New Wurly offering by Cherry Audio

Hi Ade,

The details on my rig’s time load (Intel i7 2.9 GHz 4 core @44.1KHz sample rate 32 GB RAM) is aprox 30% with it rising to over 100% when I hold the damper down and keep adding notes while it sustains. It tends to increase on the tails of the longer samples (lower range notes that have more sustain). The upper range doesn’t have a long decay so it doesn’t appear to be the issue. I was easily able to use it if I minded my manners on the sustain use. Joop must have a hell of a rig running at 192 KHz and 20% load. :slight_smile:

Dave

1 Like

I reported my rig here:

Does not seem to be so special, …

And your ASIO buffer setting?

128 samples

I noticed if I turn off the reverb, the time load goes way down.

For most of these virtual pianos, I have to mess with the velocity curve. I don’t usually use compression. Real electric pianos have a very specific keyboard action that causes the “bark.” On a Rhodes, you just boost the bass.

  • Paul

And that’s the big difference.
I also run at 128 - and the stuff that sweats at that setting compared to 256 …. Well you know.
192?
Well, it’s in da middle, ain’t it.

I’m running 128 samples on an 11Gen i7 8 core machine.

Wurliebird’s CPU load does oscillate a bit - typically between 2.5 and 6.5 average, 8 to 20% peak in Profiler. Looks like it doesn’t really have the smoothest resource usage - maybe some waiting for disk I/O for sample loading involved, I guess.

Some funky maximum values like 16106037.5% in some profiler snippets, while the overall engine cycle stays at 35% - not sure what that indicates. But that happens across most plugins in those specific cycles - not sure if that’s a Cantabile bug? No dropouts or crackles noticeable, so I assume that’s a reporting issue, not a “real” buffer throttle…

@brad: happy to provide a profiler log for diagnosis if that helps.

I typically stay away from the plugins’ internal reverbs - prefer having my “send-type” reverbs at song level to have things nice and even. But even adding copious amounts of the Wurliebird’s “Galactic” reverb didn’t really get me into glitching territory. In fact, the CPU footprint of the additional reverb wasn’t really very big - unlike @bartok2112 's experience.

Guess my systems are just a bit more robust than others. I built my mini-PCs specifically for live use with Cantabile, and I try to keep them very clean and my setups as lean as possible.

Yea, good point, it must be taken into account, my bad.

FWIW they dropped an update just today (1.0.12) but I can’t find any version change info.

One thing I noticed is that the more I compressed the Wurlybird, the more apparent it became its made with 5 sample layers (as best I can tell). Under a good amount of comp it was basically like flipping a switch between a single velocity change from (iirc) 91 to 92 because that is a place where the sample changes. Its not a problem with no comp though. That issue aside, to me it sounds more realistic without comp or with just a very small amount anyway.

That seems excessive and does sound like a reporting issue. Can you send me a profiler log showing this and I’ll check it out.

The fact there’s no audio drop outs can happen if the audio driver/hardware has a layer of buffering somewhere in the mix.

Brad

Hey Guys,

I dropped a note to the support desk at Cherry Audio and they promptly and politely replied regarding the system load that WurliBird 140b creates when adding notes while sustain is down and also regarding where the sample pack is saved and streamed from. Here is the reply, it was informative for me.

Hello Dave,

Thank you for choosing Cherry Audio!

The performance of Wurlybird will depend on a few things, including sample rate, buffer size, CPU speed (and thermals, bus speeds, etc.), interface and drivers, oversampling (if applied), and if effects are being used. Some electric piano plug-ins use physical modeling or limit voice counts that reduce the CPU load. Wurlibird has 64 voices of polyphony, which can add up when a pedal is held and notes are stacked. I tested on an older 2.2 GHz i7 and was able to get between 16-24 notes of polyphony at a 48K sample rate and 256 sample buffer before performance issues arose. With that said, there can be a significant variance in PC performance and the speed of a processor is only one factor.

As far as the samples go, the .sar file should remain in the default location **%programdata%\CherryAudio\Wurlibird 140B**.

Hopefully, this information helps.

Best Regards,

Danny

2 Likes

Interesting. I have both Neo Souls Keys and Rhodes own V8 pro, both of which have a huge sample library (especially the Rhodes V8 which took an entire day to download). Neither gives me any glitches, though I think the Rhodes utilizes physical modelling also. Those two sample sets do allow one to choose the location of the samples though. I think Windows sometimes has issues with stuff stored in the “program data” folder. The U-He chat board has some mention of this, and the newer U-He plugin versions now no longer allow you to install in the \programdata\ folder on Windows.

  • Paul
1 Like

Interesting, the spikes seem to have mostly gone away with the new version 4196. I only tested it a little bit though.