Waves AAX DSP - which ones do you want?

[sunburst79]

Originally Posted by John_Toolbox

So even if it is X times as powerful, that's not necessarily what everyone is concerned about.

I suspect hat a lot of people would rather add another HDX card if they need more power and stay entirely within Pro Tools and its workflow than have to deal with kludgey work arounds and issues with recalling the sessions. Just a guess.

[bacchus40]

BINGO!, my view is, if i cant get the job done with ONE HDX, which really, for what i tend to work with, i'd be surprised... but i'd much rather just toss another HDX in the case, add a fan or two, and call it a day!!

[TOM@METRO]

Originally Posted by bacchus40

BINGO!, my view is, if i cant get the job done with ONE HDX, which really, for what i tend to work with, i'd be surprised... but i'd much rather just toss another HDX in the case, add a fan or two, and call it a day!!

That would be a double BINGO.

[bacchus40]

Originally Posted by TOM@METRO

That would be a double BINGO.

now you're just nit-picking... lol

i have no idea what i meant by that..

[tomric]

I hope, enough pro-users are replacing their plugins with some of the great alternatives, that already (or will) support HDX. Loss of money and relevance as a studio-standard, forcing other people to buy those plugin-suites for compatibility reasons, will possibly do more than petitions...

[Marsdy]

Originally Posted by tomric

I hope, enough pro-users are replacing their plugins with some of the great alternatives, that already (or will) support HDX. Loss of money and relevance as a studio-standard, forcing other people to buy those plugin-suites for compatibility reasons, will possibly do more than petitions...

I don't buy this argument. I already have most of the "great alternatives" from McDSP, Sonnox, and some Softube and Brainworx stuff. It's not like there has suddenly been a slew of brand new, never been seen before PIs for HDX that blows the TDM competition away. It's mostly the same old stuff, there's just much less of it.

Yes, some HDX plug-ins are now operating at 64 bit precision but there's a big performance hit there..... So now we're supposed to buy TWO HDX cards to improve significantly upon the performance of an HD3 system? That wasn't what Avid claimed.

....And you're forgetting that there are currently NO HDX alternatives to Soundtoys, Eventide and all the MANY other missing PIs and there may never be at this rate. Very few were complaining about how crap Waves PIs were until they weren't available for HDX.

As I've said, I just don't see the incentive to move to HDX at present. I'm sure there are MANY in the same boat. That's more Avid's problem than it is mine.

[DaveTremblay]

Originally Posted by Marsdy

Dave Tremblay of Avid has admitted that there is a performance overhead.

There is overhead to do 64 bit double precision in plug-ins on TI processors, but the overhead to do 48 bit integer on the 56k chips is way worse, not even taking into account the processor clock increase or the number of chips we have on HDX. Additionally, there is "overhead" to do 64 bit double on Intel as well, even if you're using AVX instructions. If double precision math was free on any platform we'd always use it.

Originally Posted by Marsdy

Yes Avid are claiming massive improvement gains with it's own plug-ins. But I'd put good money on these running at 32 bit floating point precision were the TI chip shines. Again..... Avid.... prove me wrong. And if they're running at 32 bit floating point precision then do they actually sound better than TDM running at 48 bit fixed point?

There isn't a single choice of single(32bit) or double(64 bit) floating point precision in a plug-in. We do thousands of different math operations in a plug-in and for each of those, we can choose whether we want single or double precision. As experts in the field, we often choose single for performance, but know where in the computations it is important to use double. This is certainly true of Avid, and usually the case with all 3P plug-ins.

And with that last statement, you're assuming that your current TDM plug-ins are 48 bit fixed point. Most TDM plug-ins run mostly 24 bit fixed precision internally due to the huge performance cost of running a TDM plug-in at 48 bit. So yeah, all of your AAX DSP plug-ins should sound better than TDM and they won't internally clip.

Dave

[Marsdy]

Originally Posted by DaveTremblay

There is overhead to do 64 bit double precision in plug-ins on TI processors, but the overhead to do 48 bit integer on the 56k chips is way worse, not even taking into account the processor clock increase or the number of chips we have on HDX. Additionally, there is "overhead" to do 64 bit double on Intel as well, even if you're using AVX instructions. If double precision math was free on any platform we'd always use it.

There isn't a single choice of single(32bit) or double(64 bit) floating point precision in a plug-in. We do thousands of different math operations in a plug-in and for each of those, we can choose whether we want single or double precision. As experts in the field, we often choose single for performance, but know where in the computations it is important to use double. This is certainly true of Avid, and usually the case with all 3P plug-ins.

And with that last statement, you're assuming that your current TDM plug-ins are 48 bit fixed point. Most TDM plug-ins run mostly 24 bit fixed precision internally due to the huge performance cost of running a TDM plug-in at 48 bit. So yeah, all of your AAX DSP plug-ins should sound better than TDM and they won't internally clip.

Dave

I stand corrected. Thank you for the clarification.

Can you shed some light on why many third party developers are getting such lower plug-in counts with HDX than Avid's PIs? This is a complete deal breaker for me because the instance counts I'm seeing with McDSP and Sonnox plug-ins for example just don't justify the cost of upgrading from an HD3 system.

[DaveTremblay]

Originally Posted by Marsdy

I stand corrected. Thank you for the clarification.

Can you shed some light on why many third party developers are getting such lower plug-in counts with HDX than Avid's PIs? This is a complete deal breaker for me because the instance counts I'm seeing with McDSP and Sonnox plug-ins for example just don't justify the cost of upgrading from an HD3 system.

Sure.

I think the main issue with McDSP instance counts is actually on our end. I've talked with Colin about it and we're trying to address it. In most cases, those aren't limited by chip performance but by input/output channel count limitations that we imposed. We're looking at a fix on our end. In addition, those guys upgraded their signal path to double precision because they had the available power in the TI to increase their signal fidelity. And they upgraded their parameter smoothing characteristics if I remember correctly.

I haven't talked with Sonnox about instance counts, so I don't really know what changes they have made to their plug-ins. Do you know the instance count comparisons for them?

Dave

[Marsdy]

Originally Posted by DaveTremblay

Sure.

I think the main issue with McDSP instance counts is actually on our end. I've talked with Colin about it and we're trying to address it. In most cases, those aren't limited by chip performance but by input/output channel count limitations that we imposed. We're looking at a fix on our end. In addition, those guys upgraded their signal path to double precision because they had the available power in the TI to increase their signal fidelity. And they upgraded their parameter smoothing characteristics if I remember correctly.

I haven't talked with Sonnox about instance counts, so I don't really know what changes they have made to their plug-ins. Do you know the instance count comparisons for them?

Dave

Here are Sonnox's spec for their EQ
http://www.sonnoxplugins.com/pub/plu...ts/EQ-Spec.htm
Oxford Dynamics are doing better but still way below the "up to 5x improvement."
http://www.sonnoxplugins.com/pub/plu...s/Dyn-Spec.htm