… just wanted to know …
Definately a combination, but I voted samples as I really love to be able to apply pitch slides, sample offsets etc 
Normally I use a lot of VSTs, but when doing compo songs and you’re forced to use just samples it feels so right 
sometimes i use to sample some vsts just to use vibrator or pitch-slides …
[espacially when making chiptune-style songs …] but as i don’t have a
huge collection of samples i prefer to use VSTi …
I don’t think you can really give a clear answer on this one since they both have very strong good and bad sides.
For percussion sounds I use samples 99% of the time. I like the flexibility of being able to easily play the drum sounds at different pitches, chop them up, loop bits, reverse them, etc. All stuff you can do with a good VST sampler but it’s so much more efficient to just use the samples directly I think. I use the Microtonic and Drumatic drum synths a lot but most of the time I just use them to create samples, then I sequence the samples instead - unless it’s something which has a nice, changing timbre which only sounds good being synthesised in realtime by the VSTi itself (Microtonic in particular).
But then there are other VST instruments such as Albino and QuadraSID which I use regularly. Most of the time I use these in realtime to maintain the diversity of sounds they can create, the changing timbres and modulations, etc., stuff that simply gets lost in a static sample. Then again, sometimes I wanna do really weird stuff to a certain sound like a bass or synth, so I render it to a sample and then mangle that up even further. In drum&bass music for example there is usually a lot of re-sampling going on with those evil basslines. Playing back a looped bass sample at different pitches can introduce that rhythmic looping artifact which actually sounds great sometimes too. A lot of the time you simply cannot get a synth to make “that sound” on its own, so you will be forced to render a sample, process that with some filters or distortion, resample, process some more, resample, etc.
Speaking of effects, you have things like Supatrigga and BufferOveride which pretty much never produce the same result twice, they are completely random in nature. It’s not practical at all to use these things in realtime on a song since you’ll never be able to predict the final sound. In these instances I will render out a long section of my song (4 - 8 patterns or something) and then cut out the little pieces which sound nice, and sequence these back into my original song.
etc.
Overall though, I can say that for most melodic elements - bass, synths, pads, etc. - I will use VST synths in realtime. For percussion and other sounds it’s usually always samples.
Samples - using the RNI interface for everything makes things so much easier IMO. VSTis become just sample sources.
That said, a multisample function (which just turned the current VSTi patch into an RNI) and the ability to layer multiple RNIs so that a single NOTE ON would trigger them all would make this method of working easier. 
I chose VSTi because I love to program the signal-chain in some vstis. But on the other hand, those VSTis are quite expensive. I have my favourites like Legacy (heard it at my hardware-dealer) and Vanguard…
Samples, but i never used the RNI thing. I load a sample per slot and get down to work!
It is true you can get a bit more managable sound by controlling the parameters manually, but I’ve still never been able to get it to behave exactly how I want. The buffers seem to continue running on their own time independantly of the host or something, and as far as I can tell there’s no way to reset it so that it starts exactly when you manually change a parameter.
Here’s a simple demo tune which will probably highlight it better than I can attempt to explain it. You can clearly see/hear that even though I’m performing the exact same commands repeatedly it’s different each time.
http://illformed.org/downloads/bufferoverride.rns
It would be great if there was just a simple reset button, similar to the reset on Renoise’s LFO metadevice, which would force the buffers back to the start and allow you to trigger it whenever you want.
Not the end of the world really, but it’s a minor annoyance if you’re trying to be precise with it.
i am using both to an equal amount, maybe with a slight advantage for VSTi.
if the CPU wasn’t limiting me, i’d probably use 99% VSTi, the remaining 1% would go for vocals, which i merely use as FX or as a hook.
certainly depends on the kind of music you’re in to… drum&bass is certainly crying for a decent amount of sampled resources, but as i’m currently very much into the sound of analogue electro-tech stuff, i couldn’t live without the sound and flexibility VSTi offers.
concerning percussions, i also love using synthetically generated material - it’s more flexible than samples if you look deeper into it, because pitch slides, offsets and other tracker based effects are just laughable compared to the possibilities of a (for example) microtonic drumsynth.
of course that only applies to scenarios where you want to work with single-shots… using drumloops (+sample offsets) is lame anyways 
so there goes my vote for the magical software synthesis interface!
Slightly updated situation with regards to BufferOverride (sorry for hijacking this thread a bit, heh):
If I force the effect on and off like this using the bypass command it does actually behave identically each time the effect is turned back on. Essentially it’s behaving like the reset button I was talking about earlier. This was not an approach I had tried before with this particular plugin, but it seems like it might actually work.
Hurrah!
.
samples + vst
a few vsti
I only use samples (wav) and rni’s. When I use a VSTi, I render the VSTi to a rendered sample and delete the VST.
I only use VST’s for FX (Mastering and EQ)
I can’t really answer this poll. It’s the combination that makes me like Renoise. If I’d be stuck with one of them I’d use either Impulse Tracker or Cubase instead. 
I use VSTi’s more, but then again most of them are sample based.
I use a combination maybe 75% vsti and the rest samples. I use samples mainly for drums and FX.
But I would love to use samples if I could get the same quality.
Because when you use samples the commands are allways the same no need to setup any routings etc.
But to be able to use samples more would require better velocity sensitve multisamples, some more options for the automation curves and keep on disc option. Otherwise the modules would become to big.
I find it hard to tell.
both have their own advantages.
Personally I use samples for drums and (off course) recorded material.
only vsts -> running out of cpu -> render to samples -> tweak’em -> again vsts 'till no cpu cycles -> over and over…
i use no samples for drums, but maybe we will be able to import rex, someday. (convert it internal)
I went with samples only because Everything I do besides synthesis is sample based, but even then- I use a combination of samples with vst and vsti’s…
this is comparing apples to bowling balls to me.
I’ll join the club and answer “combination”, but I prefer samples as I’m more comfortable with editing RNI. Wish Renoise could load STI, tho, as I own Sample Tank L, multisamples galore but in a different environment.
sorry , can’t choose so didn’t vote…