HD Native and Latency/Softsynths

[mykhal c]

IIRC low latency mode will disable your plugs...at least on non-HD native PT it does. and i'm pretty sure it works the same way in HDN. so unless you have purposely setup the low latency monitor path AND selected it then you are usin' the latency of the PCIe card only.

[DAW PLUS]

Originally Posted by psycho_monkey

I'm not sure you're understanding me here...

I'm saying if I have a 1024 buffer latency on my pt10/Focusrite pro24 setup, the latency is like a slap back echo. Unplayable.

Sure I understand you. In this regard, I can say, try an RME instead of the Focusrite. That should explain it.

[UnderTow]

Originally Posted by DAW PLUS

Sure I understand you. In this regard, I can say, try an RME instead of the Focusrite. That should explain it.

Indeed. Not all ASIO/CoreAudio cards are made equal. They can have very different effective latencies despite what they report to the host. The Focusrite has extra (hidden) buffers beyond the one you see reported psycho.

The HD Native card is not unique in its performance or method to achieve that performance. RME have been doing the same for nearly a decade.

Low-latency performance is one of the main reasons why RME added FPGA chips to their cards all those years ago (and Avid now copied). The I/O memory addressing is handeld by the FPGA chips instead of by your CPU. This makes things more efficient and lean.

Alistair

[thehightenor]

Maybe it's using a hybrid playback engine like Logic does?

I wish Cubase had the same hybrid playback set up as Logic, it does allow for same result as your describing. I'm using an RME interface with Cubase 6 (Windows) but once I have my buffers set to 1024 for mixing VI's are unplayable.

I'm quite tempted by PT HD Native.

[psycho_monkey]

Originally Posted by DAW PLUS

Sure I understand you. In this regard, I can say, try an RME instead of the Focusrite. That should explain it.

Originally Posted by UnderTow

Indeed. Not all ASIO/CoreAudio cards are made equal. They can have very different effective latencies despite what they report to the host. The Focusrite has extra (hidden) buffers beyond the one you see reported psycho.

The HD Native card is not unique in its performance or method to achieve that performance. RME have been doing the same for nearly a decade.

Low-latency performance is one of the main reasons why RME added FPGA chips to their cards all those years ago (and Avid now copied). The I/O memory addressing is handeld by the FPGA chips instead of by your CPU. This makes things more efficient and lean.

Alistair

OK. Well, I don't have an RME to tryout, nor do I see myself owning one in the near future. If anyone who DOES have an RME setup, could they try this simple test for me - is it possible to have your native buffer size set to 1024 and play in time?

Every other interface I've also tried has this "hidden" latency too then! Which multiplies rapidly when you start increasing the buffers...

As far as I can tell, there's NO difference with soft synths between a 64 sample and 1024 sample buffer on HD Native.

Originally Posted by thehightenor

Maybe it's using a hybrid playback engine like Logic does?

I wish Cubase had the same hybrid playback set up as Logic, it does allow for same result as your describing. I'm using an RME interface with Cubase 6 (Windows) but once I have my buffers set to 1024 for mixing VI's are unplayable.

I'm quite tempted by PT HD Native.

I don't think Logic DOES have a "hybrid playback setup". Every coreaudio interface I've ever tried (off the top of my head 002/003, Mackie Blackjack, various MOTUs including the PCIe card, plus the Focusrite and probably a few others!) has had a buffer size that affects synth playback latency - that was always the advantage of PT TDM, that latency wasn't affected by processing power, and you can always drop in on a near-completed mix.

The behaviour you describe is what I'd expect from a native system - which is why I was so surprised at HD Native.

As I said - I've not checked it with audio throughput yet.

tightenor seems to be saying that RME DOESN'T work the same way. which RME do you have?

Alistair - are you only referring to the PCIe RME system? Or any RME system?

[mirrorboy]

Originally Posted by psycho_monkey

OK. Well, I don't have an RME to tryout, nor do I see myself owning one in the near future. If anyone who DOES have an RME setup, could they try this simple test for me - is it possible to have your native buffer size set to 1024 and play in time?

Every other interface I've also tried has this "hidden" latency too then! Which multiplies rapidly when you start increasing the buffers...

As far as I can tell, there's NO difference with soft synths between a 64 sample and 1024 sample buffer on HD Native.



I don't think Logic DOES have a "hybrid playback setup". Every coreaudio interface I've ever tried (off the top of my head 002/003, Mackie Blackjack, various MOTUs including the PCIe card, plus the Focusrite and probably a few others!) has had a buffer size that affects synth playback latency - that was always the advantage of PT TDM, that latency wasn't affected by processing power, and you can always drop in on a near-completed mix.

The behaviour you describe is what I'd expect from a native system - which is why I was so surprised at HD Native.

As I said - I've not checked it with audio throughput yet.

tightenor seems to be saying that RME DOESN'T work the same way. which RME do you have?

Alistair - are you only referring to the PCIe RME system? Or any RME system?

Audio throughput works, if this is what you mean:

After testing last night I had to do vox. It was a HEAVY session. Deactivated the MF, made sure that LLM thing was checked in the menu and that's it. No latency tracking vox at a buffer of 1024. No FX tho- not on sends or anything.

Scott

P.S. I had the AES 16e for a couple years and it does NOT work like this (no latency at any buffer w soft synths or overdubs). But it sounds like some of the RME stuff does....seems Leon and Alistair know a fair bit about this kinda stuff. And I remember always seeing people rave about RME cards....and I didn't really get it- maybe that's why (other than stable drivers....which is no small thing).

EDIT: So I guess this would be impossible to do overdubbing an elec guitar while using an amp sim plug on it. How would you do this? I guess you'd have to take LLM off and lower the buffer. :(

[mykhal c]

Originally Posted by mirrorboy

EDIT: So I guess this would be impossible to do overdubbing an elec guitar while using an amp sim plug on it. How would you do this? I guess you'd have to take LLM off and lower the buffer. :(

yep...but you might be surprised at what a 256 or 512 buffer feels like for an overdub...with LLM off and no plugs on the master

[DAW PLUS]

Originally Posted by psycho_monkey

OK. Well, I don't have an RME to tryout, nor do I see myself owning one in the near future. If anyone who DOES have an RME setup, could they try this simple test for me - is it possible to have your native buffer size set to 1024 and play in time?

Every other interface I've also tried has this "hidden" latency too then! Which multiplies rapidly when you start increasing the buffers...

As far as I can tell, there's NO difference with soft synths between a 64 sample and 1024 sample buffer on HD Native.

Ok, now that you mention this I got pulled back to your real question, since I got stuck in a 64 sample buffer tunnel vision thing, forgetting you got the same results with 1024 samples.
Now, THAT will not be possible with the RME AFAIK. It already has 25ms latency at 512 samples buffer, so that is no fun. I'd need to check this, but since we are moving office and have some large projects at start, I don't see us setting this up on short notice.

[thehightenor]

Originally Posted by psycho_monkey

OK. Well, I don't have an RME to tryout, nor do I see myself owning one in the near future. If anyone who DOES have an RME setup, could they try this simple test for me - is it possible to have your native buffer size set to 1024 and play in time?

Every other interface I've also tried has this "hidden" latency too then! Which multiplies rapidly when you start increasing the buffers...

As far as I can tell, there's NO difference with soft synths between a 64 sample and 1024 sample buffer on HD Native.



I don't think Logic DOES have a "hybrid playback setup". Every coreaudio interface I've ever tried (off the top of my head 002/003, Mackie Blackjack, various MOTUs including the PCIe card, plus the Focusrite and probably a few others!) has had a buffer size that affects synth playback latency - that was always the advantage of PT TDM, that latency wasn't affected by processing power, and you can always drop in on a near-completed mix.

The behaviour you describe is what I'd expect from a native system - which is why I was so surprised at HD Native.

As I said - I've not checked it with audio throughput yet.

tightenor seems to be saying that RME DOESN'T work the same way. which RME do you have?

Alistair - are you only referring to the PCIe RME system? Or any RME system?

I have an RME Mutliface - and it definitely does increase latency with buffer size, as pointed out.

What I'm meant with Logic's hybrid playback system is that you don't need to increase the buffer size as the mix grows, you can keep it on say 64 buffers for playing VI's because by default everything else is receiving a buffer size of 1024 even though your buffer is still reading 64. Thus the name hybrid.

At least that's my understanding and explains why my Logic rig can do this and my Cubase rig can't. Plus note the fact that my Logic rig (used only for writing) is using a prosumer Alesis Multimix USB 2 interface costing £250 .... so go figure!

I have noticed on the Cubase forum, people have been requesting Cubase has this same feature as Logic, which is where I originally heard the phrase "hybrid audio engine" that's what made me wonder if Avid has implemented something similar? Albeit the other way around, in so much as, no matter what size you make the buffer the live buffer is always 64 samples.

I will be honest here, I'm not certain about the technical side of these things, but using Logic and Cubase daily I am certain that Logic can do something that Cubase cannot when it comes to low latency and playing back heavy mixes.

best
tht

[mykhal c]

FWIW...as i understand it...Logic's playback tracks are always 1024 while the tracks that are record-enabled are whatever buffer size one has chosen.

[UnderTow]

Originally Posted by mykhal c

FWIW...as i understand it...Logic's playback tracks are always 1024 while the tracks that are record-enabled are whatever buffer size one has chosen.

Indeed and the same applies to native versions of PT since 10 (or 9?). Another DAW with a "hybrid" engine.

As for psycho's experience of no noticeable latency with VI's in HD¦Native, I'm surprised. I haven't gotten round to testing regular PT yet. (I don't have HD¦Native to test). I find it hard to believe that Avid didn't widely publicize this if it is true. Can anyone confirm it?

Edit: I see Scott confirmed it. Interesting...
That said, I think what you guys are experiencing is just how a good native setup reacts (which for VI's not running on the CPU is better than DSP based systems). I could be wrong...

Alistair

[mirrorboy]

Originally Posted by UnderTow

Indeed and the same applies to native versions of PT since 10 (or 9?). Another DAW with a "hybrid" engine.

As for psycho's experience of no noticeable latency with VI's in HD¦Native, I'm surprised. I haven't gotten round to testing regular PT yet. (I don't have HD¦Native to test). I find it hard to believe that Avid didn't widely publicize this if it is true. Can anyone confirm it?

Edit: I see Scott confirmed it. Interesting...
That said, I think what you guys are experiencing is just how a good native setup reacts (which for VI's running on the CPU is better than DSP based systems). I could be wrong...

Alistair

Hey Alistair-

Yeah if you go back to page one I thought Pycho was having some weird experience, too.

I tried it myself on both heavy and empty sessions, NO (SAME) LATENCY whilst tracking VI's at buffers 64 through 1024. It's all the same. Nothing changes. No latency.

Crazy I know!!!!!!!!

All that had to be done was to deactivate the MF! Which, like I said, is what I was taught to do initially (on a TDM system at Quad / Avatar NYC) I just never knew why I was doing it. Now it makes perfect sense.

I have no idea how/if any of this works on on any other system other than HD|Native although I'm sure HDX is the same. (also you wrote VIs that run on DSP chips?? I think that was a mistake, right? No VI's run on DSP chips afaik)

Also, I'm using an Aurora 16 w. the LT-HD module hooked right now into the HD Native PCIe card. PT sees it as 2 x 192s. No hiccups ever. Not once. Drivers are rock solid. But Since I've sold almost all my outboard, I'm looking for less i/o. I'm also looking for a clean pre.

I'm selling my aurora and getting an omni. (Quick OT question- does anyone know if the converters in the Omni are for sure the same as the ones in the HD IO?)

--

You're right though about Avid....I was thinking the EXACT same thing!! Why didn't they advertise this more clearly!?!?!?

Also, throughput on an audio overdub works the same as per above.


Cheers man,

Scott

[mykhal c]

Originally Posted by UnderTow

Indeed and the same applies to native versions of PT since 10 (or 9?). Another DAW with a "hybrid" engine.

Alistair

hey Alistar...i can say with 99.999999% confidence that PT does 'not' work in this way

[UnderTow]

Originally Posted by mirrorboy

Then I opened up the heavy session I'm on now (UAD-2 a plenty) and all I had to do was deactivate the master fader, make sure there's nothing on the instrument track but the instrument vi and make sure it's going directly out your converter to your speakers (usually 1-2) ie no bussing, etc...

<snipped>

The truth is, this is the exact way the in house staff works at quad in NYC. I always did it as well when I was there but I didn't know why I was doing it (de activating the MF).

Yes, delay compensation is always relative to the full path. If you route a track so that it doesn't go through your main bus (for instance) it doesn't need to have the delay compensation applied relative to whatever is on that main bus. That is also why you should always create a separate bus as your main bus and not just use a master fader. This way you can always do an overdub or record a VO without suffering from any latency on your main bus (or any other busses in the "main" signal path).

My basic post-production templates always have a separate bus with an R128 meter on it which is fed by a send from the main bus. This metering has significant latency but by routing this bus to nothing, it doesn't affect the overall latency on the rest of the mix (including any ADR or VO tracks).

Here is an example of a session's PDC:



M1 goes to the MUSIC bus; FX1 goes to the SFX bus, both buses go to the MIX bus. The DIRECT track has heavy latency but as it is not routed through the MIX bus (like all the other tracks) the MIX doesn't experience this latency.

So in short, never use a Master Fader. Always create a separate bus for the main MIX. Then you can overdub without suffering plugin latency. (This applies to all versions of PT that I am aware of).

Alistair

[mykhal c]

Originally Posted by mirrorboy

I'm selling my aurora and getting an omni. (Quick OT question- does anyone know if the converters in the Omni are for sure the same as the ones in the HD IO?)

Scott

yes....and if you go to Black Lion's site you can read their discussion on both the Omni and HD I/O...where they talk about both and converters in the Aurora...some converters are used in all 3 interfaces.

[UnderTow]

Originally Posted by mykhal c

hey Alistar...i can say with 99.999999% confidence that PT does 'not' work in this way

How can you be sure? This is not about live tracks. It is about all the other tracks. The only way to know what the real latency of non-live tracks is is to insert something that reports the real latency back. Like VE Pro. If you have it, try it.

Alistair

[mirrorboy]

Originally Posted by mykhal c

yes....and if you go to Black Lion's site you can read their discussion on both the Omni and HD I/O...where they talk about both and converters in the Aurora...some converters are used in all 3 interfaces.

Great thank you!


Scott

[mykhal c]

Originally Posted by UnderTow

How can you be sure? This is not about live tracks. It is about all the other tracks. The only way to know what the real latency of non-live tracks is is to insert something that reports the real latency back. Like VE Pro. If you have it, try it.

Alistair

from direct discussions PT achieves its' performance due to it's real-time sample processing demand on the CPU unlike the other DAWs that wait their turn in the CPU scheduling queue (block based). this real-time sample processing CPU demand was a result of DSP cards requirements...where the code stayed in for PT native (for better or worse...dependin' on ones rig)

[UnderTow]

Originally Posted by mirrorboy

Hey Alistair-

Yeah if you go back to page one I thought Pycho was having some weird experience, too.

I tried it myself on both heavy and empty sessions, NO (SAME) LATENCY whilst tracking VI's at buffers 64 through 1024. It's all the same. Nothing changes. No latency.

I'm still sceptical. What happens at 2048 samples? Could it be you just don't notice the 1024 samples? (Hard to believe, I know).

All that had to be done was to deactivate the MF! Which, like I said, is what I was taught to do initially (on a TDM system at Quad / Avatar NYC) I just never knew why I was doing it. Now it makes perfect sense.

Yes, see my other post.

(also you wrote VIs that run on DSP chips?? I think that was a mistake, right? No VI's run on DSP chips afaik)

Oops, yes, I meant not running on DSP (as opposed to for instance the Virus TDM plugin).

You're right though about Avid....I was thinking the EXACT same thing!! Why didn't they advertise this more clearly!?!?!?

I'm still reserving judgement until I can test this myself but yes it seems weird...

Alistair

[mirrorboy]

Originally Posted by UnderTow

I'm still sceptical. What happens at 2048 samples? Could it be you just don't notice the 1024 samples? (Hard to believe, I know).



Yes, see my other post.



Oops, yes, I meant not running on DSP (as opposed to for instance the Virus TDM plugin).



I'm still reserving judgement until I can test this myself but yes it seems weird...

Alistair

Yeah try it on a HD Native rig.

You're in for a surprise.....

Scott

P.S. My rig doesn't show 2048 samples. Also, if there was any more latency...even a teeny bit....at 1024 vs 64......I'd know for sure....been dealing with latency as long as anyone.

[UnderTow]

Originally Posted by mykhal c

from direct discussions PT achieves its' performance due to it's real-time demand on the CPU unlike the other DAWs that wait their turn in the CPU queue. this real-time CPU demand was a result of DSP cards requirements...where the code stayed in for PT native (for better or worse...depends on ones rig)

All DAWs need "real-time" response and all processes on both Windows and Mac OS have to "wait their turn in the CPU queue". Pro Tools (or Avid) have no control over this. The only way to affect the behaviour is to change the thread priority. Any DAW can do this.

And again you won't notice extra latency on non-live tracks. They are just played a little early to compensate.

Alistair

[mykhal c]

Originally Posted by UnderTow

All DAWs need "real-time" response and all processes on both Windows and Mac OS have to "wait their turn in the CPU queue". Pro Tools (or Avid) have no control over this. The only way to affect the behaviour is to change the thread priority. Any DAW can do this.

And again you won't notice extra latency on non-live tracks. They are just played a little early to compensate.

Alistair

ok...just tryin' to share what has been shared with me...

[UnderTow]

Originally Posted by mykhal c

ok...just tryin' to share what has been shared with me...

And it is appreciated but I think somewhere between the developers and you something has been either embellished, oversimplified or simply misunderstood (I don't mean by you although that is also possible).

With the exception of the VI behavior described in this thread, the low-latency performance of PT is actually rather abysmal compared to other native DAWs. I think that what you were told must have been marketing tinted.

Alistair

[mykhal c]

Originally Posted by UnderTow

And it is appreciated but I think somewhere between the developers and you something has been either embellished, oversimplified or simply misunderstood (I don't mean by you although that is also possible).

With the exception of the VI behavior described in this thread, the low-latency performance of PT is actually rather abysmal compared to other native DAWs. I think that what you were told must have been marketing tinted.

Alistair

no doubt about me not havin' a clear and concise view of the under-the-hood workings of PT. however my discussions don't involve the marketing folks. if i can find one of the DUC discussions where we talked about this at length i'll post the link here. and FWIW my discussions are not limited to the DUC.

[psycho_monkey]

Originally Posted by UnderTow

I'm still sceptical. What happens at 2048 samples? Could it be you just don't notice the 1024 samples? (Hard to believe, I know).



Yes, see my other post.



Oops, yes, I meant not running on DSP (as opposed to for instance the Virus TDM plugin).



I'm still reserving judgement until I can test this myself but yes it seems weird...

Alistair

It might not be "zero latency" but it's not not noticeable for my keyboard drumming. It's nothing like the same latency as a regular 1024 buffer - it really is super obviously different.

[mirrorboy]

Just wanted to post that after tonights skype session (actually it was Face Time) with my writing partner / producer (who's on the east coast), I can verify another suprised and happy user.

We were listening to his HDX rig (Mac Pro 8 Core). The latency was bugging him so much he couldn't get a simple progression down in time. I told him about all this (which is pretty lame me telling him something about Pro Tools, or unusual rather)....anyways, he did indeed have a MF active.

He deactivated it and the latency was gone (I'm sure it wasn't 0.00 but it was not noticeable anymore, like w. a 64 buffer).

He was as surprised as I was when I told him to open the playback engine and look at the buffer. It was set to 1024 and he got no latency.

So we laughed about it a bit and how great this feature is (and how did we not know this? etc, etc). And about how horrible PT is when it comes particularly to running Kontakt!! :-) (but i think he said that NI is coming out w an AAX version very soon)

So thanks again pychomonkey!! Seriously thank you for posting this info!!


Cheers,

Scott

[UnderTow]

Well it sure doesn't work on my own regular PT rig (with an Orpheus). VI latency is noticeable. I purposely set the latency high to 30ms but I'm not exactly sure what the real latency is. The Orpheus reports it in two ways with a basic buffer in microseconds and extra buffers in ms rather than samples but PT only accepts certain buffer sizes... The way the system is behaving it seems like PT is forcing a particular buffer size on the Orpheus. It is all a bit unclear with this combo.

Next time I'm on a HD|Native rig I'll test to compare how that VI latency feels compared to my own rig.

Alistair

[psycho_monkey]

Originally Posted by UnderTow

Well it sure doesn't work on my own regular PT rig (with an Orpheus). VI latency is noticeable. I purposely set the latency high to 30ms but I'm not exactly sure what the real latency is. The Orpheus reports it in two ways with a basic buffer in microseconds and extra buffers in ms rather than samples but PT only accepts certain buffer sizes... The way the system is behaving it seems like PT is forcing a particular buffer size on the Orpheus. It is all a bit unclear with this combo.

Next time I'm on a HD|Native rig I'll test to compare how that VI latency feels compared to my own rig.

Alistair

Exactly - it seems to be an HD Native feature (and hdx, but you'd expect that).

[UnderTow]

Originally Posted by psycho_monkey

Exactly - it seems to be an HD Native feature (and hdx, but you'd expect that).

Maybe someone should inform Avid's marketing department.

Alistair

[DAW PLUS]

Originally Posted by mirrorboy

Just wanted to post that after tonights skype session (actually it was Face Time) with my writing partner / producer (who's on the east coast), I can verify another suprised and happy user.

We were listening to his HDX rig (Mac Pro 8 Core). The latency was bugging him so much he couldn't get a simple progression down in time. I told him about all this (which is pretty lame me telling him something about Pro Tools, or unusual rather)....anyways, he did indeed have a MF active.

He deactivated it and the latency was gone (I'm sure it wasn't 0.00 but it was not noticeable anymore, like w. a 64 buffer).

He was as surprised as I was when I told him to open the playback engine and look at the buffer. It was set to 1024 and he got no latency.

So we laughed about it a bit and how great this feature is (and how did we not know this? etc, etc). And about how horrible PT is when it comes particularly to running Kontakt!! :-) (but i think he said that NI is coming out w an AAX version very soon)

So thanks again pychomonkey!! Seriously thank you for posting this info!!


Cheers,

Scott

You mean his HDN rig, not his HDX rig, correct?