That is like saying that a modern 100 horse power car isn't designed to work well with anything external because you can't hook up a horse in front of it to add 1 HP. This kind of argumentation completely misses the point.

Your arguments in these discussions never made any sense to people that really understand the technology but you had the benefit of the doubt because it wasn't possible to perform a direct performance comparison. Now that we can, you should just accept reality and move on.

If people like the UAD plugins then fine. That is a good reason to invest in that platform but arguing from a performance point of view, which this thread is about, is plain crazy.

Alistair

+1

And the next Mac Pro line which should be here in a few months... we will see the hexacore model as the entry level MP and very likely a 16 core model as the high end model.

If there are good native alternatives to UAD plugins, such as Ampex, Studer, Manley and Reverbs, I'm very happy to move to native. But there is none, unfortunately.

UAD card dongle copy protection scheme works perfectly so far, but if they could find any other solution that works as good as dongle card, they should move to native.

+1

Anyway, I was wrong bout the Brainworx Digital V2 being the first "multiplatform" plugin.

Here is the result with the TDM / UAD / Native versions of SPL Vitalizer.

17 TDM Stereo Vitalizers ( Each instance occupied 56 percent of one DSP so there is room for more plugins but no more Vitalizers ) HD2 PCIe.
56 UAD instances on a UAD-2 Quad.
390 Stereo AU´s in Logic. ( with heavy crunchy settings )
Mac Pro Hexacore 3.46.

Are you actually feeding the plugins with audio as well, playback, algorithms hitting the signal? In native hosts, this is how the CPU gets hit. Especially when stacking multiple inserts with compressors e.g., later processors might not get a (high enough) signal to process.

It is not a matter of stacking plugins, but actually feeding a signal which triggers all algorithms, yet spitting out a signal of the same level if possible to feed the next processor. It takes a bit of work to do this properly, but the test will be more realistic when comparing to DSP based versions.

I used to max out my processor. (mac pro dual quad)

Since I got a UAD Quad, I dont.

If thats just because Im making different decisions or using different plugins or my DAW differently I dont care really........producing music is about getting to the destination, a finished mix.

My vehicle aint stalling on the highway anymore.

Putting your "cars" up on blocks and revving their engines is all well and good but driving them "on the road" in a variety of different situations is where it counts for me. YMMV ! heh

The O.P did confirm he is passing audio thru the plugins.

*I see your point about stacking the plugins via inserts lowering the actual signal level to the later plugins , in turn lowering the needed processing. However , from experience with my own benches where we did an experiment replacing the full range Sine waves with recorded silence or even a drum loop at a lower peak level for example, the variable we witnessed was measurable, but it was not any where near large enough to account for the numbers presented for the plugins in the example(s) posted here. In that respect I don't think it is representative.

Also, the example you noted is only more realistic empirically due to having to compensate for the fact that the DSP cards reserve the uppermost DSP requirement for the respective plugins algorithms whether they are used or not. That IMO is not relevant to working environments unless the plugins are always used with all algorithms active , and even when that is the case, the variable was still huge.

Not for a static EQ plugin like the one tested here. The De-esser might use a little less processing if it isn't actually de-essing but I am guessing the difference is marginal.

For what it is worth, I fed the plugins with full mastered mixes in my test. The de-essers of all chained instances will have been active due to the way they were setup. So this is a full load on all the plugins.



I'm really tired, can you say that in English Vin? :-D

Are you saying that the noted performance differences can not be accounted for by any instances of the plugins working less hard because the signal levels (in this case that would be the high-frequency content triggering the de-essers) have been reduced by the time they reach that particular plugin instance? If so, agreed. :-)

Alistair

Trying to maintain the level through all stacked plugins gave similar results as before but of course using a low cut filter etc gave the next plugin less bottom to process so I rearranged the test and put only 2 plugins on each track and the computer maxed out on 182 stereotracks or 364 Vitalizer plugins but as there was 158 more tracks playing than the 26 stacked tracks when reaching 390 plugins I should say the trackcount itself put some drain on the computer too...conclusion, my previous plugin counts are still relevant.

Edit: Also when reaching 364 plugins I was using MOTU 896HD instead of built in audio as in the "390" test, I also had Pro ToolsHD running in the background, not playing of course but used to monitor the Optical Output from the 896HD which was not the case when using built in audio so the "stacked" numbers in the first post are still very valid.

I'm totally agree with you here, and this is the only reason why most people accept the UAD expensive bill (wich is a good reason in a way),

I just don't see any interest in UAD version of existing native plugin :

- same algo (, no upsampling, same algo porting to UAD system, brainworx already said it, even if some peoples continue to claim they sound better))
- same price (what is the advantage for consumer if an existing plugin running on a crack proof system is not cheap)
- less instances (not a big deal but it's true)
- more audio latency
- less flexibility (even an external dsp box is more heavy compared to an usb dongle).

you must be a crazy fanboy to buy things that already exist in their native counterpart (especially when you see how effficient those native plugins are, it's not like if they were CPU intensive).

you will have less for the same price (not exacly because you have a 1500 usd quad to buy before running it).

the only advantage is for UA because the more you own plugins that run on UAD the more you need to buy expensive dsp cards (they continue to sell 5x less powerfull dsp at 5 x more the price of intel i7 chip)

but if someone have a good reason to justify it i'm open (emotional or fanboyism are not good reasons, nor stability wich is a fake argument considering the fact those native plugins work very well if not better wich lower audio latency).

Hey Alistair,

True enough, I was referring more to plugins like compressors which I use in my benches that can impact the level as they are stacked.

Sorry Mate, my fault , rereading what I posted I realised a bit of Diagonalese had crept in.. LOL

Compensate was not the word I was needing there , I have amended the post above.

Here is the amended and hopefully clearer version.

In short , yes I agree that the performance differences presented here cannot be accounted for by the plugins working less hard due to the reduction of signal.

Don´t miss my previous post . I rearranged the test with only two plugins on each track , similar results.
I also did the the test with only one plugin on each channel. I maxed out the 255 track Logic limit with 650 percent CPU load in Activity monitor showing that 390 tracks should be realistic as Logic can utilize up to 950 percent out of 1200 on my hexacore machine. So my numbers are valid even if the DAW PLUS had relevant input about the test.

Hey Evan ,

That is exactly what I am saying , I must be getting lost in the mix... :-(

I got a little lost in reading this thread spread over 2 days, a pc and an iphone... but I was curious about audio level being lower when stacking inserts since it might render later inserts "unused".

I have noticed this too, but it depends on the plugins used - or better stated, I did not want this to overlooked despite it most probable not making much difference. If you stir, it is better to have all ingredients in first... heh

100% use of algorithms indeed is not close to a realistic working environment, that's why I mentioned in regard to the DSP version comparison.

Hey Leon,

All Good , we are on the same page ... :-)

Just to throw another curve into this mix , I just fired up a rig with a UAD2-Duo and noticed that "DSP Load Lock" is ticked by default.

So the reservation of DSP resources is actually an option , which throws some new variables into the mix.

I apologise if this has already been covered on this thread , but it would be interesting to test with the Load Lock Off on the UAD-2 so it can dynamically allocate the DSP resources according to algo's used, as is the usual practice with the Native versions of the plugins.

Might also be worth testing with the Extra Buffering Off as well.

That will place a final dot on the i.. IMO

If its already been done and I have missed it in the mix , might be worth a quick summary if it has been already tested and documented

@TAKFAT I think the DSP load lock is only relevant to a few plugins, as the channel strip plugins, where you can bypass the EQ section inside the plugin for example.

Look at this chart
UAD Instance Chart
and you can see the plugins that may benefit from the DSP load lock feature.

Waves and Softube just to name a couple, but all of this is subjective ;o)

Cool,

Just thought it was worth a closer look.