card, you should be able to get right at about 3ms roundtrip including any latency from converters coming in AND going out.

With any other connection, its going to be down to the quality of the drivers really. Thunderbolt and PCIe should have similar latency numbers, as Thunderbolt is a direct connection to the PCIe Bus, just through a cable.

That being said, you can get similar latency performance from something like RME over USB.
__________________
www.dungeonbeach.com
analogexplosions is offline   Reply With Quote
Old 04-03-2019, 09:55 PM   #30
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Krefeld, Germany
Posts: 7,111
Default

Quote:
Originally Posted by Cableaddict View Post
I don't want to buy a pre-made PC.
Obviously, I'm wasting my time here.
As always in life, you either need to take a lot of time to build own competence in constructing stuff, or to shell out money for somebody doing this for you
-Michael
__________________
www.boa-sorte.de
mschnell is online now   Reply With Quote
Old 04-26-2019, 03:56 PM   #31
mlprod
Human being with feelings
 
Join Date: Jul 2015
Location: Stockholm, Sweden
Posts: 635
Default

Quote:
Originally Posted by Cableaddict View Post
I'll be upgrading both my main PC and my audio system soon.

I'll be using a Lynx Aurora(n), but I'm a bit puzzled by current connection options. I know that Dante is ridiculously expensive (due to licensing issues?) plus it currently has latency issues. So my two choices are Thunderbolt, or staying with PCIe

With audio, I see that PCIe cards are pretty much going the way of the Flintstones, with Thunderbolt-equipped mobos starting to be standard. therefore, I'll probably go that way, but I'd REALLY like to understand more. I've read countless articles, but I have yet to find one that fills me with confidence.
I trust some of the computer gurus here more than on most other forums.

SO, MY QUESTIONS:
----------------------------------------

1: Does Thunderbolt have the potential to offer lower DAW latency than even a PCIe card?

I read something, rather badly explained, about how TB uses DMA so that the audio goes straight to & from your ram, thus creating no cpu overhead. That's sounds tasty to me, but is that also the case with PCIe cards, or not?


2: If using a PCIe Thunderbolt adapter card, with my current non-TB motherboard, should I expect LESSER performance, vs a mobo with built-in TB-3?


3: Do all TB3-equipped motherboards offer the exact same TB performance?

I ask because I read something (again, very badly explained) about how some TB implementations use some kind of controller chip, whereas the "real thing" has a more direct path to the promised land.

---------------------------------

### LAST: Do any gains that might happen exist because the system will allow for a smaller hardware buffer, or are those gains created elsewhere? (With the buffer size still determined by cpu speed, driver design, etc)
Regarding question 1, i think not.

Btw, I am running an RME HDSPe MADI FX on an Asus WS x299 Pro with i9 9900x at 4,5GHz and the performance is just ridiculous.
__________________
Magnus Lindberg Productions - VRTKL Audio - Redmount Studios
magnuslindberg.com
mlprod is offline   Reply With Quote
Old 04-27-2019, 08:44 AM   #32
cyrano
Human being with feelings
 
cyrano's Avatar
 
Join Date: Jun 2011
Location: Belgium
Posts: 4,169
Default

I don't know for sure, but I've always assumed TB doing DMA. How else do you sustain these speeds?

EDIT

It just dawned on me it does DMA, as there is a security mechanism to prevent attacks through it. You could fi, read keys in ram via TB or FW.

I don't know if DMA is used for audio, though.
__________________
“It has become appallingly obvious that our technology has exceeded our humanity” Albert Einstein
cyrano is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 04:05 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.