Physical Modeling vs. Sample Based Instruments (A comparison)


I always avoided the physical modeling instruments beacuse i felt that

  • computational load is much higher
  • sound quality is lower

So for that concert grand piano sound I started out with the ‘The Grandeur’ from Galaxy Instruments [ ] wich resembles a sampled Steinway Concert Grand. (99€)

It provides an average to good quality grand piano sound. But it never really made me feel at home with it.

After searching for better quality sound, I decided on the ‘Digital Grand’ from C. Bechstein (249€). That was a huge step upwards. Not only in cost and sound quality but also in loading times and time load (never measured just felt that way.)

Recently, with the release of Pianoteq 6 (99€, 2 Instruments) there was some hype around their product. I decided to give it a chance and downloaded the trial product Pianoteq Stage including all optional instrumunts but restricting usage to 20 minutes, after which a relaod is required, just enough time to perform some tests.

In general I rate the sound quality of the grand pianos to be equal to the Digital Grand, slightly better in the low and high registers.

Now with the profiler included in build 3543 one can directly compare the load profile of the three (using Kontakt Player VST plugin for the sample based instruments)

  • The Grandeur, Instrument: The Grandeur.nki
  • Digital Grand, Instrument: Digital Grand Player 1.1.0.nki
  • Pianoteq VST plugin, Instrument: Steinway D Classical

Here are the results (using out of the box ettings, no tweaking):

The Grandeur:

  • Total Sample Size 7.16 GB
  • Loaded Patches in Memory 0.18 GB
  • Profiler Memory Working Set: 0.6GB
  • Profiler Max. Time Load (max/avg): 10%/4%
  • Profiler Max CPU Load: 0,9%

Digital Grand

  • Total Sample Size 14.34 GB
  • Loaded Patches in Memory 0.83 GB
  • Profiler Memory Working Set: 1.5GB
  • Profiler Max. Time Load (max/avg): 31%/12%
  • Profiler Max CPU Load: 1,8%


  • No Samples
  • Profiler Memory Working Set: 0.4 GB
  • Profiler Max. Time Load (max/avg): 15%/9%
  • Profiler Max CPU Load: 2,7%

I expected a much higher CPU load, but it is just 50% higher than the Digital Grand, while the time load is 50% down. Very impressed.

Looks like physical modeling of grand piano has made it into the top league.

There are some other differences, like in the usability, user interface etc. which i did not investigate into. There is also a Standard and Pro version for those into tweaking the physical models.

Are you using physically modeled instruments? What kind? What are your thoughts and experience?



I’m using the SWAM Saxophones. They are 4 physically modeled saxes, and they are lightyears ahead of any sampled saxophones I’ve ever used. You really do need a breath controller to get the most out of them, though.

The whole line of SWAM instruments are physically modeled, and they are all amazing. I wish I could afford more of them.


SWAM horns are amazing. The strings, not so much. They feel very perfect and lifeless. It might be that with a lot of controllers and humanizing (mainly pitches, all sampled strings have too-perfect intonation imho) it would be better. I haven’t been a fan of PainoTeq up too now but I haven’t tried the new one yet. (There’s a devious way to cheat the demo too if you’re… well… devious lol Only for the purposes of doing a better evaluation of course.)


I respectfully disagree about sound quality of modelling compared to sample based instruments, unless you are listening to poor quality models.

Sampling will never be able to fully create the nuances that a well modelled instrument will. I have been a fan of modelling ever since I acquired my Yamaha EX5 in 1999. The AN modelling (Virtual Analog), VL modelling (Virtual Acoustic) and also the FDSP modelling (models electric piano and guitar pickups amongst other tricks) in the EX5 made it totally stand out from pure sample based instruments.


Shall I submit VB3 ?

Or all GSI instruments? Or Lounge Lizard?


It’s all about implementation. Yeah, a modeled instrument objectively has the potential to beat sampling on paper, but that doesn’t mean any given modeled VST is going to sound great in reality; few of them really do ino though the ones that work, work pretty well (SWAM horns being the best probably). And it depends on the complexity of the instrument involved. Though modeled pianos have a lot going for them I just don’t think the sound rivals a sampled piano (Painoteq 6 may change my mind, we’ll see. Electric pianos, otoh, do very well. All just depends. Modeled strings will almost certainly be the future but for now they stink. (Of course, sampled solo strings mostly stink too. But less.)


(That said, the day is probably not far off when the concept of sampling will seem like a quaint joke…"Yeah, people used to waste TERABYTES of storage on that crap! lolz ")


Still love my E-MU samples, but not for everyday gigging use. Studio or solo gigs, probably. Just not worth dealing with most of the time.


My original post was in regard of accoustic instruments, mainly accoustic grand pianos. But I do agree that electric/electronic instruments can be modelled as well. I just do not have any experience with such modelled instruments.

I would also expect that it should be easier to model electric/electronic instruments as the physics of theirt parts are well understand. Which is not the case for accoustic instruments, at least not in the same depth.


The Sampled vs Modeled is never-ending discussion for the Performers & Programmers. :slightly_smiling_face: The real question that I see is, presented with a good performanc in a great mix, does the audience really care about the origin of the sound? Most won’t hear the difference between Samples, Modeling & a decent GM module. I suppose we can, but we’re playing to the audience.


I guess you have not heard what Yamaha VL can do then? :slight_smile: and that is 20 year old wind/string modelling technology


Please all are my favorites!!!


I think depends on the company. I have “go to’s” for both. For piano I use UVI ‘s Ravenscroft 275 which is sample based ; but for Rhodes I can use Keyscape (sample) or Lounge Lizard (modeled) . I like the reliability of modeled and the size file/program size factor. I’ve had a few times where I’m sound checking before a show and I don’t get any sound from one of my sample vst’s and I have to reboot the machine. Modeled has never been a problem. I also use piano teq modeled piano for live with good results. And for Hammond B3, GSi ‘s VB3 is my first goto before. My Nord 5d! You have great options either way!


At the risk of sounding out of touch. You are saying that there is a non sample based Piano VST that is Modeled? Resource wise, that would be great. I just purchased Addictive Keys, which sounds pretty good and is lighter on the resources, but a Modeled Grand would be epic. I love VB3, but I favor NI Classic Keys Rhodes to LL, but I use LL for the Wurley. That is what I got LL for. I should have looked more into their Rhodes, but I liked NI, so I didn’t seek any improvement. Anyway, if Pianoteq is modeling based, I may give it a go.


It definitely is modeled!


Wow. It is pricey. Will have to save for it.


Luckily you can demo it for free though to make sure it’s going to fit your needs. Pianos are very subjective! I’ve downloaded v6 and haven’t tried it yet.


I just tried the demo!
To be honest: I didn’t expect much. I tested it some years ago and I was disappointed. But nowadays it seems to work quite well! To make it fit my style of playing I have to make some dynamic-adjustments and also edit the velocity curve. But all in all it sounds good! And of course it’s loading speed is like hell in comparison to my beloved 275.
I still need to play a bit with it to figure out, if it really fits my needs and my idea of how a piano should sound like.


Pianoteq has the best CP70 sound I have ever heard (to these subjective ears) as a free model (once you have purchased the main package).


Love the newest Pianoteq acoustic piano instrument: Steingraeber.