I continue to wrestle with maximizing performance on my single-machine rig, and I'd like to know if I have things maxed out, or if I should be able to squeeze a few more voices out of it. Here's the system:
2 gigs of Kingston DDR2700
Dedicated PATA sample drive, which benchmarks to about 40 megs/second average read
Running Kontakt 2.0.2 within Sonar 4
When I load a sizable piano sample, I can play forearm clusters up to 120 voices before getting violent sputters. With a full complement of samples loaded, I get up to about 50-60 voices. In each case, the DFD meter hits the red just before dropouts occur.
This leads me to believe that the disk is the bottleneck, but that leads me to two new questions:
1 - How to tweak the new DFD settings, other than the basic slider? I think I understand the old-style DFD parameters, but the new ones aren't the same.
2 - How likely am I to get more voices from a hard drive upgrade? Will a typical SATA drive help any, or is it time to upgrade the rest of the system? My CPU is usually around 50-60%, with occasional surges to the 70% range when the poly count goes higher.
If you can provide any advice, you have my thanks!
I think your very high CPU usage is because of the latency. Do you have it very low like 10ms or perhaps lower? Try taking it up to around 20-25ms and you should see a decrease of around 20-30% cpu. 20-25ms is a bit big for realtime playing but for a orchestral mix its ok.
Try setting your DFD settings to around 70kb for each sample and see how that works out. If you hold sustain notes and you see them cutting off after a small period, increase your buffer size to double. Generally i believe its good to have different settings for each song. For example if you are only loading a piano, take the buffer up to 250k as long as you don't see "insufficient ram" message from Kontakt.
If it is a really heavy orchestration and you need ALL the ram, you can just have the Buffer at 24kb just so you can hear the song play in realtime. I can never load enough instruments and have them playing correctly so i usually make a "sketch" song temporary. Afterwards, i render each instance of Kontakt seperately and i increase the sample buffer size to around 200kb or even more, as much as my ram can hold.
A bit of info around what styles of music you compose would be very helpful to determin the best settings.
Thanks, Rich & Theodor. I'm working on pretty dense orchestrations right now, loading samples up to the threshold where K2 gives me threatening messages about stopping the DFD engine & what not. Perhaps it's time to add another gig of memory - I know that I'm not using all of my 2 gigs just for samples. I'm trying to avoid adding a 2nd system to the rig. I've been there & done that, and didn't like the extra complexity.
I hadn't seen the DFD tabs per-instrument. I'll fiddle with it a bit. It's also taking some experimentation on the polyphony limit. I'm using a few QLSO Gold sounds with release trails, and they eat up the poly. And, I'm using a couple of the K2-included VSL legato instruments, and they consume RAM like no other.
In the meantime, my workaround has been to open an extra instance of K2 as my "freeze" bin - whenever I get a track settled that I don't think I'll use for a while, I'll load up that track's instrument in the new instance and refreeze it. Cumbersome, but I hope it will get me through this project.
The best thing you can do in K2 is..once you get a part down in QLSO with a patch, Purge it. You can do it isnt. by inst. or the entire instance..Brilliant..saves a ton or ram and lets you load more patches per instance..Best RIch