The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Latency with my Interface, get another one?
Old 10th July 2016
  #1
Gear Head
 

Latency with my Interface, get another one?

Hi i bought a Focusrite 18i2o interface just recently

1. with an empty project samplerate 30, i feel like i have very subtle latency, the monitoring doesnt sound like when i playback the recording, is that okay?

2.bigger problem -> with the highest samplerate, which is 1024, i get crackls in the sound if I load "Ozone 6" 13 times onto an audio track. Its a different track, not the one im recording on.

is this a problem with my PC?
Would a RME Interface change anything?

thank you i so appreciate ur knowledge !!

my PC : WIN8 64x 16GB RAM, 1TB SSD, IntelCore i7-4500U @ 1.80GHz 2.40GHz
DAW: BITWIG/ABLETON (same problem)
RAM : 8GB used
CPU : 51-77% not hitting 100 !!

Ive read some other threads but since my cpu isnt hitting 100 im confused

Thanks Again !!!!
Old 10th July 2016
  #2
Lives for gear
 
JulenJVM's Avatar
Hello,

I think you are confusing sample rate with buffer size here. The bigger the buffer size, the more time your CPU has to process audio, which means that you can load more plugins. But it's normal to get crackles with 13 instances of Ozone, its a plugin pretty heavy on CPU. Your ASIO load is maxed out, not your CPU.

Also, if you are monitoring from the DAW, it's normal to get roundtrip latency. You need to monitor yourself from the interface directly, before the signal enters the DAW. Make sure to turn up the fader in Scarlett Control of the input used to record, to feed the signal to your mix output.

Hope this helps!
Old 10th July 2016
  #3
Gear Head
 

Okay !! Thx so much for ur help

So to sum it up. This latency is normal. RME would not change anything major.

i just have to use direct monitoring, "light weight" VSTS, and bounce finished tracks to audio ...

correct sir?
Old 10th July 2016
  #4
Lives for gear
 
JulenJVM's Avatar
Sounds good, although you won't find yourself very usually using more than 1 Ozone plugin at a time - that one is meant for mastering mainly. You should be able to load many many more regular plugins.

And yes, use direct monitoring for tracking, it's the only way of getting rid of latency
Old 10th July 2016
  #5
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by rat010104 View Post
Would a RME Interface change anything?
Probably better latency and better performance. Focusrite don't have a very good reputation on this front while RME are pretty much the gold standard when it comes to this stuff.

You will probably still run out of juice when trying to run 13 instances of Ozone on one track though. It is a heavy plugin and most DAWs will run all the plugins for a single track on the same core. In this scenario things are not being balanced between the cores. Luckily it isn't normal use when making music.

Alistair
Old 11th July 2016
  #6
Lives for gear
 
s12512's Avatar
What do you need 13 instances of ozone for???
Old 11th July 2016
  #7
Gear Head
 

The ozone thing was just a test :D i dont need that many of them ...

so thank u everybody for explaining the plot to me

i think im gonna stay with the focusrite, because its 1000 € less, and i feel its a computer thing anyway, focusrite is not calculating to slow, its my computer !!!

have a nice week everybody !!!
Old 11th July 2016
  #8
Lives for gear
 
JulenJVM's Avatar
Don´t underestimate UnderTow´s advice on the optimization of the interface drivers, they have a direct impact on the ASIO load that your CPU can handle. And ASIO Load is what determines how many plugins you can run at a time, not your CPU. I also have a 1st gen Scarlett 18i20, and I´m starting to explore other USB interfaces with better drivers and performance that might improve my ASIO Load. In this case I´d keep the Scarlett as a slave device just for the extra analogue inputs.

RME Babyface Pro
Old 11th July 2016
  #9
Lives for gear
 

The CPU is fairly low powered, so your overhead per core isn't all that high. If you throw something intensive on there like Ozone then it's going to max them out fairly quickly. Due to the way larger plug in's balance, it's fairly common to see ASIO crap out around 80 - 90% of the real CPU overhead even on powerful setups, so what your seeing there isn't all that surprising given its a mobile dual core.
Old 11th July 2016
  #10
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by JulenJVM View Post
Don´t underestimate UnderTow´s advice on the optimization of the interface drivers, they have a direct impact on the ASIO load that your CPU can handle. And ASIO Load is what determines how many plugins you can run at a time, not your CPU. I also have a 1st gen Scarlett 18i20, and I´m starting to explore other USB interfaces with better drivers and performance that might improve my ASIO Load. In this case I´d keep the Scarlett as a slave device just for the extra analogue inputs.

RME Babyface Pro
Indeed. Here are two extreme cases to demonstrate the differences. Two different interfaces hooked-up to the exact same computer running the exact same DAW and both set to a 64 samples buffer (This Focusrite can't do 32 samples):

The Focusrite Saffire 6 USB has a roundtrip latency of 9.887ms and can run 68 instances of the Reaper multiband compressor.
The RME HDSPe AES has a roundtrip latency of 4.4ms and can run 149 instances of the Reaper multiband compressor.

That is half the latency and more than double the plugins on the exact same computer with the exact same settings. You could set the RME to a 128 sample buffer (Round trip of 7.7 ms) and run even more plugins and still have lower latency than the Saffire 6. That is how much the interface connection, drivers and firmware affect the performance of your DAW.

Alistair
Old 11th July 2016
  #11
Lives for gear
 
JulenJVM's Avatar
Thanks Alistair.

I've been checking on the Low latency database thread, there's lots of interesting info in there, and a lot of work put in it. By looking the tests, it looks like RME interfaces are the ones that deliver the better performance over USB in Windows. If you don't mind me asking, what parameters should I be looking at when trying to determine which one can handle a higher plugin count when mixing? I'm not so worried about RTL as I monitor directly from the interface and rarely record VSTis live. It's maxing out the CPU what matters the most to me.

Thank you in advance!
Old 11th July 2016
  #12
Lives for gear
 

@ JulenJVM

My recommended approach is to check Vin's page, where is summary of the thread.
DAW Bench : DAW Performance Benchmarking
So for example..
pick some suitable latency limit for you.. not buffer size, but real roundtrip or output latency.. say 10ms. This is kind of worst acceptable RTL latency, you can use for tracking or performance through computer.
Then look for real-world RTL values and find the interfaces which will fit into that limit.. say one interface would have 64s buffer, another with better performance can have 128s and still there will be some some delay margin.
Then look for columns with RXC (count of Reaper's multiband compressor), CV (count of Kontakt instances with convolution reverb - this is heaviest and most demanding test in DAWBench) and NCV (count of Kontakt instances without convolution reverb, which is lighter).
You'll see, why RME or Lynx is so praised.. because there can be very significant differences.

Michal
Old 12th July 2016
  #13
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by msmucr View Post
@ JulenJVM

My recommended approach is to check Vin's page, where is summary of the thread.
DAW Bench : DAW Performance Benchmarking
So for example..
pick some suitable latency limit for you.. not buffer size, but real roundtrip or output latency.. say 10ms. This is kind of worst acceptable RTL latency, you can use for tracking or performance through computer.
Then look for real-world RTL values and find the interfaces which will fit into that limit.. say one interface would have 64s buffer, another with better performance can have 128s and still there will be some some delay margin.
Then look for columns with RXC (count of Reaper's multiband compressor), CV (count of Kontakt instances with convolution reverb - this is heaviest and most demanding test in DAWBench) and NCV (count of Kontakt instances without convolution reverb, which is lighter).
You'll see, why RME or Lynx is so praised.. because there can be very significant differences.

Michal
Hello Michal,

Thanks a lot for the info, the numbers finally make sense. RME looks like THE contender on this one, but what is puzzling me right now, is that most of RME USB interfaces provide more or less the same performance (with 1-2 ms of RTL difference).

In my case, I have a Focusrite Scarlett 18i20, so I have my input tracks count, preamps and direct monitoring solved with that, but performance-wise, RMEs look like they are in a different universe. UFX, UCX and Babyface provide a very similar performance, but at different prices. Now I need to find out if performance wise, it´s worth investing the extra money on a Fireface UCX, or go for the cheaper one, Babyface Pro.

Anyway, I just want to say thanks to everyone here, the advice provided has opened my eyes!
Old 12th July 2016
  #14
Lives for gear
 

Quote:
Originally Posted by JulenJVM View Post
In my case, I have a Focusrite Scarlett 18i20, so I have my input tracks count, preamps and direct monitoring solved with that, but performance-wise, RMEs look like they are in a different universe. UFX, UCX and Babyface provide a very similar performance, but at different prices.
Focusrite use a XMOS controller for the USB transfer, as do a number of other firms I may add. In fact at this point it's a widely adopted off the shelf solution often leveraged by many firms to keep the cost of the units down.

RME design their controller around a FPGA chip and they program that from the ground up, allowing them to tweak the ever living daylights out of it.

The costs involved in maintaining and supporting a full R&D code team is obviously far higher than buying in a solution, which is pretty much where you money goes when paying out the RME premium. As you'll note by the RTL scores however, it's pretty clear where that money is going!
Old 12th July 2016
  #15
Lives for gear
 

Quote:
Originally Posted by JulenJVM View Post
...
Thanks a lot for the info, the numbers finally make sense. RME looks like THE contender on this one, but what is puzzling me right now, is that most of RME USB interfaces provide more or less the same performance (with 1-2 ms of RTL difference).

In my case, I have a Focusrite Scarlett 18i20, so I have my input tracks count, preamps and direct monitoring solved with that, but performance-wise, RMEs look like they are in a different universe. UFX, UCX and Babyface provide a very similar performance, but at different prices. Now I need to find out if performance wise, it´s worth investing the extra money on a Fireface UCX, or go for the cheaper one, Babyface Pro.
...
It depends what you exactly plan to do.. Babyface and UCX are quite different interfaces.. One is bus powered compact interface, while UCX has more of everything, better preamps and size isn't main thing.
Also do you plan to keep 18i20? If yes, then it can be used as ADAT converter for another audio interface, because it can be used in standalone mode.

That way you can use not only USB interfaces, but for example also RME HDSPe AIO or even used RME 9632 (if you have PCI slot), which are internal cards with ADAT I/O with really good performance.. and 18i20 can be connected there.

Michal
Old 12th July 2016
  #16
Lives for gear
 
JulenJVM's Avatar
Hi Michal,

Thanks a lot again for the replies. At this point, I´m keeping the Focusrite for recording mainly, and I want something with more juice to deal with the audio on the laptop. Unfortunately, the only option for me is a USB interface as I work on a laptop, so the PCIe cards are not an option. I want the RME as main interface, and the Focusrite slaved to it via ADAT. I´m trying to decide if I need the extra preamps that the UCX or UFX provide, but so far it is not a must. My only priority is improving the ASIO load that my laptop can handle, that´s why I´m also looking at the Babyface. There´s a 400€ difference between the Babyface and UCX, and I have enough I/Os as it is, so I´m trying to decide if I should invest those 400€ of difference somewhere else, like a microphone, or a few extra plugins....

Decisions, decisions, decisions
Old 12th July 2016
  #17
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by Pete Kaine View Post
Focusrite use a XMOS controller for the USB transfer, as do a number of other firms I may add. In fact at this point it's a widely adopted off the shelf solution often leveraged by many firms to keep the cost of the units down.

RME design their controller around a FPGA chip and they program that from the ground up, allowing them to tweak the ever living daylights out of it.

The costs involved in maintaining and supporting a full R&D code team is obviously far higher than buying in a solution, which is pretty much where you money goes when paying out the RME premium. As you'll note by the RTL scores however, it's pretty clear where that money is going!
Hi Pete,

Yep, the numbers add up! The performance is up there, so no doubts that it´s money well invested. Just need to decide if I invest a little bit or a bit more
Old 12th July 2016
  #18
Gear Addict
 

Quote:
Originally Posted by JulenJVM View Post
Hi Michal,

Thanks a lot again for the replies. At this point, I´m keeping the Focusrite for recording mainly, and I want something with more juice to deal with the audio on the laptop. Unfortunately, the only option for me is a USB interface as I work on a laptop, so the PCIe cards are not an option. I want the RME as main interface, and the Focusrite slaved to it via ADAT. I´m trying to decide if I need the extra preamps that the UCX or UFX provide, but so far it is not a must. My only priority is improving the ASIO load that my laptop can handle, that´s why I´m also looking at the Babyface. There´s a 400€ difference between the Babyface and UCX, and I have enough I/Os as it is, so I´m trying to decide if I should invest those 400€ of difference somewhere else, like a microphone, or a few extra plugins....

Decisions, decisions, decisions
At larger buffer sizes, the difference in the number of plugins you can run, RME vs Focusrite, is not as pronounced as it is at the smaller buffer sizes.

If you want more processing power, put the money into a new machine. That will get you much more than a switch to RME.
Old 12th July 2016
  #19
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by GoldDragon View Post
At larger buffer sizes, the difference in the number of plugins you can run, RME vs Focusrite, is not as pronounced as it is at the smaller buffer sizes.

If you want more processing power, put the money into a new machine. That will get you much more than a switch to RME.
The thing is that I'm working on a mix where I'm starting to max out on ASIO load, but not on CPU (don't have the exact figures right now), and I'm starting to wonder if the Focusrite is the bottleneck. I also want to explore Nebula, so I need to optimize my system. That's on Cubase 8.5, Win 8.1, laptop with i7 4790 8 GB RAM
Old 12th July 2016
  #20
Gear Addict
 

Quote:
Originally Posted by JulenJVM View Post
The thing is that I'm working on a mix where I'm starting to max out on ASIO load, but not on CPU (don't have the exact figures right now), and I'm starting to wonder if the Focusrite is the bottleneck. I also want to explore Nebula, so I need to optimize my system. That's on Cubase 8.5, Win 8.1, laptop with i7 4790 8 GB RAM
Increase the buffer 1024 or larger. RME is not a magic bullet to make your system process more plugins.
Old 13th July 2016
  #21
Lives for gear
 

Quote:
Originally Posted by JulenJVM View Post
The thing is that I'm working on a mix where I'm starting to max out on ASIO load, but not on CPU (don't have the exact figures right now), and I'm starting to wonder if the Focusrite is the bottleneck
Done the tweaks recommended for your OS?

https://support.focusrite.com/hc/en-...o-on-Windows-7
https://support.focusrite.com/hc/en-...-on-Windows-10

The high performance power tweak is quite crucial for stopping it maxing out the CPU at lower usage levels, if you've not applied it already, it may help here.

Other than that, it could be a badly optimized plug in hogging a CPU core and making it fall over earlier than it should. Worth checking through your plugins with the aim of seeing if removing one helps it balance. It might prove as simple as updating an older plugin build in order to reduce CPU load. *glances sideways at DIVA and Serum*
Old 13th July 2016
  #22
Lives for gear
 
JulenJVM's Avatar
Quote:
Originally Posted by Pete Kaine View Post
Done the tweaks recommended for your OS?

https://support.focusrite.com/hc/en-...o-on-Windows-7
https://support.focusrite.com/hc/en-...-on-Windows-10

The high performance power tweak is quite crucial for stopping it maxing out the CPU at lower usage levels, if you've not applied it already, it may help here.

Other than that, it could be a badly optimized plug in hogging a CPU core and making it fall over earlier than it should. Worth checking through your plugins with the aim of seeing if removing one helps it balance. It might prove as simple as updating an older plugin build in order to reduce CPU load. *glances sideways at DIVA and Serum*
Hi Pete,

That´s very good advice, thank you. I´ll try those tweaks and then check for misbehaving plugins, but I don´t think it´s a particular one...but I´ll keep an eye.

As for the buffer size, I was already using the largest one allowed by the interface: 20 ms or roughly 960 samples @ 48 kHz. CPU reads constant at 38% and Memory Consumption is at 40%. So I don´t think the processor is the bottleneck....

Off to tweak Windoze!
Old 13th July 2016
  #23
Lives for gear
 
JulenJVM's Avatar
@ Pete Kaine

Actually, I have to eat my words, you had a very good point with the plugins! It turns out el cabrón was the trial version of Pink 2412, it was overloading a core. I have just downloaded the new released version, and things went back to normal. I was able to use many instances of the plugin!

I also checked the Windows tweaks, and I had done most of them. I was surprised to find a "Steinberg Power Scheme" which I changed for the High Performance one, and I need to check if I need to fiddle with hyperthreading or not.

All in all, despite having improved a lot the situation, I still get to max out the ASIO load while the CPU & Memory hover around 55%, so there´s a lot of room for improvement. Off to investigate some more
Old 13th July 2016
  #24
Lives for gear
 
JulenJVM's Avatar
Actually, now that I think of it, 20 ms - 960 samples is too small of a buffer size! And that's the maximum allowed by the Scarlett - unless I've missed something....I could do with a larger buffer size!
Old 13th July 2016
  #25
Lives for gear
 

Quote:
Originally Posted by JulenJVM View Post
Actually, now that I think of it, 20 ms - 960 samples is too small of a buffer size! And that's the maximum allowed by the Scarlett - unless I've missed something....I could do with a larger buffer size!
Hmm, usually it's not really necessary to go over 512s, as you usually don't gain too much going higher.. In some cases and setups, I've actually experienced point of diminishing return, where longer buffer was worse.

All this of unfortunately depends on DAW (it's scheduling and efficiency), project (what is routed where and how many serial plugins are at one track - this is btw. one of most common mistake.. user place too many heavy plugins to buses or master, where all runs at one core), particular used plugins and instruments (some plugins has also internal settings for chunk size used to processing data from DAW, like iZotope).. So there are some generic things, which you can apply, but it's really necessary to experiment for finding sweetspot.

IMHO, if DAW performance with real world projects after some tuning and optimization won't be acceptable with 512s ASIO buffer, then it's more matter of CPU which can't cope with particular workload and further fiddling with audio interface doesn't change that.. (this has much more significant effect at latency sensitive workloads with shorter buffers)

Nebula or Aqua based effects are still one of most demanding out there and most of its fans and heavy users here seems to be either using pretty powerful machines (like 4970k or 5820k) or possibly freezing.
I haven't tried its latest updated versions versions with multithreading Core optimization, but I can imagine.. its performance can vary a bit among different DAWs and buffer sizes.
In some cases..when some plugin (I'm not talking about Nebula/Aqua in particular) is doing it's own multi-threading (besides typical fashion with one processing and one UI thread per plugin), it can sometimes "fight" with plugin scheduling at DAW and quite counterintuitively make situation bit worse. So some experimenting is really necessary to avoid mentioned choking of one core, which leads to dropouts, although overall load doesn't necessarily indicate some processing issue.

Michal
Old 13th July 2016
  #26
Tui
Gear Guru
 
Tui's Avatar
Quote:
Originally Posted by Pete Kaine View Post
The costs involved in maintaining and supporting a full R&D code team is obviously far higher than buying in a solution, which is pretty much where you money goes when paying out the RME premium. As you'll note by the RTL scores however, it's pretty clear where that money is going!
And ongoing support, I might add. I can still get fresh drivers for my 15-year old Digiface.
Old 14th July 2016
  #27
Lives for gear
 

Quote:
Originally Posted by JulenJVM View Post
It turns out el cabrón was the trial version of Pink 2412, it was overloading a core. I have just downloaded the new released version, and things went back to normal. I was able to use many instances of the plugin!
Acustica plug ins are insanely power hungry, it's pretty much all I get phone calls about at the moment! Sound amazing mind, so worthwhile trade off I guess

Quote:
Originally Posted by JulenJVM View Post
I also checked the Windows tweaks, and I had done most of them. I was surprised to find a "Steinberg Power Scheme" which I changed for the High Performance one, and I need to check if I need to fiddle with hyperthreading or not.
Steinbergs power scheme is fairly close to the high performance one, they do the essential tweaks with it to ensure Cubase is running properly, certainly shouldn't be any issues using that one.

Quote:
Originally Posted by JulenJVM View Post
Actually, now that I think of it, 20 ms - 960 samples is too small of a buffer size! And that's the maximum allowed by the Scarlett - unless I've missed something....I could do with a larger buffer size!
Scary numbers, almost sounds like your using WASPI not ASIO. I'm running with a 6.5ms / 128 buffer and even then I sometimes complain it's too laggy!

Quote:
Originally Posted by Tui View Post
And ongoing support, I might add. I can still get fresh drivers for my 15-year old Digiface.
For all the interfaces I keep floating around the studio, it's my RME 9632 from well over a decade ago that remains the daily workhorse. It's last driver update was around the start of the year, so can't really complain for value for money when all that's considered.
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…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump