At least, it is what I found out today
I finally decided to make a test on GS3 and K2 my self, and the results did not follow the common sense that GS3 has superior performance than K2, both in RAM management and voice output.
The fist thing that I noticed is that each note in k2 counts as two voices, wich means that if we have a 32 note count in Kontakt, it is equivalent to 64 in GS3 and so on. This already surprised me a lot, because I frequently hear user´s complaints that K2 can´t reach the same amount of voices than GS3, and this perception is actually based in a wrong interpretation of data. So, that "huge" 400 voice limit that one can achieve with GS3 is actually a 200 note limit in K2, wich is perfectly normal even for a slow setup like mine. Maybe I just misinformed, but I already felt very disappointed with many GS3 user´s opinions just because of that, for this fact is rarely mentioned, and so far has misleaded many people regarding K2 performance. Looked a little like "cheating", if you know what I mean.
Whatever, maybe I am just late to find out about this important detail.
Let´s go on with the tests:
Library: Sam Trumpets – Patch (close) Sustain 1 Rel_S
Computer: Athlon XP 1700 – 1 GB Ram – 120 Gb Seagate Barracuda exclusive for samples, newly formatted drive (I know, it´s an old computer, but the point is testing performance, right? )
Kontakt 2 – latest version, DFD set to 256 notes (512 voices)
GS3 – latest version, set to 512 voices max polyphony
Sequencer – Cubase SX3 + giga VST adapter
Results after loading the instrument:
GS3 – 31% of total memory used (according to GS3 interface)
Msg32.exe – 263.168 K
Gstudio.exe – 32.348 K
Total memory load: 295516K
Voices result: 270 voices – dropout
240 voices - clicks and pops
Cubasesx3.exe (before kontakt+patch loaded) - 83.276 K
Cubasesx3.exe (after kontakt+patch loaded) - 231.076 K
Total memory load = 147.800K (calcutaling the difference)
Voices result: 256 notes (512 voices!!!) – dropout
no clicks and pops until dropout
MY Conclusion: Kontakt could play twice more voices than gigastudio, despite the fact that the two samplers were configured with similar settings. There was also a large advantage for Kontakt regarding use of RAM, since its load was considerable smaller than GS3.
The test was also made using other patches, with the same results, including pianos, and string instruments.
Ok, so far I am taken to the conclusion that, at least for my system, is it not true that GS3 is more efficient in any way than K2. Other people could do the same tests and post the results here, to make a definitive comparison between the two platforms.