Voice usage with AAE

[dcboucher]

If the new mixer in PT 11 is the Avid Audio Engine, and there are no variants like TDM or RTAS, do we know what the voice usage is per track of audio? My LCR film mixes love to eat voices, and everything is 96K. With some developers refusing to make AAX DSP formats, the back and forth with Native vs. DSP plugins will be unavoidable for me. Will I still suffer from the voice burning of that DSP->Native->DSP transition on an HDX system? How about HD Native? Any new rules of thumb to minimize the negative effects if they exist? I have searched the DUC, and not really found a definitive answer for PT11 so please be kind if the answer was sitting there under my nose.

Thanks,
d

[nst7]

Can't answer the other stuff, but in HD Native, the voice count is strictly for audio tracks, so assuming your computer can handle everything, you are just limited to the built in track count of HD software. So that would be 128 mono tracks at 96k, with stereo using up 2 tracks/voices, 5.1 using up 6 tracks/voices, etc. Or 256 mono tracks at 44/48k, etc.

Also, HD Native does include the little meter bar showing how many voices are currently in your session, just like HDX or TDM.

[dcboucher]

That would be spectacular if the advertised number of voices actually represented the number of tracks. Certainly my sessions on a PT10HDX rig don't yield that 1:1 ratio. RTAS plugs were burning voices all over the place. Anyone reporting with an HDX rig for PT11 yet?

[Carl Kolchak]

Originally Posted by dcboucher

That would be spectacular if the advertised number of voices actually represented the number of tracks. Certainly my sessions on a PT10HDX rig don't yield that 1:1 ratio. RTAS plugs were burning voices all over the place. Anyone reporting with an HDX rig for PT11 yet?

I think it's the same situation as TDM + RTAS, although more efficient voice-wise.

Everything streaming off the computer hard drive is in Native land, as are AAX Native plugins.

Once you enter the DSP mixer you use a voice, so as long as from that point on, you stick with AAX DSP plugins, you're not going to waste voices.

If however you place a Native plugin after a DSP plugin, you used a voice going in to the DSP mixer initially, another voice gets used coming out of DSP land, in to Native land for Native plugin processing, and then a third voice is used going back in to the DSP mixer (you can see how this would escalate, if you repeatedly mix plugin types, even on a single track).

Rule of Thumb : place your Native Plugins ahead of your DSP plugins.

From what I remember, HDX is more efficient (somehow) with the number of voices it uses if you come in and out of DSP land repeatedly, though I'm assuming it's still the case that Native plugins on an Aux track cost a voice?

[Danny Caccavo]

With HDX, 11 is no different from 10 in terms of "voice stealing" for DSP->Host "round trips." But HDX steals only half the number of voices that TDM stole (uses one voice per channel for a round trip rather than two) so there's *some* consolation. It's one of those design limitations that comes from "serving two masters" (DSP and Host) that Avid has not been able - and may not ever be able - to overcome.
Yep - we gotta keep lobbying those plug-in developers to make DSP versions of their AAX plug-ins. Maybe the development process will get easier as time goes by, I don't know. But the problem sure does create a price gap in Avid's offerings, as an HDX-1 system underperforms the cheaper HD Native offering in this area. I'm sure Avid is VERY aware of this problem.

[dcboucher]

That was my experience on the last mix I did in the box. I actually got a better track count from my laptop with HD10 running in native mode than from the HDX. The HDX offered DSP for some DSP hungry effects which was very attractive, but if I am running some big mama CPU, the native mixer plugin seems to end up with higher track counts until you jump to HDX2

Would the HDX be better served by offering a choice for a native mixer plug in for mixing and dsp mixer for tracking? I know that wouldn't kill the round trip ticket but it might deal with auxes running native plugs.

I hoped that 64 bit might offer a higher voice count for an HDX card. I didn't realize how solid that ceiling was. After all, the voices are being allocated dynamically and the sheer number of paths changes based on sample rate. It stands to reason that the DSP could be coded to allow a higher number of I/O considering the wider bit path. Maybe it was maxed already?

[Chief Technician]

Originally Posted by dcboucher

I hoped that 64 bit might offer a higher voice count for an HDX card.

You get 256 voices per HDX card.

[dcboucher]

Originally Posted by Chief Technician

You get 256 voices per HDX card.

128 @ 96k. I have had zero clients at 48k in the last 5 years and haven't recorded at a 1x sample rate in a decade. 1x sample rates might be a delivery standard but it is definitely not an accepted multitrack standard in my recent experience.

[Chief Technician]

Originally Posted by dcboucher

128 @ 96k. I have had zero clients at 48k in the last 5 years and haven't recorded at a 1x sample rate in a decade. 1x sample rates might be a delivery standard but it is definitely not an accepted multitrack standard in my recent experience.

I see. I work in post production, and I can count on one hand how many times a client has asked for 96kHz over the last 12 years. Our experiences differ, and that makes sense. 256 voices will be enough for me when I upgrade to HDX. If 128 is not enough voices for you, maybe you need an HDX2?

[dcboucher]

That's what I am afraid of. I really don't need even 128 tracks but because of losing voices to native plugs, I'll need to rethink the stem mixer I use to print stems and composites. Do you know if you lose a voice in a native scenario sending out to dsp based plugs on a UA card?