Old 10-01-2018, 02:06 PM   #1
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default Clock Speed Vs Core Count?

At the moment it is somewhat impractical for me to use one central pc for gaming and audio work, so I'd like to piece together a dedicated machine that is going to run as efficiently as possible. I have considerable experience building machines, so I have a pretty good idea of how to get the best performance with a very quiet machine at a low cost. What I need to know is how Reaper utilizes resources when recording and loading vsts, is it heavy on a single core, infinitely scale-able, or somewhere in between? How well does it handle ram, is it dual/triple/quad channel friendly, does ddr4 provide any noticeable performance benefit over more affordable ddr3 given the added latency? Is Reaper stable at 8 gigs, 32 gigs, 128?

I have plenty of good options either way to build a nice quiet machine with a low power draw, or a more powerful machine that can be quiet when it needs to be, and go full tilt when required, but I need to know what Reaper can do first, to avoid dropping money for little/no benefit.

Some use cases:
I need as low a latency as possible for real time tracking of various instrument vsts and vst effects. It could be as simple as tracking a rock tune with Guitar cab sims/impulses, piano vst's, and drum vsts, all the way up to compositions with dozens of orchestral instruments and synthesizers. Right now my weak link is my interface, it's limited to 64 samples at 48Khz, I'm shooting for 16-32 samples at 96-192Khz, which should destroy my latency problem, but I need a chip that can handle that without crackling.

Right now I've been using an I52500 at 3.3ghz, and I've pushed it to it's limit on a few occasions at 64 samples, with the native M-audio ASIO driver. I'm also running 8 gb's of ddr3. For all I know Reaper might only be using 2 cores, in which case a high clocked dual core would fit the bill, or it might scale perfectly across six or more, in which case an Xeon might be more efficient per watt requiring less cooling power. I might also be hitting a memory limit as well when loading in vsts. Basically I need to understand what makes Reaper tick, so that I can maximize a pc build and get the best performance per watt and $.
Taffer is offline   Reply With Quote
Old 10-01-2018, 03:29 PM   #2
Steviebone
Human being with feelings
 
Join Date: Jul 2018
Posts: 325
Default

I'm not an expert and I'm sure there are others here that can chime in with more specific expertise. That said, I have built numerous machines intended for studio use over the last 20+ years. I am, however, new to Reaper.

When I fire reaper up on my machines with numerous VSTs loaded all of the cores are active. This leads me to believe that the cores are being utilized appropriately. How many cores will reaper utilize? I don't know. However I do have one 8 core machine and it looks like reaper utilizes them fine. In light of this experience I would say that after a fashion cores are more important than processor speed. Especially if you utilize a lot of plug-ins. How it factors into track count I don't know.

With regards to memory I believe reaper is mostly a streaming technology so audio probably doesn't require boatloads of RAM. However, once again, if using a lot of plug-ins each plug-in consumes memory so the more memory the more plug-ins. You definitely don't want Windows paging during operations. That said, many plug-ins don't require a lot of memory so even eight gig would probably suffice for many users. However, if you are using sample based libraries memory can become critical. I have a Kontakt instance on a separate machine that utilizes 32 gigs of RAM by itself.

As for memory speed I'm not sure that makes as much of a difference. That would be attributable to overall computer performance and of course faster is always better. How much better? I would think that there would be other areas where you would see a more noticeable improvement. I would rather have more cores than faster memory for example.

With regards to latency I think this has to do primarily with your interface and its driver than raw computing speed. Again, while I have a good deal of experience I am not the expert here. Good luck.

PS: Using the same computer for gaming and DAW work is probably not optimal. All of my musical computers are single purpose machines. This avoids having a unnecessary drivers and services loaded that reduce your available CPU resources. You'd be surprised at how many things Windows loads at startup that you don't need and that are consuming base resources. On a single purpose machine you can disable dozens of unneeded services that will make the machine more stable and run faster for audio work.
Steviebone is offline   Reply With Quote
Old 10-01-2018, 06:21 PM   #3
mhuss
Human being with feelings
 
mhuss's Avatar
 
Join Date: Jun 2009
Location: West of Philly
Posts: 285
Default

The problem as I see it is that it appears all available processor threads are being used, but at any given time it seems 7 are waiting for the 8th to finish some blocking task! :-\

When you have serial VSTs in a chain, VST2 can't start working on a chunk of sound data until VST1 gets finished with it, and so on down the line. If VST1 is a reverb or fancy convolution modeler, you may be waiting for a (relative) while!

I guess the ideal is fast as possible with at least 8 threads.
mhuss is offline   Reply With Quote
Old 10-01-2018, 06:33 PM   #4
tack
Human being with feelings
 
tack's Avatar
 
Join Date: Jan 2014
Location: Ontario, Canada
Posts: 445
Default

Quote:
Originally Posted by Taffer View Post
I need as low a latency as possible for real time tracking of various instrument vsts and vst effects.
Then this should be simple: you need to optimize for clock speed.

Maybe hold out for a 9900k, where you can get a generous number of cores (for your non realtime work) and superb single core performance?

Reaper can make excellent use of many cores -- it's really quite good here -- but you said the magic words: "low latency" and "real time tracking." Little else matters for that but single core performance.
tack is offline   Reply With Quote
Old 10-01-2018, 07:07 PM   #5
Steviebone
Human being with feelings
 
Join Date: Jul 2018
Posts: 325
Default

Quote:
Originally Posted by tack View Post
Then this should be simple: you need to optimize for clock speed.

Maybe hold out for a 9900k, where you can get a generous number of cores (for your non realtime work) and superb single core performance?

Reaper can make excellent use of many cores -- it's really quite good here -- but you said the magic words: "low latency" and "real time tracking." Little else matters for that but single core performance.
Well I have older machines that used the same audio interface and latency was essentially the same for me. I think latency has more to do with your interface than CPU clock speed after a point. In the past I have upgraded interfaces and seen immediate drop in latency, and upgraded computers with much less noticeable difference in latency. While overall clock speed factors to an extent, getting the signal INTO your DAW is mostly a function of your interface after a point, unless your computer is really old.

There's no such thing as too much speed. But if you have a super-computer and a crappy interface/driver you're still going to have latency. I think this bears out in that some interfaces have settings for much lower latency than others. Doesn't matter how fast your CPU is if your interface has an inefficient driver and/or hardware design.

There are others here tho that know the inner workings of Windows multiprocessing better than I. But that's my two cents. Of course if you can do both....
Steviebone is offline   Reply With Quote
Old 10-01-2018, 07:24 PM   #6
tack
Human being with feelings
 
tack's Avatar
 
Join Date: Jan 2014
Location: Ontario, Canada
Posts: 445
Default

True, overall latency is the product of a multi-part signal chain. If you're doing very trivial things (like a basic rompler with no post-processing) then indeed other things (like your audio interface, driver quality, etc.) start to matter a lot more in proportion.

For more involved processing, especially at such extremely small buffer sizes like 16, it's going to challenge the best of processors to keep those buffers consistently full. Windows in particular isn't the greatest at latency sensitive realtime workloads, so having a generous amount of headroom is going to be important to keep things consistent. A 16 sample buffer at 96kHz with realtime tracking means processing a block every ~160 usecs or so, and one small hiccup that might go completely unnoticed at 128 samples means a clicky-poppy buffer underrun.

You know, I'm not so sure Windows is up to the task.

Does anyone here successfully run such low latency with Windows?
tack is offline   Reply With Quote
Old 10-01-2018, 09:44 PM   #7
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 22,083
Default

I think RME doesn't even give you the option of doing 16 samples at 96k, only 32. Or perhaps even 64.
EvilDragon is offline   Reply With Quote
Old 10-02-2018, 04:36 AM   #8
tack
Human being with feelings
 
tack's Avatar
 
Join Date: Jan 2014
Location: Ontario, Canada
Posts: 445
Default

My Babyface Pro only goes to 48 samples at 48kHz. (And unsurprisingly the Threadripper 2950X I have in my system right now isn't capable of driving that without underruns even with a fairly basic piano instrument in Kontakt.) At 96kHz, my lowest buffer available is 96 samples.

I imagine the higher end RME devices could run with smaller buffers but I'd need a different processor. (Again that's no surprise -- Threadripper wouldn't be the right choice for extremely low latencies.)

Last edited by tack; 10-02-2018 at 04:51 AM.
tack is offline   Reply With Quote
Old 10-03-2018, 02:15 PM   #9
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 7,441
Default

For an interface to be used in live sound low latency needing situations, shop for one that will be at least 10ms round trip latency or lower at a 128 sample block size running at 48k. (48k is still the sweet spot for CPU vs. latency. 96k may have a much lower latency with the same number of samples but it crosses the line into more CPU cycles needed.)

You don't want a live system that you'd need to crank down to 64 or 32 samples just to hit the 10ms mark or there goes your headroom.

For cores vs. speed in a DAW system, shop for speed first.
serr is offline   Reply With Quote
Old 10-03-2018, 02:44 PM   #10
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Right now my weak link is my interface, it's limited to 64 samples at 48Khz,
So that is what 1.3 ms? 48000/1000 = 48 = 64/48 = 1.3 - Maybe 3 ms RTL? - 128th notes @ 120 BPM is around 30+ ms? I'm going to propose that there may be additional unaccounted for latency unless you are playing passages that are extremely fast where the playing catches up with the buffer so to speak. Or I'm just misreading your post (possible).

That said, both core count and speed matter IMHO, count for distributing load, speed for the real time thread which services the audio driver and lives on a single core IIRC.
karbomusic is offline   Reply With Quote
Old 10-03-2018, 03:03 PM   #11
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 7,441
Default

I'd guess that meant 64 samples was the lowest the block size could be set and still have stable operation. But it doesn't tell us what the round trip latency was at that block size setting at all.

The lowest base level latency a hardware audio interface can do at any block size setting in your DAW app is what it is regardless of the block size.

Any reasonable modern computer can probably be made to run stable at a 64 sample block size depending on the plugins being used. The fast CPU systems from the last 10 years can be made to run at 32. But you don't want to go with just whatever from an interface and think you're going to make it up on the computer with a lower block size or you'll have used up all your headroom right out of the gate. So shoot for hitting 10ms or less round trip latency running at 48k and with a block size of 128 samples. Most of the USB units don't make the cut for that and would be a PITA for live work.

Get a proper low latency supporting interface to begin with and actually get to use the computing power you have and have some headroom there.
serr is offline   Reply With Quote
Old 10-03-2018, 03:07 PM   #12
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Originally Posted by serr View Post
I'd guess that meant 64 samples was the lowest the block size could be set and still have stable operation. But it doesn't tell us what the round trip latency was at that block size setting at all.
It means "at least" but the napkin math shows there's likely other latency that is unaccounted for even if we double it for round trip. That said, my setup is only 53 samples off of what is reported to reaper but I've always found 'some' discrepancy for the last 20 years because there are enough moving parts that only a loopback can determine true RTL.

One point I was hinting at is I'm not sure needing less than 64 samples at the interface level is going to save the world here, minus details I'm not yet aware of.
karbomusic is offline   Reply With Quote
Old 10-03-2018, 03:11 PM   #13
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 7,441
Default

That was my point. Only a loopback test can tell you what your actual latency is.
The block size and sample rate by themselves tell us nothing about the actual latency of the hardware system. Truly nothing. Not even a hint of what might be possible.

And then the bit: Don't try to make up for a higher baseline latency interface with a low block size as a workaround. You want headroom here. The higher you can set the block size to achieve the needed low latency, the more bigger plugins you will be able to run without issue.
serr is offline   Reply With Quote
Old 10-03-2018, 04:40 PM   #14
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Originally Posted by serr View Post
That was my point. Only a loopback test can tell you what your actual latency is.
The block size and sample rate by themselves tell us nothing about the actual latency of the hardware system. Truly nothing. Not even a hint of what might be possible.
We can be 100% sure it as at least 64 samples per, on my system it is roughly exactly double +53 samples (actually @64 it's less than 53 but can't measure right now). So "truly nothing an not even a hint" isn't really correct, it gives us everything but the unaccounted for pieces which shouldn't be 'that' much. Semantics yes but those block sizes absolutely do tell us part of the story. IOW, if you set it to 64 but in reality it's 320, something is terribly wrong.
karbomusic is offline   Reply With Quote
Old 10-03-2018, 04:48 PM   #15
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 7,441
Default

Fair enough! (Obvious logic lost on me for a moment there it seems.)

We really do need to know what that hardware interface adds though. But you knew that.
serr is offline   Reply With Quote
Old 10-03-2018, 04:48 PM   #16
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Originally Posted by serr View Post
Fair enough! (Obvious logic lost on me for a moment there it seems.)

We really do need to know what that hardware interface adds though. But you knew that.
Yea I'm with you, in reality we are on the same page.
karbomusic is offline   Reply With Quote
Old 10-05-2018, 09:15 PM   #17
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

All right, so what do I want to buy then? Clearly there is more to the latency than sample size and sample rate, that I'm not currently aware of. So am I to understand that usb devices are incompetent for low latency vst instruments? If so perhaps we should squash the optimism of broke college kids everywhere, who think they can buy a cheap usb interface, and fire up Guitar Rig or Amplitube as an affordable amp replacement, or using Addictive drums with a 300$ ekit. Why is it that I can get a reasonable latency with an M-Audio Keystation into VSCO, and yet when I try out Amplitube or an Ekit with Mtpower, suddenly latency is unusable? I want consistently low latency and I don't want to pay anymore than I need to, what exactly do I want?
Taffer is offline   Reply With Quote
Old 10-06-2018, 12:43 AM   #18
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

Why is it so damn difficult to get a straight answer regarding usb interface latency? I've asked various Long and Mcquade associates, gearslutz forum member, harmony central forum members, and no one seems to have a clue whether or not usb interfaces are capable of undetectable latency. I remember back in the day when I first started on this journey there was talk of firewire being superior to usb 2.0. More recently I asked around at gearslutz and was told usb 3.0 is competent, and a Komplete 6 was recommended. Then I called up L&M only to be told that the Klomplete 6 is usb 2.0, as are all of the other usb interfaces I've seen. They also don't allow returns on interfaces due to software bundles. I'm so damn tired of this decade long search for a competent product, what do I NEED!? Not how much $ can we convince op to throw away today, not more of the same "less than 16ms is undetectable so usb interfaces should be fine". I need hard data and options. USB 2.0? USB 3.0? Firewire? Thunderbolt? (Is that even an option on pc?...Hackintosh?) Pci/Pcie? I don't mean any disrespect to forum members here, I can't quite put into words how frustrating this quest for competent latency performance has been. I don't need 8-12 inputs, I don't need top of the line mic pre amps, I just need competent latency performance, and I don't want to pay extra $ for the aforementioned stuff I don't need. I like to shred on guitar, Guthrie Govan, MAB, Rusty Cooley, I'm not going to pretend I'm as competent a guitarist as they are, but I've developed my technique for over a decade and I play every bit as fast as they do. I'm not bragging here, but if there is even slight latency I'm going to notice it.
Taffer is offline   Reply With Quote
Old 10-06-2018, 02:49 AM   #19
mikemana
Human being with feelings
 
Join Date: May 2012
Posts: 18
Default

Split your signal. Run one though a hardware emulator ie.Tech21 character pedals or JOYO clones. record both but only monitor the pedal and on zero latency option.
mikemana is offline   Reply With Quote
Old 10-06-2018, 03:00 AM   #20
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 22,083
Default

Quote:
Originally Posted by tack View Post
My Babyface Pro only goes to 48 samples at 48kHz. (And unsurprisingly the Threadripper 2950X I have in my system right now isn't capable of driving that without underruns even with a fairly basic piano instrument in Kontakt.) At 96kHz, my lowest buffer available is 96 samples.

I imagine the higher end RME devices could run with smaller buffers but I'd need a different processor. (Again that's no surprise -- Threadripper wouldn't be the right choice for extremely low latencies.)
I checked. My UFX+ goes to 32 samples at 48k, 64 samples at 96k.


To answer to Taffer: setting the audio buffer size is just one part of the equation. This is only what the ASIO driver reports as processing latency. There are other latencies involved: ADC and DAC latencies (which are usually just a few samples, in general about 1-1.5 ms for both, but the main problem is USB bus latency. Presonus has a pretty good picture that explains things:



I always like to use this picture because it's very illustrative of what's happening. Except in our case, instead of the green box in the middle, we have ASIO input/output buffer latency.

In any case: USB2 is ok when you have really well written drivers and don't use off-the-shelf USB controllers (which almost all audio interface vendors do, but NOT RME - they have their own USB controller done in FPGA). USB3 is better, but Thunderbolt is the best (and yes, it will work on Windows, at least Thunderbolt 2 - again, RME has a solution here, but it's quite expensive, and you don't need that many inputs for sure). Forget Firewire, it's dead for some time now.

Get a Babyface Pro would be my recommendation. But just remember that there will always be some latency. It's unavoidable. What can be done, though, is bringing it down to acceptable levels (which would be below 10 ms round-trip latency). This should be possible with Babyface.

Last edited by EvilDragon; 10-06-2018 at 03:12 AM.
EvilDragon is offline   Reply With Quote
Old 10-06-2018, 08:33 AM   #21
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 7,441
Default

Quote:
Originally Posted by Taffer View Post
All right, so what do I want to buy then? Clearly there is more to the latency than sample size and sample rate, that I'm not currently aware of. So am I to understand that usb devices are incompetent for low latency vst instruments? If so perhaps we should squash the optimism of broke college kids everywhere, who think they can buy a cheap usb interface, and fire up Guitar Rig or Amplitube as an affordable amp replacement, or using Addictive drums with a 300$ ekit. Why is it that I can get a reasonable latency with an M-Audio Keystation into VSCO, and yet when I try out Amplitube or an Ekit with Mtpower, suddenly latency is unusable? I want consistently low latency and I don't want to pay anymore than I need to, what exactly do I want?
USB connecting interfaces aren't all just flat out unable to run at low enough latency for live sound work or live performance with MIDI instruments. Just the cheap and/or older ones.

Back when the first firewire models came out and USB 2 wasn't yet a thing, you had the situation where "USB isn't an option, you need firewire." Nowadays you have all these very low budget USB interfaces made. Cheapness is cheapness. In between you have some pro models that are both USB and run at a base line low latency. One of the expensive firewire or thunderbolt models may be faster still but there are USB options that work.

And yeah, the kids buying the cheapo USB interfaces and trying to use them for live performance are making a mistake. Trying to make up for a cheap interface with a more bangin' computer would also be a mistake. More mileage the other way.
serr is offline   Reply With Quote
Old 10-09-2018, 09:36 PM   #22
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

Quote:
Originally Posted by EvilDragon View Post
I checked. My UFX+ goes to 32 samples at 48k, 64 samples at 96k.


To answer to Taffer: setting the audio buffer size is just one part of the equation. This is only what the ASIO driver reports as processing latency. There are other latencies involved: ADC and DAC latencies (which are usually just a few samples, in general about 1-1.5 ms for both, but the main problem is USB bus latency. Presonus has a pretty good picture that explains things:



I always like to use this picture because it's very illustrative of what's happening. Except in our case, instead of the green box in the middle, we have ASIO input/output buffer latency.

In any case: USB2 is ok when you have really well written drivers and don't use off-the-shelf USB controllers (which almost all audio interface vendors do, but NOT RME - they have their own USB controller done in FPGA). USB3 is better, but Thunderbolt is the best (and yes, it will work on Windows, at least Thunderbolt 2 - again, RME has a solution here, but it's quite expensive, and you don't need that many inputs for sure). Forget Firewire, it's dead for some time now.

Get a Babyface Pro would be my recommendation. But just remember that there will always be some latency. It's unavoidable. What can be done, though, is bringing it down to acceptable levels (which would be below 10 ms round-trip latency). This should be possible with Babyface.
So I need to spend $1k to get the best of an inferior connection method (USB)? If I actually need to drop $1k to get a competent usb interface, I wouldn't mind spending a few hundred more for a competent interface utilizing a better connection method (Thunderbolt, Pcie etc). If cutting down on inputs and pre amp quality (without sacrificing latency performance) doesn't save many $, I might as well spend the money for the best of the best, not the best of the mediocre...
Taffer is offline   Reply With Quote
Old 10-09-2018, 09:52 PM   #23
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Germany
Posts: 6,000
Default

In the German "Sound and Recording" magazine, I recently read a test of a Windows PC dedicatedly built for DAW use by a company who offers several of such types. This one did feature a Thunderbold card and I seem to remember that this company created Windows drivers for same.

AFAIR, the comments were enthusiastic.

See -> https://www.soundandrecording.de/equ...olt-3-im-test/ :
"Nach dieser positiven Erfahrung empfand ich es überaus schmerzlich, von der DA-X Audio Workstation Extreme mit Thunderbolt 3 wieder auf meinen eigenen Audio-PC mit USB 2.0 wechseln zu müssen" ("After this positive experience it hurt extremely to need to return to my own Audio PC with USB2.0")

So maybe you need to be "spending a few hundred more" (at least $ 2500, the Thunderbold card is $ 100 ) .

-Michael
__________________
www.boa-sorte.de

Last edited by mschnell; 10-10-2018 at 06:34 AM.
mschnell is online now   Reply With Quote
Old 10-09-2018, 09:57 PM   #24
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

I'm just going to start throwing some options out there, hopefully you guys can tell me why they suck:
https://www.long-mcquade.com/16490/P..._Interface.htm
https://www.long-mcquade.com/12886/P...te_Audio_6.htm
https://www.long-mcquade.com/64050//..._Interface.htm
https://www.long-mcquade.com/98051/P...-Interface.htm
https://www.long-mcquade.com/103975/...ansformers.htm
https://www.long-mcquade.com/54623//..._Interface.htm
https://www.long-mcquade.com/67712//...o_Inteface.htm
https://www.long-mcquade.com/88456/P...Deep-Black.htm
https://www.long-mcquade.com/104884/...-Interface.htm

And so on: https://long-mcquade.com/departments...top-pagination

What makes the premium priced 1000$ Rme Babyface Usb, better than any of the other 400-600$ usb, firewire, and thunderbolt interfaces, and why doesn't L&M have any Pci or Pcie options?

Last edited by Taffer; 10-09-2018 at 10:49 PM.
Taffer is offline   Reply With Quote
Old 10-10-2018, 12:24 AM   #25
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 22,083
Default

Babyface is not $1000, it's $750. But I see, you're talking Canadian dollars.

What makes it premium? The whole package. You're getting drivers nobody else in that (or lower) price range has. They are simply the best.

Ask L&M why they don't stock any PCIe options?


But better question for you: if you don't need to have a bajillion of I/O, why are you concerned if something is USB or TB? TB is most often used on audio interfaces that have a lot of I/O, MADI, etc. (Clarett being one notable exception, but Focusrite's drivers, while ok, aren't at RME level). For up until 32 ins and outs, USB suffices just fine, and Babyface showcases that perfectly. Plus, it is compact and bus powered, which means portable as well.
EvilDragon is offline   Reply With Quote
Old 10-10-2018, 04:08 PM   #26
JSMastering
Human being with feelings
 
JSMastering's Avatar
 
Join Date: Aug 2018
Posts: 169
Default

It's not bus powered, but I've been using a Native Instruments A10mk2 as a DJ interface for years with 8 mono channels in and 10 out. The drivers have been stable through most of that, and the latency is low enough that I can mix and scratch just fine. Or, at least, I could back when I did that. (I still use it to DJ for my dogs; I just don't use vinyl control anymore).

USB is plenty capable as long as stupid things aren't happening, and that setup always had a lot of stuff hanging off the USB ports, so it was on the stupid side.
JSMastering is offline   Reply With Quote
Old 10-10-2018, 04:54 PM   #27
Patrice Brousseau
Human being with feelings
 
Join Date: Apr 2013
Posts: 62
Default

There is also the Zoom UAC2, reputed to have really low RTL latency, particularly if connected to Intel USB3 controller...
https://www.long-mcquade.com/61635/P...gestive-search
Patrice Brousseau is offline   Reply With Quote
Old 10-10-2018, 05:50 PM   #28
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
But better question for you: if you don't need to have a bajillion of I/O, why are you concerned if something is USB or TB?
Pretty much ^this. USB2 got an unnecessary bad rap due to companies who's primary business isn't the sound card business or even when it was, but their product was more a hodgepodge of known components and third-party no name prebuilt stuff using class compliant drivers. RME rolling their own with FGPA pretty much proves USB2 concerns are an old wives tale until you need fairly high track counts.

I easily record 24 simultaneous audio tracks in real-world conditions with out issue using my RME Fireface and measly Microsoft surface and all that going to a frigging micro SD card for that matter with zero issues.
karbomusic is offline   Reply With Quote
Old 10-10-2018, 05:56 PM   #29
Glennbo
Human being with feelings
 
Glennbo's Avatar
 
Join Date: Mar 2008
Location: Planet Earth
Posts: 2,576
Default

Quote:
Originally Posted by karbomusic View Post
Pretty much ^this. USB2 got an unnecessary bad rap due to companies who's primary business isn't the sound card business or even when it was, but their product was more a hodgepodge of known components and third-party no name prebuilt stuff using class compliant drivers. RME rolling their own with FGPA pretty much proves USB2 concerns are an old wives tale until you need fairly high track counts.

I easily record 24 simultaneous audio tracks in real-world conditions with out issue using my RME Fireface and measly Microsoft surface and all that going to a frigging micro SD card for that matter with zero issues.
Isn't true that USB2 is streaming packets, vs the now almost obsolete Firewire method of streaming data?

I've often wondered if RME is possibly circumventing the default nature of USB2 and doing something more like Firewire to achieve their unmatched performance.
__________________
Glennbo
Soundclick - http://www.soundclick.com/glennbo
--
Glennbo is offline   Reply With Quote
Old 10-10-2018, 06:07 PM   #30
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

Quote:
Originally Posted by EvilDragon View Post
the main problem is USB bus latency.

In any case: USB2 is ok when you have really well written drivers and don't use off-the-shelf USB controllers (which almost all audio interface vendors do, but NOT RME - they have their own USB controller done in FPGA).

USB3 is better, but Thunderbolt is the best (and yes, it will work on Windows, at least Thunderbolt 2 - again, RME has a solution here, but it's quite expensive, and you don't need that many inputs for sure). Forget Firewire, it's dead for some time now.
Quote:
Get a Babyface Pro

What makes it premium? The whole package. You're getting drivers nobody else in that (or lower) price range has. They are simply the best.

But better question for you: if you don't need to have a bajillion of I/O, why are you concerned if something is USB or TB? TB is most often used on audio interfaces that have a lot of I/O, MADI, etc. (Clarett being one notable exception, but Focusrite's drivers, while ok, aren't at RME level).


I don't know whether to congratulate you, or question your sanity. That's some next level contradiction the likes of which I will probably never again witness
Don't get me wrong, I appreciate you taking the time to provide advice, but I assume you have an explanation for the conflicting advice? I don't mind having a bunch of inputs or high end pre amps, I just don't want to pay significantly more for it. If a competent 2 io usb interface is 1000$, and a superior 8 io thunderbolt interface is 1000$, why would anyone buy the inferior product?
Taffer is offline   Reply With Quote
Old 10-10-2018, 06:08 PM   #31
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Originally Posted by Glennbo View Post
Isn't true that USB2 is streaming packets, vs the now almost obsolete Firewire method of streaming data?

I've often wondered if RME is possibly circumventing the default nature of USB2 and doing something more like Firewire to achieve their unmatched performance.
Streaming bytes instead of packets, as far as I know, yes. RME wrote their own (probably to the edge or to circumvention) instead of relying on class-compliant - where class compliant means if you meet certain default criteria, the OS built-in driver will just work (meaning virtually no labor by you the manufacturer for the driver development) but that is both slower and limits custom functionality.

IMHO it's very tempting and in some cases required - A) you don't have to really be a sound card company to offer a product that is a "sound card", just get someone to make you a vanilla one, stick you logo on it and you're now a sound card company or B) you can't offer a card at a price low enough for users with tiny budgets to afford so the trade off is latency and full featured driver support for price.

As far as I can tell ^that is the biggest difference between OK, good and great soundcards; those who design and develop sound card technology vs those who buy and resell it.

The bigger point though is that USB2 is perfectly capable of handling real-world demands until you get to some high track count where the bandwidth needed exceeds the bandwidth it is capable at transmitting - regardless of the underlying method of transmission. My Fireface UFX is BOTH USB2 and Firewire and I honestly can't tell bit of difference, at least up to somewhere less that 30 simultaneous incoming streams.
karbomusic is offline   Reply With Quote
Old 10-10-2018, 06:14 PM   #32
Taffer
Human being with feelings
 
Join Date: May 2018
Posts: 45
Default

Quote:
Originally Posted by Patrice Brousseau View Post
There is also the Zoom UAC2, reputed to have really low RTL latency, particularly if connected to Intel USB3 controller...
https://www.long-mcquade.com/61635/P...gestive-search
Now we're talking, 300$ seems like a reasonable place to start when updating from a 100$ outdated interface. At least it properly advertises it's usb version.
Taffer is offline   Reply With Quote
Old 10-10-2018, 06:27 PM   #33
Glennbo
Human being with feelings
 
Glennbo's Avatar
 
Join Date: Mar 2008
Location: Planet Earth
Posts: 2,576
Default

Quote:
Originally Posted by karbomusic View Post
The bigger point though is that USB2 is perfectly capable of handling real-world demands until you get to some high track count where the bandwidth needed exceeds the bandwidth it is capable at transmitting - regardless of the underlying method of transmission. My Fireface UFX is BOTH USB2 and Firewire and I honestly can't tell bit of difference, at least up to somewhere less that 30 simultaneous incoming streams.
I never record more than four simultaneous inputs, but with the awesome performance I get from a pair of ancient M-Audio Delta 2496 cards, if I were to get something to replace them I would consider this card that I have looked at about a million times.

https://www.sweetwater.com/store/det...fall-hdsp-9632
__________________
Glennbo
Soundclick - http://www.soundclick.com/glennbo
--
Glennbo is offline   Reply With Quote
Old 10-10-2018, 06:30 PM   #34
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 22,631
Default

Quote:
Originally Posted by Glennbo View Post
I never record more than four simultaneous inputs, but with the awesome performance I get from a pair of ancient M-Audio Delta 2496 cards, if I were to get something to replace them I would consider this card that I have looked at about a million times.

https://www.sweetwater.com/store/det...fall-hdsp-9632
I worked in a music store when the original 2496 cards came out and they were always raved about and highly regarded, never heard a bad thing about them. They were designed pre-Avid when M-Audio was the brand and the company IIRC. Actually they came out a couple years after the Delta-66s, I had gotten two of those not long before the 2496s came out. I wanted a 1010, another historic workhorse but I couldn't afford it at the time. I ended up with two 66s and an Omni IO.

The Hammerfall should be pretty bad ass though. RME doesn't mess around IMHO.
karbomusic is offline   Reply With Quote
Old 10-10-2018, 06:56 PM   #35
Glennbo
Human being with feelings
 
Glennbo's Avatar
 
Join Date: Mar 2008
Location: Planet Earth
Posts: 2,576
Default

Quote:
Originally Posted by karbomusic View Post
I worked in a music store when the original 2496 cards came out and they were always raved about and highly regarded, never heard a bad thing about them. They were designed pre-Avid when M-Audio was the brand and the company IIRC. Actually they came out a couple years after the Delta-66s, I had gotten two of those not long before the 2496s came out. I wanted a 1010, another historic workhorse but I couldn't afford it at the time. I ended up with two 66s and an Omni IO.

The Hammerfall should be pretty bad ass though. RME doesn't mess around IMHO.
In my case, since I'm using the Linux version of REAPER pretty exclusively now, the M-Audio cards perform as low latency as I see running the Windows version of REAPER.

If I had to replace them, the RME HDSP 9632 would likely be my first choice due to it being a PCI device and running at PCI bus speed whether in Windows or Linux, just like the M-Audio cards.

RME's USB2 devices work in Linux, but not as well as they do in Windows, due to them only having class compliant drivers in Linux.

Edit:

BTW, it was Tim Ryan who started M-Audio, sometime after he started MidiMan, which was sometime after he wrote my favorite C64 midi sequencer "Studio One" by "Third Street Software" which was after he wrote my previous midi sequencer that was marketed by Sequential Circuits.

https://en.wikipedia.org/wiki/Tim_Ryan_(engineer)

OK, enough trivia, back to audio interfaces.
__________________
Glennbo
Soundclick - http://www.soundclick.com/glennbo
--

Last edited by Glennbo; 10-10-2018 at 07:05 PM.
Glennbo is offline   Reply With Quote
Old 10-10-2018, 09:49 PM   #36
mschnell
Human being with feelings
 
mschnell's Avatar
 
Join Date: Jun 2013
Location: Germany
Posts: 6,000
Default

Quote:
Originally Posted by Glennbo View Post
Isn't true that USB2 is streaming packets, vs the now almost obsolete Firewire method of streaming data?
I understand that TB is more or less just serial PCI, and hence the protocol introduced by TB itself is close to zero. Hence there is not "Standard Sound" driver, nor a "Standard Packaging" behavior imposed, nut the driver needs to be dedicatedly crafted for the hardware and intended purpose. Hence it is dynamically possible to use small packets for low latency (e.g. with small buffer size setting and/or low channel count) or larger packets for improved average performance (e.g. with higher buffer size setting and/or higher channel count.

-Michael
__________________
www.boa-sorte.de
mschnell is online now   Reply With Quote
Old 10-10-2018, 10:20 PM   #37
Glennbo
Human being with feelings
 
Glennbo's Avatar
 
Join Date: Mar 2008
Location: Planet Earth
Posts: 2,576
Default

Quote:
Originally Posted by mschnell View Post
I understand that TB is more or less just serial PCI, and hence the protocol introduced by TB itself is close to zero. Hence there is not "Standard Sound" driver, nor a "Standard Packaging" behavior imposed, nut the driver needs to be dedicatedly crafted for the hardware and intended purpose. Hence it is dynamically possible to use small packets for low latency (e.g. with small buffer size setting and/or low channel count) or larger packets for improved average performance (e.g. with higher buffer size setting and/or higher channel count.

-Michael
Smaller chunks of data at higher speeds can be better for critical timing. Juggling larger chunks of data at a slower rate, timing becomes much more critical because a single late packet will contain a larger slice of playback time with a very obvious glitch if it happens. That said, it is all a balancing act and if the ratio of speed to byte/bit size is not done right, either method can be inefficient.

Reminds me of the one bit DACs of some CD players vs the more traditional 16 bit DACs that are more common. The one bit DACs were essentially reading the CD serially and at a higher rate.
__________________
Glennbo
Soundclick - http://www.soundclick.com/glennbo
--
Glennbo is offline   Reply With Quote
Old 10-10-2018, 11:24 PM   #38
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 22,083
Default

Quote:
Originally Posted by Taffer View Post


I don't know whether to congratulate you, or question your sanity. That's some next level contradiction the likes of which I will probably never again witness
Don't get me wrong, I appreciate you taking the time to provide advice, but I assume you have an explanation for the conflicting advice? I don't mind having a bunch of inputs or high end pre amps, I just don't want to pay significantly more for it. If a competent 2 io usb interface is 1000$, and a superior 8 io thunderbolt interface is 1000$, why would anyone buy the inferior product?

What's the contradicition you find? All the info I wrote is correct.

TB is better for large I/O count interfaces. USB2 works perfectly fine with well written drivers (-> RME) for up to 30 I/O or so, so there's no need to make the audio interface more expensive by having TB on it (because TB costs money for licensing the tech from Intel).

See two posts karbo posted after your post that I quoted. RME is "the shit", for good reason. So it's up to you to decide:

* how many I/O you need
* how opposed to USB you are
* do you want the best drivers in the biz or something else
EvilDragon is offline   Reply With Quote
Old 10-11-2018, 07:20 AM   #39
brainwreck
Human being with feelings
 
brainwreck's Avatar
 
Join Date: Jul 2006
Location: The 90's
Posts: 19,627
Default

Quote:
Originally Posted by Glennbo View Post
In my case, since I'm using the Linux version of REAPER pretty exclusively now, the M-Audio cards perform as low latency as I see running the Windows version of REAPER.

If I had to replace them, the RME HDSP 9632 would likely be my first choice due to it being a PCI device and running at PCI bus speed whether in Windows or Linux, just like the M-Audio cards.

RME's USB2 devices work in Linux, but not as well as they do in Windows, due to them only having class compliant drivers in Linux.

Edit:

BTW, it was Tim Ryan who started M-Audio, sometime after he started MidiMan, which was sometime after he wrote my favorite C64 midi sequencer "Studio One" by "Third Street Software" which was after he wrote my previous midi sequencer that was marketed by Sequential Circuits.

https://en.wikipedia.org/wiki/Tim_Ryan_(engineer)

OK, enough trivia, back to audio interfaces.
For linux, it seems that RME pci isn't going to have any benefit, the same as usb2. Which is a shame, since their drivers seem to be so good on windows (where anything lesser is a stability / performance gamble).

Those old maudio pci cards are pretty dang good for low cost devices on windows, if you have a machine with pci slots. I had a 1010lt at one point that both sounded good (noticably better than any usb or firewire device I have owned, even though having unbalanced connections) and performed well (sub 5 ms latency on windows xp). But these days, things have gone backward with audio devices. Devices costing the same and even much more don't sound as good and perform worse.
__________________
The media are misleading the public about...pretty much everything.
brainwreck is offline   Reply With Quote
Old 10-11-2018, 07:58 AM   #40
Glennbo
Human being with feelings
 
Glennbo's Avatar
 
Join Date: Mar 2008
Location: Planet Earth
Posts: 2,576
Default

Quote:
Originally Posted by brainwreck View Post
For linux, it seems that RME pci isn't going to have any benefit, the same as usb2. Which is a shame, since their drivers seem to be so good on windows (where anything lesser is a stability / performance gamble).

Those old maudio pci cards are pretty dang good for low cost devices on windows, if you have a machine with pci slots. I had a 1010lt at one point that both sounded good (noticably better than any usb or firewire device I have owned, even though having unbalanced connections) and performed well (sub 5 ms latency on windows xp). But these days, things have gone backward with audio devices. Devices costing the same and even much more don't sound as good and perform worse.
While it's true that there are no drivers written by RME for Linux, the difference between running a generic driver for a USB2 audio device vs a PCI bus connected device should be noticeable.

I'm not using a driver written by M-Audio for my 2496 PCI cards in Linux, yet I can run them at 64 samples latency (44.1kHz 24 bit 1.4/2.9ms) with no artifacts in the audio. I plan to build a new machine once there are Spectre-proof processors, and the Asus mobo I select will be one with at least two PCI slots so I can use my cards in yet a sixth machine since buying them back in the 90s.
__________________
Glennbo
Soundclick - http://www.soundclick.com/glennbo
--
Glennbo 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 12:40 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2018, vBulletin Solutions Inc.