nice work Vin!

Interesting. Question is how much this relates to real-life practice? I still suspect that most studios use their DAW as glorified digital recorder with basic editing functionality. As such they don't really need a lot of "performance", just stability and functionality.

Under that light it likely doesn't matter much if PT9 does worse or not, because most people ain't gonna come close to these borders.

Producers/artists who like to use lots of ITB software effects instead of 19" outboard racks on the other hand might want to consider if exporting stems from a better performing DAW isn't more productive to just buying a brand-name.

Hey Timur,

Horses for course , of course... :-)

With the release of PT there was a huge flurry ( I would even call it close to hysteria ) of interest in Protools from many who had never dipped their toe. I had numerous clients contact me who's primary work scenario is film/TV compositon, who use large session templates based on sample based virtual instruments. The question was asked whether they could navigate to PT9 from their current DAW Hosts and still maintain those large session templates .

The above indicates not.

Yes there are some workarounds being investigated using VE Pro that alleviates some of the variable , but its not a real solution to the native RTAS VI performance issue.

Also for those using Cubendo on Windows and wanting to have a direct comparative of VI /Sampler performance on OSX , it will be of interest.

For those that use their system as glorified tape machines, as discussed in another thread and as you noted, they could still be using XP /OSX 10.4/10.5 on 5 year old hardware and 4 year old DAW Hosts, so this information is obviously of no interest or concern, but for those that use the DAW's more extensively , this info is quite pertinent. With the current tech systems being so powerful, the later is fast becoming the norm for the vast majority who can and will work totally ITB.

If the performance variable was small, it would be a non event, but for anyone not surprised/concerned by the actual performance variables displayed by PT9/RTAS VI performance , then I really have no answer past - no matter what AVID put out there , it will be consumed regardless because its the "industry standard "

That how ever will not see the issue addressed... :-(

I know for a fact that AVID is considerably concerned about the VI performance of RTAS , however I suspect that in this instance the ship is going to take quite a while to turn around.

BTW: I am going to ask the mods to change the title of this thread to - DAWBench DSP / VI Universal - Cross Platform DAW Benchmarks :

those results really are absolutely shocking.

both for ProTools and for Mac OS.

they don't really matter to me right now because of the work i do and what it demands of the system that i choose to run it on, but who knows what the future may bring for me so i am happy to have this information.

fair play to you and thank you once again Vin.

Vin - your work is VERY much appreciated.

As you know I'm a long term fruity fellow - and am somewhat disappointed to know how far behind we are. There are ways around the performance issues (just add more computers and VEP) but as someone who does use 300+ track templates with Vienna, EastWest (AGHHH), LASS, Spectrasonics etc, one has to start thinking that perhaps life IS greener on the other side of the fence...

Brendan.

Hey Vin,

you indicate you used Cubendo 5.x.x, could you be more specific?

I assume you used the 32-bit version on both platforms?

Also, which specific version of OSX did you use? Did you boot it up in 32-bit or 64-bit mode?

And is a C6 (64-bit) comparo on the way?

Hey Z,

That last test run was on C5.5.x , can't remember the exact version now sorry , if you notice I use the Name Cubendo to signify Cubase 5.5 / Nuendo 5.

I cross referenced the results with Nuendo 5 , and they were on par as C5.5 and N5 use the same seq version of the audio engine.

OSX version was at 10.6.2 running in 32 Bit Mode, I have Cubase 6 / 10.6.6 installed now and will do some future testing shortly.

Cubase 32 used across the board as the DAWbench VI test only uses 2.7GB total, which is well within the 4GB assignable memory. ( My understanding is that OSX 10.6.x using PAE even running in 32 bit mode has a 40 Bit memory address space , so assigns the full 3.5-4GB to Cubase. )

I'll look into doing a 64 bit version , I'll need to beef up the template to get the memory usage more heavily engaged.

Hey Brendan,

I missed you post earlier , cheers for the support.. :-)

You know only too well how much time and energy I and my test partners have put in over the years to try and get Steini to focus more attention on the performance of OSX , and I have come to the simple conclusion that it isn't going to get any better than what we have now without a major paradigm shift - i.e . Hybrid Engine.

A lot of the underlying issue is within OSX / Core audio, but of course that is slippery slope to try and navigate and discuss . The only comfort Steinberg can draw from the results that I see are that they are not on an island when it comes to performance issues on OSX , as pretty much all cross platform DAW hosts suffer to a certain degree with the cards that have been dealt.

I'll be interested how Samplitude performs on OSX when its released next year, as that has a hybrid engine. It may just place the cat amongst the pigeons on OSX... :-)

Re the life being greener , well you can always just hang the greener pastures onto the MAC system as slaves as you mentioned, hide them away so as to keep the illusion for the clients that you have remained loyal to the fruit... LOL

It should be remembered that from a VI developer's point of view the biggest market by far is Cubase on the PC and that's exactly where they'll expend the most resources to get optimal performance.

When you change platforms and/or hosts, it's never really a direct comparison. That said, if you really need serious VI power, Cubase on the PC is where you'll find it.

I am not so sure its that clear cut, and I suspect a lot of Logic/OSX users would question that assertion.

PT9 has changed the playing field to an enormous extent tho, I am told from a reliable source that the end user base increased 3 fold since its release, whether that is accurate I can't say for sure, but either way its in AVID's best interest to resolve the RTAS VI performance issue. One way to ensure its remains high on the agenda is to keep it right up in their peripheral view at all times... :-)

I'm sure it's probably a top priority at Avid however I don't think one can simply assume that all of the problem is Avid's or Apple's.

OK, but since you ran Win7 in 64-bit shouldn't OSX be booted in 64-bit as well? Not saying this would affect performance noticeably but just to cross the t's and dot the i's, if you will.

Lest you have guys like me come in and split hairs, you know... heh

And the latest version of OSX is 10.6.7 BTW.

I'll run thru some OSX 32/64 comparatives when I get a chance , but considering that we are not hitting any memory walls , nor is there a 64 bit version of PT9 , it hasn't been a priority. I have not seen any reports of performance improvements using the 64 bit kernel, but I am happy to look into it further

Re 10.6.7 , I know some automatically update to the latest and greatest, I tend to update less frequently - if it aint broke, don't fix. Most of the point releases of OSX are so minimal , it doesn't change the playing field or results. Anything past 10.6.4 is just fodder like the app store and some extra compatibility for the newer SB architecture , nothing associated to X58.

That is certainly a factor, but Apple itself is to blame here, too. They have habitually frustrated 3rd-party developers with midstream changes that potentially destroyed thousands of hours of labor.

For example: Adobe (and Steinberg, IIRC) had Carbonized 64-bit OSX versions of their popular software ready to go when Apple pulled the plug and they were forced to recompile for Cocoa.

Add to that the switches from OS9 to OSX, from PPC to Intel and it becomes clear that we should just be happy that 3rd-party developers did not completely abandon Apple in the first place! They cover a smaller segment of the market but have required a lot more effort and upkeep over the last decade or so.

The upshot is that OSX is a better platform from a technological POV, but that is of little consequence to the end-user when the apps fall behind.

Windows is a bit of a technological flea-market, but it has provided a consistent platform for more than 15 years now. That combined with its dominant market share makes it much more attractive environment for developers.

A bit of a strange comment in a thread about benchmarks that show Windows consistently beating OSX performance wise.

Talk about backwards compatibility, check out this film (if you have 10 minutes to waste that is. Otherwise jump to 7:30) YouTube - Chain of Fools : Upgrading through every version of windows (HQ)

Windows (7) is backwards compatible for nearly 25 years! Amazing!

Alistair

Well with RTAS its certainly not reserved to Apple , as the issue is cross platform , so that is 100% AVID's bag. Re the performance issues of the other cross platform Hosts , well Apple have to hold at least one hand up there, the other the devs. I am willing to bet each will point the finger at the other.

3 FOLD?!?! That's 8 times their pre PT9 user base!

3 times I could have seen, but 8 times? That's HUGE!!!

ok interesting

I have another idea for the testing protocol:

run this session 4x a day for 6 months inserting a mix of plugins and take a note of crash frequency/stability on each platform...



I know - but it would be interesting.

Hold On, hmmm, 3 x is what I meant...., need to double check something..

Threefold : three times as great or many; "a claim for treble (or triple) damages"; "a threefold increase"

We are getting lost in the mix somewhere, not sure how you are coming up with 8 ?

Either way, even 3 x is more than substantial. I am betting any of the other devs would give their left knacker to have achieved anywhere near that.

@ SWAN, I'll get on it... LOL

Hmmm... perhaps I must sit corrected!

I was corrected on that by my physics teacher 20 years ago - he stated that a fold is litterally folding something (like folded steel), where each fold doubles the value of the preceding, thus 5 fold is 32 times the original number.

According to the interwebs that logic is incorrect in these circumstances.

Apologies.

Hey Matt,

No apologies needed at all, as your explanation makes total sense.

All Good, so to be clear , PT core user base has allegedly increased by 3 x since PT9's release.

Well, it can make total sense, but still be wrong. ;-)

And you're right, tripling the user base is substantial, but semi expected initially as you open a closed platform to the masses. I suspect that growth rush has passed in favor now of steady incline.

I'd be very curious to do this kontakt test with logic.. I noticed that Cubase performs really well (in my own tests also, even though logic beat it for FX), with instruments.

Notice how abysmal the Cubendo OSX results are even on 512 buffer.

I'd test by having logic on the different buffer settings and have that instrument track selected. Which means logic will be in "live" mode and the real buffer setting being used.

T,

I am interested in seeing the Logic results myself, just don't have the time and energy to do the ports.

I need a volunteer to help me with Logic that is familiar with Cubendo / Protools , you are more then welcome to port the sessions across , full credits applied , of course... :-)

The Reaper ports are almost done for those interested..

Does that mean that Logic will no longer benefit from the "hybrid" approach?

Hey Z,

Hmm, interesting point..., but All instruments would have to be selected and its not really a valid comparison as no one would use Logic in that manner.

Worth a look see tho.. :-)

FWIW: The Audio Interface is extremely important in this test as well - case in point, had one of these newer Mixer come FW interface units that are flooding the market on the bench to do some qualifying recently, and although it did O.K with the DSP test, on the VI test in collapsed compared the RME reference hardware.

Something to keep in mind

I will help porting Tafkat but if i need Cubase i can't :(

I don't own it anymore.


I thought the test was maximum voices out of one instance of Kontakt so i misunderstood that part. Yes it is impossible to have all instruments channels in logic in low latency mode at once.

However the next best thing would be to test at small safety buffer which is 512 which would at least give us a good comparison at the cubase 512 buffer where the windows versions gets 3x the polyphony of the OSX one.

With a motu i think it's one of the best FW interfaces driver wise (in mac land) so i'd be a good test candidate. I do have an m-audio lying around somewhere too. If Tafkat wanted to loan me a babyface for a weekend i could do RME USB tests also. In any case my back is not bad at present which will be short lived so if there's any way i can help. this is the time.

Hey T,

Without Cubendo as a reference/template to port the session, it will be pretty tough :-(

FYI - There are 6 individual Kontakt instances in the template, all with 16 multitimral parts , 5 of those are for the polyphony loading so not all being used until called. Of the 5 x Polyphony instances each one delivers 320 notes of poly , for a total of 1600 notes available before the current session is tapped.

We haven't come close yet... :-)

144 reaxcomps, 64 samples without any tweaks in BIOS

For those interested,

The Reaper ports of the DAWbench VI sessions have been completed and have now been uploaded to the DAWbench site.

Get them Here

I'll post some preliminary results shortly.