The No.1 Website for Pro Audio
Audio Interface - Low Latency Performance Data Base
Old 1 week ago
  #4171
Gear Guru
 
monkeyxx's Avatar
Cool thanks everyone. It looks like the Quantum will run at a lower RTL, but the RME's HDSPe will handle more plugins, at similar buffer sizes. This kind of "evens out" in a way, since you could run the Quantum at 128, get a "similar" RTL as the RME at 64, and the plugin numbers are a little closer now. Just to give one example.

Think I will stick to my Presonus and consider a new PC build that would handle big sessions and video rendering better than my current PC.
Old 1 week ago
  #4172
Gear Guru
 
monkeyxx's Avatar
This thread inspired me to do a very non-scientific test of my audio system. Kind of a "real world" example.

My Win 10 PC uses an i7 4790K CPU running at 4.4 GHz clock speed. Interface is a Thunderbolt 2 Presonus Quantum. EDIT: forgot to mention than my standard project settings are 48 kHz, 32 bit float in Cubase 10.

I found that 64 buffer size is sort of a "sweet spot." Dropping to 32 samples can cause CPU overloads in a "heavy" session in Cubase. But it can also be stable at the beginning of a project for round trip tracking a guitar, bass or something. However, interestingly, increasing buffer to 128 or anything above that doesn't really seem to lower the CPU load bar at all, compared to 64 samples. There was no observable benefit to cranking up the buffer size on my system past 64.

Softube Tape is a relatively CPU "heavy" plugin. The Brainworx Console plugins are very light on CPU, I could insert dozens of them and the CPU load would climb very, very slowly.

Then I looked at my UAD Quad Satellite DSP load which was around 20-30%. I inserted 4 more UAD SSL E channels and the load on the Satellite quickly approached maximum. In comparison to the bx Console workflow, it was embarrassing, almost a shame if you want to go that far. The return on investment is not there at all. Every time I do this sort of test I get reminded why I don't spend money at UA any longer. For people with slow computers and inefficient interfaces, UAD might "seem" impressive. But to anyone with a capable native system, the UAD stuff is way, way behind in terms of efficiency.

Thought this was relevant enough to to the topic of this thread. No I don't have any hard numbers to offer right now. My point is not to slam UAD necessarily just a quick description of what my total system is capable of, or not capable of in a typical DAW session. My main take away is to leave the Quantum parked at the 64 buffer for now, with this computer.

Last edited by monkeyxx; 1 week ago at 04:15 PM..
Old 1 week ago
  #4173
Lives for gear
 
Jim Rosebrook's Avatar
 

Quote:
Originally Posted by monkeyxx View Post
The return on investment is not there at all. Every time I do this sort of test I get reminded why I don't spend money at UA any longer. For people with slow computers and inefficient interfaces, UAD might "seem" impressive. But to anyone with a capable native system, the UAD stuff is way, way behind in terms of efficiency.
Thanks for sharing his observation, which has been mine as well. I have multiple rigs.. some with UAD.. some with Quantum.... and the return on investment is clearly with the Quantum for me.. HUGELY so.
Old 1 week ago
  #4174
Lives for gear
 
daskeladden's Avatar
 

Quote:
Originally Posted by monkeyxx View Post

My Win 10 PC uses an i7 4790K CPU running at 4.4 GHz clock speed. Interface is a Thunderbolt 2 Presonus Quantum. EDIT: forgot to mention than my standard project settings are 48 kHz, 32 bit float in Cubase 10. My main take away is to leave the Quantum parked at the 64 buffer for now, with this computer.
On that setting Quantum is already down to RTL 3.17ms, so you cannot not expect that any system should work on heavy projects on that setting. On 48KHz on buffer size 32 you are down to RTL 2.50ms which is pretty much unusable on any projects and on any system. RTL 2.50ms can only be used recording live vocals/guitar with minimum of plugins and tracks or you will be punished by clicks and pops. Don't trust people that saying otherwise. As of today, technology (software/hardware) has it's limitations when it comes to RTL. And RTL will never be zero, but maybe in a few years you can work with big projects on RTL 1.5ms; one setting that does it all.
Old 1 week ago
  #4175
Gear Nut
 

Quote:
Originally Posted by Jim Rosebrook View Post
Thanks for sharing his observation, which has been mine as well. I have multiple rigs.. some with UAD.. some with Quantum.... and the return on investment is clearly with the Quantum for me.. HUGELY so.
i do love my old quantum 2 for what it does, which is simple and fast and that's all. helped my workflow immensely. i did replace it with an antelope D8 SC because i wanted to rack everything up, but it still resides on my pedalboard for when i'm using a hybrid analog/digital setup with an ABY box. i also use it in the rehearsal space for quick macgyver-style recordings with whatever mics happen to be in there, and you don't have to fuss with it at all and it's digitally controlled and fast and clean and for simple interface use, you can't really ask for much more.
Old 1 week ago
  #4176
Lives for gear
 

Quote:
Originally Posted by Christof View Post
The drummer: He needs tight timing for reference. But he will get that from the drumset itself, even without monitoring. A signal arriving later will not affect the "precedence effect" as long as the delay is not longer than the upper limit of the haas-area.
Sound plausible to me.


Quote:
The singer: The delay itself shouldn't be such a great problem for many singers. But here's why it DOES matter: Sound coming from the outside (inear or monitor speakers) will mix with the sound heard "internally" thus comb-filtering will be changing the sound of his own voice. He knows his voice very well and will recognize any deviation in sound quickly. Even a small delay can do quite bad things to the sound perceived!

Anecdotally I have observed this complaint from vocalists many times here on GS, particularly vocalists that mention wearing headphones. They talk about a strange, weird, phasey/flangey or hollow sound in their headphones.
Old 1 week ago
  #4177
Gear Guru
 
monkeyxx's Avatar
Quote:
Originally Posted by DanRand View Post


Anecdotally I have observed this complaint from vocalists many times here on GS, particularly vocalists that mention wearing headphones. They talk about a strange, weird, phasey/flangey or hollow sound in their headphones.
Some specific microphones are more prone to this effect than others. So it depends somewhat on the microphone being used, and obviously what the singer is hearing and reporting on.

I first heard this when watching a tutorial with a "top professional" but I forget which human it was. Got my attention, and I noticed it with the mic I use the most now, but not the mic I used previously.

I don't really like voice in my headphones anyway so most often I don't even turn it on.

You can play with polarity and phase to get the cue sounding "right" in the head of the singer.

I don't even think this is a super common occurrence, but it's something to be aware of, in case it comes up.
Old 1 week ago
  #4178
Gear Nut
 

The rtlutility link on the first page appears to broken.
Old 1 week ago
  #4179
Lives for gear
 

Quote:
Originally Posted by PeteJames View Post
I've been doing a bit of research into products like the ERM Multi-clock and perhaps having a reliable clock is more what I need rather than the absolute lowest latency?
The ERM Multi-clock will remove jitter from the MIDI clock & Din Sync, but won't completely stop the jitter coming from hardware synths and drum machines.

It can also help to make the alignment/synchronisation of clocks going to your synths easier because the ERM Multi-clock gets its timing from the DAW's audio and therefore has a more predictable timing/alignment relationship to your DAWs audio than a normal MIDI interface.

However, if your clocked hardware synths and drum machines sound in sync/aligned with your DAW's audio before or during recording, and aren't jittering too much for you, then all you need to do is configure your DAW properly for recording. There's a simple rule of thumb when recording external sound sources into a DAW. If an external sound source sounds in sync with your DAW's audio, before and during recording, then all you have to do is configure your DAW properly and your audio recordings will have the exactly same alignment to your DAW's audio when you play back the recording.

Quote:
Maybe the lowest latency 3ms RTL vs 10ms or so RTL won't help me that much?
As I said previously, I'm not that sensitive to latency. My current roundtrip is 10.8ms with a buffer size of 128 samples (plus MIDI in/out and synth MIDI response time latency or sync lag), and the audio recordings from my synths are all in sync when I play back those recordings. The timing relationship to my DAW's audio I hear when recording the audio from my external synths is the exactly the same when I play back the the recordings, to the sample. So, provided everything sounds in sync with my DAW's audio during recording, having highish latency doesn't affect the recording alignment of my synth's audio recordings because I've configured my DAWs correctly for recording external audio sources.

I would prefer a roundtrip of around 7ms so that I could use hardware inserts (External Audio Effect) when composing, but I don't feel the need to go any lower than that.

Quote:
I would have imagined buying an RME UFX+ should be perfect for midi timing and midi clock sync with no jitter etc when spending that kind of cash.
RME interfaces do generally have good MIDI timing. Unfortunately I need more MIDI outputs so need some kind of dedicated MIDI interface.

Quote:
Is DAW triggered midi audio not lining up with the gird when recording to your DAW or live jamming with sequencers a sync issue not a latency issue?
As I've tried to stress above, if an external sound source (not just synths) sound in sync with your DAW's audio before/during recording then all you have to do is configure your DAW properly for recording. (again, see here if recording software monitored synths through Ableton)

Where lower latency can help is by making it easier for you to keep your live keyboard playing in sync with your DAW's audio in the first place.

But for pre-recorded MIDI clips and midi clock then higher latency generally makes no difference unless you've configured your DAW incorrectly.

If you haven't correctly configured your DAW (i.e. your MIDI notes or MIDI clock are firing at the wrong time, or you've incorrectly configured recordings in your DAW), then lower RTL will mean the misalignment with your DAW's grid will be smaller.

Hope some of this helps! I'm off to cut the bloody grass. And not the good kind
Old 1 week ago
  #4180
Lives for gear
 
PeteJames's Avatar
Quote:
Originally Posted by DanRand View Post
The ERM Multi-clock will remove jitter from the MIDI clock & Din Sync, but won't completely stop the jitter coming from hardware synths and drum machines.

It can also help to make the alignment/synchronisation of clocks going to your synths easier because the ERM Multi-clock gets its timing from the DAW's audio and therefore has a more predictable timing/alignment relationship to your DAWs audio than a normal MIDI interface.

However, if your clocked hardware synths and drum machines sound in sync/aligned with your DAW's audio before or during recording, and aren't jittering too much for you, then all you need to do is configure your DAW properly for recording. There's a simple rule of thumb when recording external sound sources into a DAW. If an external sound source sounds in sync with your DAW's audio, before and during recording, then all you have to do is configure your DAW properly and your audio recordings will have the exactly same alignment to your DAW's audio when you play back the recording.



As I said previously, I'm not that sensitive to latency. My current roundtrip is 10.8ms with a buffer size of 128 samples (plus MIDI in/out and synth MIDI response time latency or sync lag), and the audio recordings from my synths are all in sync when I play back those recordings. The timing relationship to my DAW's audio I hear when recording the audio from my external synths is the exactly the same when I play back the the recordings, to the sample. So, provided everything sounds in sync with my DAW's audio during recording, having highish latency doesn't affect the recording alignment of my synth's audio recordings because I've configured my DAWs correctly for recording external audio sources.

I would prefer a roundtrip of around 7ms so that I could use hardware inserts (External Audio Effect) when composing, but I don't feel the need to go any lower than that.



RME interfaces do generally have good MIDI timing. Unfortunately I need more MIDI outputs so need some kind of dedicated MIDI interface.



As I've tried to stress above, if an external sound source (not just synths) sound in sync with your DAW's audio before/during recording then all you have to do is configure your DAW properly for recording. (again, see here if recording software monitored synths through Ableton)

Where lower latency can help is by making it easier for you to keep your live keyboard playing in sync with your DAW's audio in the first place.

But for pre-recorded MIDI clips and midi clock then higher latency generally makes no difference unless you've configured your DAW incorrectly.

If you haven't correctly configured your DAW (i.e. your MIDI notes or MIDI clock are firing at the wrong time, or you've incorrectly configured recordings in your DAW), then lower RTL will mean the misalignment with your DAW's grid will be smaller.

Hope some of this helps! I'm off to cut the bloody grass. And not the good kind
Thanks very much, that's definitely the most helpful explanation i've ever had. So really the RTL only matters for live playing - the rest is just down to synchronisation and compensation. Therefore it's not necessarily easier to compensate for ultra low latency than medium latency - it's just a case of upping the compensation? I'm not really sure how to do that though. I'm familiar with Ableton Live Preferences/Audio/Latency settings but really don't know how to set them up properly.
Old 1 week ago
  #4181
Lives for gear
 
PeteJames's Avatar
Quote:
Originally Posted by DeadPoet View Post
AMD Ryzen 3950x as mentioned (briefly) in my post.


When using only softsynths I never have the need to up my buffer size (for now). When I'm mixing (using multiple parallel paths and well over 40 channels of outboard on both inserts as sends) things get ugly real fast.


I don't use midi a lot in the hardware realm (at the moment I only have one Behringer Model D but hope to expand that in the future) so I can't comment on that.
Playing through a master midi keyboard (on its own usb connection) never has gotten me problems.



Herwig
Have you got any examples of the softsynths you're using and track counts a 32 buffer size? Doesn't sound like daskeladden is convinced.
Old 1 week ago
  #4182
Lives for gear
 

Quote:
Originally Posted by DeadPoet View Post
Define not realistic? I just did a mix with over 80 tracks (mostly stereo tracks, very few mono).
All of which have Kazrog TrueIron, Softube Tape and Console1 on them.
A few parallell mixbus-routes, a few plugins and hardware on the 2buss, about 8 FX sends.

48k/32float, Win10, CubasePro10.5, 3950x, RME HDSP MADIFx.


32 buffer size and not sweating for a nanosecond.



Writing sessions (with a dozen or so modern/present-day VSTi's) have the same scenario. When come mixtime I bounce my midi to audio because I'm old-skool like that.


32 buffer size is a very real world scenario for me ever since I got that CPU.



Herwig
Interesting, I have a 3950x and if I have a project with say 60 channels, say 30 of those tracks have around 4 plugins on average on each track, for examples (pro q x 2, black box, a compressor or an eventide reverb, a valhalla delay) my comp completely breaks down at 32 buffer.

I just tried it out and my ableton cpu is up at 80% and pops and crackles all over the place. I am using a terrible UR44 interface and driver but surely your RME is not making that much difference in performance.

256 samples, cpu is around 40%. PC deals with it fine.
Old 1 week ago
  #4183
Lives for gear
 
PeteJames's Avatar
Quote:
Originally Posted by Primativ View Post
Interesting, I have a 3950x and if I have a project with say 60 channels, say 30 of those tracks have around 4 plugins on average on each track, for examples (pro q x 2, black box, a compressor or an eventide reverb, a valhalla delay) my comp completely breaks down at 32 buffer.

I just tried it out and my ableton cpu is up at 80% and pops and crackles all over the place. I am using a terrible UR44 interface and driver but surely your RME is not making that much difference in performance.

256 samples, cpu is around 40%. PC deals with it fine.
What happens if you get rid of the other 30 tracks and just keep the 30 tracks with 4 plugins? That's probably all I need really but I would be using soft synths like Diva / Repro 5 and Serum plus a few effects on perhaps 10 of the tracks.
Old 1 week ago
  #4184
Gear Addict
Quote:
Originally Posted by PeteJames View Post
Thanks very much, that's definitely the most helpful explanation i've ever had. So really the RTL only matters for live playing - the rest is just down to synchronisation and compensation. Therefore it's not necessarily easier to compensate for ultra low latency than medium latency - it's just a case of upping the compensation? I'm not really sure how to do that though. I'm familiar with Ableton Live Preferences/Audio/Latency settings but really don't know how to set them up properly.
There's on exception and that is that Ableton does not delay compensate the midi clock. So when using high latency plug-ins (e.g. Fabfilter Q3 with natural phase mode) and clocking external synths or drum machines you will run into problems
Old 1 week ago
  #4185
Lives for gear
 

^ Yes, that's true, but that's additional PDC/plugin latency, not the roundtrip latency determined by the combination of audio interface and DAW buffer size.

PeteJames is trying to decide which audio interface to buy and how low he needs to go on the RTL front.
Old 6 days ago
  #4186
Lives for gear
 

Quote:
Originally Posted by PeteJames View Post
What happens if you get rid of the other 30 tracks and just keep the 30 tracks with 4 plugins? That's probably all I need really but I would be using soft synths like Diva / Repro 5 and Serum plus a few effects on perhaps 10 of the tracks.
Ok so I ran an expirement for you. 32 buffer, 48k sample rate, UR44 RTL is 6ms.

37 channels. Average plugin count is around 4 or 5 per channel. 4 instances of UA Lion. 1 instance of fabfilter one. I also opened a couple of extra instances of Serum.

CPU is hovering around 48%.

No pops or crackles at all. Runs smoothly.

You can only improve performance from here with a bettter driver and interface.

Last edited by Primativ; 6 days ago at 02:22 PM..
Old 6 days ago
  #4187
Lives for gear
 
DeadPoet's Avatar
Quote:
Originally Posted by PeteJames View Post
Have you got any examples of the softsynths you're using and track counts a 32 buffer size? Doesn't sound like daskeladden is convinced.
I can do something tomorrow. I remember loading up an 8-channel multi in Omnisphere and copying that 24 times? The ASIO meter was at 20% or so.
I'll get some numbers running tomorrow if you want to.


Herwig
Old 6 days ago
  #4188
Lives for gear
 
daskeladden's Avatar
 

32 buffer size isn't that impressive as long as RTL is over 3ms. I can run almost anything when RTL is over 5ms. RTL is "all" that matters when recording live
Old 6 days ago
  #4189
Lives for gear
 
PeteJames's Avatar
I'm really struggling to find a post i'm sure i've come across before that shows every audio interface tested so far thunderbolt and usb in this detailed format. Anyone know where it is?
Attached Thumbnails
Audio Interface - Low Latency Performance Data Base-screenshot-1181-.jpg  
Old 6 days ago
  #4190
Lives for gear
 
PeteJames's Avatar
Quote:
Originally Posted by DeadPoet View Post
I can do something tomorrow. I remember loading up an 8-channel multi in Omnisphere and copying that 24 times? The ASIO meter was at 20% or so.
I'll get some numbers running tomorrow if you want to.


Herwig
Yes that would be really cool. I'm curious about ram usage too in relation to Omnisphere. I'm most familiar with the CPU usage of DIVA / Repro 5 and Serum. As well as an audio interface i'm looking for either the new 16 inch macbook pro 8 core or a desktop PC with a AMD beast like yours or Intel equivalent. I've always owned laptops really and am used to apple when it comes to audio but some of these desktop PCs are putting out double the performance. Whether it's complete overkill of me is another matter but if I was able to get latency that low in my workable projects that would be amazing.
@ Primativ - Thanks for going to the effort to test. I was really hoping to see would could be accomplished at more of a 2-3 second RTL.

Last edited by PeteJames; 6 days ago at 11:23 PM..
Old 6 days ago
  #4191
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by PeteJames View Post
I'm really struggling to find a post i'm sure i've come across before that shows every audio interface tested so far thunderbolt and usb in this detailed format. Anyone know where it is?
Its supposed to be listed on the first post and all the subsequent latest updated posts, but looks like GS is again blocking it because its too large.

Find it Here

I'll message the admins and see if I can get the picture size limit lifted again.

Old 6 days ago
  #4192
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by PeteJames View Post
Yes that would be really cool. I'm curious about ram usage too in relation to Omnisphere. I'm most familiar with the CPU usage of DIVA / Repro 5 and Serum.
Just remember that Cubase uses a hybrid playback engine , so the 32 sample buffer doesn't come into play unless you are track armed.

Also take note that in Cubase 10/10.5 any time the realtime engine is in play, it can trigger some threading issues as it kicks in the MMCSS thread limiting imposed on those versions. I have covered that extensively in a dedicated thread.

Lets not get too sidetracked with memory/CPU usage on this thread, as there are dedicated threads that cover those areas specifically.

Old 6 days ago
  #4193
Lives for gear
 

Quote:
Originally Posted by PeteJames View Post
Yes that would be really cool. I'm curious about ram usage too in relation to Omnisphere. I'm most familiar with the CPU usage of DIVA / Repro 5 and Serum. As well as an audio interface i'm looking for either the new 16 inch macbook pro 8 core or a desktop PC with a AMD beast like yours or Intel equivalent. I've always owned laptops really and am used to apple when it comes to audio but some of these desktop PCs are putting out double the performance. Whether it's complete overkill of me is another matter but if I was able to get latency that low in my workable projects that would be amazing.
@ Primativ - Thanks for going to the effort to test. I was really hoping to see would could be accomplished at more of a 2-3 second RTL.
No worries. Unfortunately the UR44 cannot reach that low RTL. I will try it when I pick up a new RME, if of course you are still interested by then.
Old 6 days ago
  #4194
Gear Head
 

Quote:
Originally Posted by pangea2003 View Post
That's why is so important constructive criticism from user experience

Otherwise, what else makes you think Presonus switched the 'nebulous marketing' for an actual product that actually performs well now?

Are you listening Antelope??

Now, I don't believe that the quantum will stop working with Studio One.. and surely Studio One won't get discontinued anytime soon..

what makes me think..

Antelope is actually the number one in releasing and DISCONTINUING products very quickly.. right??
I think Focusrite had headed that direction which is why the Clarett was my last Focusrite interface. I just got the Motu 828ES and never looked back. I've owned the first generation Scarlett 18i20, The Forte and the Clarett 8Pre and all of them I experienced flaky drivers. Both Forte and Scarlett gave me the BSOD when using Studio One of O bump the USB cable or accident unplug the hardware. Focusrite biggest pet peeve is to release a new product and discontinue and abandon support for them rendering them obsolete. They have such a short development life cycle such as 3 versions of Scarletts released with in the past 6 years. The Clarett Thunderbolt only had its run for 4 years and now its discounted and abandoned driver support. I had enough of that and just stopped buying Focusrite interfaces all together when RME, Lynx and Motu still make drivers for its products that are 10+ years old. Focusrite likes to take forever when it comes to releasing new drivers, many of them stuck in beta or completely abandoned them. I did however notice a signficant better latency performance with my 828ES and the drivers are very rock solid since Motu has came a long ways. I remembered their legacy products had some driver issues with windows but certainly isn't the case with its current AVB line up.
Old 6 days ago
  #4195
Lives for gear
 
PeteJames's Avatar
Quote:
Originally Posted by Primativ View Post
No worries. Unfortunately the UR44 cannot reach that low RTL. I will try it when I pick up a new RME, if of course you are still interested by then.
I'm trying to nail down a final decision this week. I might just go for broke and get the UFX+ because you just can't go wrong with it and if it does me 10+ years without any hassle i'll be a happy camper. I see the Presonus as more of a risk or possibly short term solution that I may regret. Could be wrong though - could turn out to be an awesome decision and give me more money to spend on synths! Hopefully, I'll have chosen a powerhouse PC for the studio this week too.

Thanks for the link TAFKAT
Old 4 days ago
  #4196
Here for the gear
 

Would love to see where the new gen Lynx Aurora (n) sits on the list. Maybe it has the legs to replace the king of the hill?
Old 3 days ago
  #4197
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by Uberphat View Post
Would love to see where the new gen Lynx Aurora (n) sits on the list. Maybe it has the legs to replace the king of the hill?
Lynx TB has already been tested and reported on, the driver is the same across the PCIe/TB product lines.

The only variance over the original Aurora would be if the AD/DA has lower latency.

The original Aurora had low latency already , so even if was lower on the (n) it wouldn't be enough to take the reference position as the overall efficiency/results on the benchmarks at the respective latencies is not there.

Old 6 hours ago
  #4198
Lives for gear
 
Marando's Avatar
Thank you Tafkat and contributors. It's because of this thread that I bought a RME Babyface Pro FS. It's a gigantic upgrade over the Apollo Twin Duo USB with regards to latency/ASIO performance, seriously insane. I also believe to hear a subtle upgrade in audio quality too, as if there is a little bit more clarity, if that makes sense.

Suffice to say that the Babyface is fantastic and I shouldn't have hesitated to buy it.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump