Waves and AAX DSP....

[DaveTremblay]

Just a couple of comments...

1) We have seen chip (TI) for chip (56k) performance gains of up to 7x through optimization of plug-ins. This benefit is in addition to increased signal fidelity and headroom with the switch to floating point. Not all plug-ins will optimize this well though. Most will be about 2-3x faster with a reasonable amount of optimization. Some algorithms aren't a good match for the chip architecture and we'll just see a 1 to 1, when you just consider chips.

2) Chip to Chip comparison is not an accurate assessment of HDX performance. That is like trying to compare Intel chip performance core to core, even though your new one has 4 cores and your old one only had two. No one does that. Our 5x performance claim is Card to Card. Looking at the unoptimized numbers for SSL looks like it's hitting 4x performance card to card if your numbers are correct. (twice the number of cores)

3) Getting code to run on the DSP is the first step. Optimization is the second, obviously. How much a plug-in developer wants to optimize is up to them. Trying to squeeze the last bit of performance out of a plug-in that is rarely used is not typically a good use of time. A sound design focused plug-in for instance, are you really going to have 100 of those in a session? Optimization is typically prioritized by real user impact. You can see in our own plug-ins that we spend considerably more time on EQ3, Dyn3, and ChannelStrip than some of our older plug-ins that are not used as much.

4) Comparisons between Native instance counts and DSP instance counts is not an apples to apples comparison. The HDX signal network operates with a much higher level of determinism and way lower latency than any Native system could claim. We even chose to give up a significant amount of theoretical DSP performance to run the network that way. It's like comparing a Porsche 911 to a Minivan, on gas mileage or cargo space. It's just not why you buy a Porsche.(Unfortunately, I can't claim that last line is from experience) :)

5) Optimization of your first plug-in is difficult as you learn what tricks will work and what won't. The first plug-in might take someone a few weeks to get a good result, but latter plug-ins will go much quicker. It's the same with Native plug-ins.

Waves as always been a strong partner of Avid. And the discussion surrounding their product plans doesn't have to be so polarizing. They have a great collection of plug-ins developed by a great team of engineers. I had the opportunity to work with one of them a couple weeks ago and was impressed with the solutions he was willing to dive in and investigate. I think it is clear to everyone that they have a giant collection of plug-ins, so any transition is going to be difficult. Seriously, even taking a single day to optimize each plug-in is over a man year of work.

As Waves customers, you might want to consider something other than the all-or-nothing petition. Possibly a more useful bit of information to collect and give to Waves is a prioritized list of which plug-ins of theirs you'd like to see on the DSP. Which ones do you need in the low latency domain for tracking? Which sit in your track/bus structure in a way that cause large latency compensation or voice count hits?

Dave

[Marsdy]

Dave.... Thank you for responding....

We have seen chip (TI) for chip (56k) performance gains of up to 7x through optimization of plug-ins..... Most will be about 2-3x faster with a reasonable amount of optimization

Most will be about 2-3x faster? That's a LOT less than your marketing guys have been claiming.... And this nine years down the line from TDM....?

With regards to...

This benefit is in addition to increased signal fidelity and headroom with the switch to floating point

Most have agreed in Avid's recent survey that there are marginal differences between TDM and HDX in terms of audio fidelity. The increase in headroom of HDX is indisputably an advantage over TDM as is the increased voice count. As I have said, Avid would be far better focusing on improvements in their latest A/D, D/A converters IMHO.

However, at the end of the day... surely the end user is equally, if not more interested in raw plug-in counts. In other words, is an HD3 user REALLY going to notice a significant improvement upgrading to an HDX1 system? The jury is out.

Dave... please don't take this as a personal attack on your good self.... I truly appreciate your contributions here on the DUC....

[dave911]

Thanks for the info Dave.

[DaveTremblay]

Originally Posted by Marsdy

Most will be about 2-3x faster? That's a LOT less than your marketing guys have been claiming.... And this nine years down the line from TDM....?

The marketing claims are based on card to card performance, not chip to chip. Just like Intel/Apple/AMD don't claim Core to Core performance when comparing performance of new platforms. If you multiply that chip to chip number by the number of chips, you will see 4-6x performance card to card, roughly, which was stated later in my post.

Originally Posted by Marsdy

However, at the end of the day... surely the end user is equally, if not more interested in raw plug-in counts. In other words, is an HD3 user REALLY going to notice a significant improvement upgrading to an HDX1 system? The jury is out.

If your question is really comparing an HD3 system to an HDX1 today, maybe not. Depends on the user. But an HD3 system is quite costly compared to an HDX1, so surely you could grant the price has come down, by 70% (ish, I'm not a pricing guy). And an HDX3 is clearly a win over an HD6. And any DSP accelerated system is a combination of all of the Native CPU you've got PLUS the DSPs. It's not an either or. Take the most massive 32 core Intel system. It runs a ton of plug-ins. If you want more, or lower latency, or higher determinism, add HDX cards and it is still more. Right?

Originally Posted by Marsdy

Dave... please don't take this as a personal attack on your good self.... I truly appreciate your contributions here on the DUC....

Thanks. I don't take it that way at all. We are going through a technology transition with the platform and that always raises questions, concerns, and a bit of good ol' rage. But trust me, we didn't take this course lightly. And AAX isn't done being developed, it's only beginning. Contrary to what you might read or hear, we have a really killer Engineering and Product Management team and some of the "pain" is part of a bigger picture transition. And a lot of plug-in developers get that. We've even had our fair amount of "thanks" from developers who are sick of dealing with all of the problems related to RTAS and TDM plugs.

Dave

[TOM@METRO]

Thanks Dave,

I for one, really do appreciate your presence here.

You touched on a very appealing possibility here with your thoughts on Waves. I know that for me, a dozen or so DSP PIs from Waves would sit well. That is, if I could pick ‘em. :)

Perhaps this could be an opportunity to list those PIs that we feel are most needed in DSP, and see if there is a consensus. Perhaps we could dedicate a thread to this.

[Tobias Eichelberg]

Thanks for chiming in Dave!

Before I moved to HDX, I was on HD3 and on large projects I had to carefully consider which plug-ins to use as TDM, so here's my priority-list of Waves plug-ins that usually sit on audio subgroups and master busses (I have Mercury+SCC TDM by the way):

C4/C6 multiband compressor (top priority for me as there are no good substitutes for this one yet)
RenDeEsser
VEQ4
H-EQ
PuigChild 660/670
PuigTec EQP-1A
CLA-76
CLA-2A
CLA-3A
MV2
L2
L3-LL (the linear phase ones have high latency anyways)

That's about it for me. I know for a fact that many top engineers still use the Renaissance-stuff a lot, if you'd add in:

RVox
REQ
RComp

[TOM@METRO]

For me.

L3-all
MV2
Aphex Exiter
Kramer Tape
H-EQ
V series
H-Deay
H-Comp
C6 Multi

This would bring me back into the club.

[Tobias Eichelberg]

I would like to add a few words to explain the somewhat frustrated sentiments towards Waves in this discussion:
My Waves bundle is a 12.000$ software bundle, that's a considerable investment (for me). It's been my main tool-of-choice for a long time.

While I totally understand that Waves has to carefully consider how to spend their developer's time and salaries, it actually took quite a few man-days for their customers to come up with the 12.000$ they paid for the product. And I chose Waves because of their reputation as a solid company with almost universal platform support.

Now there's a bunch of customers who are asking for AAX DSP support, yet we're given "One Knob", "The King's Microphones", NS1 - all this stuff that has been there before in one way or another. That hurts me as a customer.

[JFreak]

Okay, let's play along...

Here's my list of Waves plugs that IMO must get the AAX-DSP treatment fast:

- SSL E-channel
- API 2500 comp
- C6
- L2/L3LL
- Kramer PIE comp
- Dorrough meters

...and some 2nd priority plugs:

- SuperTap delay
- Aphex Exciter
- Vocal Rider
- V-EQ4
- LoAir
- GTR

The rest I can wait for very much longer.

Oh, and I'm a Mercury+SCC(TDM) customer who can't justify new hardware until my list of 1st priority plugs is made available as AAX-DSP.

[WorldStudios]

I will happily settle for just some Waves AAX Dsp first and more to follow- hopefully. I use the SSL plugins a lot, plus the MV2, the CLA comps and the H-Delay. After that, perhaps API. But the one plugin that NEEDS to be DSP is, of course, GTR. If I cant play it with no latency, it will not get used a lot.