VSampler 2 FAQ

It depends very much on your usage, but here are a few numbers in short, more details with the next 2 questions:
minimum: 200 MHz PII, 64 MB RAM, Windows 95, 98, ME, NT4, 2000, XP
recommended: P3, P4 or Athlon, 256 MB RAM or more, Windows 2000 or XP

And a working ASPI layer is required for the AKAI CD import, check the first question of the AKAI import section of the FAQ.

Yes of course, VSampler loves Win2000 and WinXP. The only "disadvantage" against Win98/ME is that Win2000 and XP don't come with ASPI drivers, but a working ASPI layer is required for VSamplers AKAI import. So you have to install ASPI drivers manually, if your CD recording software didn't do that already, e.g. Nero). But that's no problem, check the first question of the AKAI import section of the FAQ.

It depends very much on your usage. You need enough CPU to run at least your sequencer and VSampler, probably some other instruments and effects as well.

The good news: VSampler is one of the fastest software samplers around, which has also been confirmed by several mag reviews in the past. The bad news: there are a few very "popular" ways to waste VSamplers power and CPU cycles for nothing, check the question VSampler takes sooo much CPU - what can I do to speed it up? at the end of this page.

To give you a number what to expect, on a Pentium III 700 VSampler is able to playback around 80 stereo voices in Standard quality, when all instruments use filters and LFO's that drops down to around 40 voices in worst case. The voices indicator at the upper right corner shows the number of active voices.

VSampler is able to perform bank loading in two different modes, and sample playback as well. All modes can be freely combined. While you usually don't have to care about those details (especially if you got enough RAM), you can increase the performance of VSampler on systems with less RAM to fit your song situation perfectly, without performance loss. The following chart explains the two different bank loading modes.

Under Current Bank preferences you can activate the current Preset and Multi only option for dynamic instrument loading. If it's not activated VSampler pre-loads the complete soundbank - all instruments. "Pre-load" usually means "into RAM", check next question for the other sample loading options. Note: it's a bank setting, not a general setting, so if you want to change the behaviour of a certain bank you have to re-save the bank with your new setting.

VSampler is able to perform bank loading in two different modes, and sample playback as well. All modes can be freely combined. While you usually don't have to care about those details (especially if you got enough RAM), you can increase the performance of VSampler on systems with less RAM to fit your song situation perfectly, without performance loss. The following chart explains the two different sample playback modes.

"Sample streaming from disc" means VSampler will play a sample directly from disc, without loading it into RAM before. This applies to the complete sample, it can be streamed from disc or not. This simple kind of streaming is not as powerful as the famous streaming engine of Gigasampler, which devides each sample into a RAM-part (first part of the sample) and a streaming part (last part of the sample) and then switches between those two parts when needed and thus allows fast sequences to be played with high polyphony as well, because fast sequences usually mean only the first portion of the sound is needed - coming from RAM.

Under Bank / Preset Defaults --> Streaming you can define VSampler's default behaviour, above which sample size it should automatically activate disc streaming for a new sample. Note: this setting is a sample-setting, not a general setting. You can enable/disable streaming for every single sample or set a complete Preset or Bank to be streamed or not, either through menu (Edit --> Sample Streaming) or at the Audio-Pool (right mouse) of VSamplers "Sample" page.

You might ask "Why do I you have to care about all this stuff, can't it just work?" Well, a hardware sampler/synth has been designed for a defined hardware from the ground up, you get a guaranteed polyphony no matter whether your instruments use all features at once or not (exceptions prove the rule), e.g. 32 voices polyphony with an AKAI S3000. But with a software sampler/synth there is no defined hardware and no defined processing power, there isn't a guaranteed polyphony - the faster your CPU and the less complex your VSampler instrument, the higher the polyphony - it's in your hands!

The good news: VSampler is one of the fastest software samplers around, which has also been confirmed by several mag reviews in the past. The bad news: there are a few very "popular" ways to waste VSamplers power and CPU cycles for nothing. To give you a number what to expect, on a Pentium III 700 VSampler is able to playback around 80 stereo voices in Standard quality, when all instruments use filters and LFO's that drops down to around 40 voices in worst case. The voices indicator at the upper right corner shows the number of active voices.

Let's begin with the simple things, the general settings at the Preferences menu.

#1: The most obvious item would be Performance Settings. Using "Standard quality" saves 50% CPU against "Production quality", for most sounds you won't notice a big difference, and you can just re-activate "Production quality" before you do the final mixdown.

#2: Under New Bank/Preset Defaults check the Streaming tab. If you got enough RAM you should not use Streaming, enter a high MB number to not activate Streaming for "normal-sized" samples. Streaming means VSampler will play a sample directly from disc, without loading it into RAM before. While this saves RAM (obviously) it stresses your disc to the max and on most (IDE-)systems heavy disc load stresses the whole system as well. Your disc performance has to be shared with the audio tracks of your song as well as with the Windows swap disk traffic, and if your disc can't cope with the data stream your sound begins to stutter, while the CPU is not even at 10% maybe. Only use disc streaming if you don't have enough RAM to load the instruments into RAM. Anyway, VSampler offers a clever solution to save RAM without using Disc Streaming, be sure to read the "How can I lower VSampler's RAM usage?" question below.

#3: The most "popular" and most effective (=worst) way to waste CPU cycles are unnecessary long release times at the Volume envelope, a very common problem with imported SF2's and their logarithmic envelopes. The longer the voice is active the longer it will eat CPU cycles. Unfortunately the current VSampler uses a much to soft curve for logarithmic envelopes (will be changed in VSampler 3), it falls down very slllllowwwwww at the end, so the voice takes ages before it actually stopps playing (and using CPU). Example: you play a Piano phrase and think you hear 10 voices maximum, but VSamplers voices indicator (upper right corner) shows 60! After you imported an SF2 sound you should do a quick test for every instrument you use in your song: hit a single key, hold it down for a second and release it, now watch the voices indicator. If you do not really hear the sound anymore but the voices indicator still shows "1" that means wasting CPU - the release time is way to long.

But there's a simple and effective solution: use a linear release phase in the volume envelope instead of a logarithmic one and shorten the release times to around 1/4 of the times they had with logarithmic envelopes. Watch the animation below, compare the curve of the release phase (logarithmic vs. linear) and the number for the release time (marked yellow). Get the point? You can approximate the logarithmic curve with a line. Nobody will notice the difference in the mix, but the voices will stop earlier, which saves much CPU when you play a fast sequence.

For the SF2 import you can set the default to linear envelopes at New Bank/Preset Defaults, but that won't help much when not combined with shortening the release times.

#4: Many of the available SF2's have filters activated at very high cutoff frequencies and resonance set to 0. When played back at an SB Live it often makes no audible difference to a disabled filter, the filter has just been abused to "smooth" bad sample quality a bit. The SB Live hardware doesn't care about that, but your CPU does! A filtered voice takes around two times the CPU of a voice without filter. The rule: after importing a sound from an SF2 and it got got the filters enabled, check if you can hear a difference between filter enabled/disabled, if not, disable the filter! For the SF2 import you can set a default behaviour under New Bank/Preset Defaults, e.g. you could tell VSampler to not arm a filter when it's cutoff frequency is higher than 5 kHz and the resonance is below 10%, a nice "set and forget" setting.

#5: When your CPU is too slow to handle a certain instrument played in a fast and complex sequence (e.g. a fast Piano run), you should limit the polyphony of this instrument. Select all Splits of the instrument (upper right corner, set the Edit Group to "All Splits of Preset"), put them into a Split Group (same place, change "off" into "1") and limit the polyphony of this group from "Max" down to 5, for instance. Now VSampler will automatically stop the first voice as soon as a 6th voice starts.

You could save an edited version of the original soundbank and use one version for faster realtime playback and the other one when rendering the final mixdown to WAV.