Official Serum 2 thread!

[Igro]

Yes. But CPU load builds up from there. I can easily create a preset in Pigment, that would kill my CPU too.
Some of newer filters do affect CPU indeed.

My CPU isn't the neewest, but I can run multiple instances of heavy Phase Plant or Vital patches. There are two synths that kill my CPU with ease though - Zebralette 3 and Serum 2. There are some bass patches in Serum 2 that my CPU can barely handle with low latency, even when I play them in a monophonic manner. That seems a bit excessive to me.

Could you name me that patch? I want to try and report back.

Okay, I finally had a look after upgrading my system. Now on Ryzen 5800x f.ex. a bass patch called "Kitchen Sink" hits good 50 per cent of real time CPU use with a single voice, so on my old system that would likely be hitting the limits. I'm on 512@48k so not even that tight latency there.

Yepp, 50% on my system (FL-Studio meter). 50% of one core.

The preset creater used 32 modulation targets and 20 effects for that preset. He needs to go back to school and complete education from scratch.

I guess that’s why it’s called Kitchen Sink - probably created to test the limits of what Serum can do cpuwise

)). I think the name "CPU benchmark" would fit better.
...
Anyway, don't be fulled with presets when testing CPU load. It's just not worth it. If I would make such a preset in Arturia synth, that would be 100% CPU load in a second.

[nanostream]

They didn't even get an AMD to test on until after they shipped. It's not a testament to Apple, but to rushing it out the door.

It certainly looks that way.

The broader point I wanted to make is that this doesn’t seem to be a CPU issue in general. Serum 2 just doesn’t gel with AMD CPUs at the moment, and that’s it.

[Igro]

Are here any Intel laptop users to test the same preset (kitchen sink)? Cause I was testing on my laptop, AMD Rayzen 7 5700U.

[liquidsound]

On iMac late 2015 Reason 13.2
Kitchen Sink on poly and 7 notes no problem.
8 or 9 notes it begins to struggle.

[HcDoom]

Very weird they didnt test the code on AMD cpus till they released the plugin. Now its probably way to late to update /optimize anything or it will take a lot of time till we see some improvements.

[Echoes in the Attic]

Is there a way to save the UI size after adjusting it? It always opens up again at the default smaller size (Ableton).

[liquidsound]

Is there a way to save the UI size after adjusting it? It always opens up again at the default smaller size (Ableton).

In the same dropdown menu where you pick the size, at the bottom you can set it as a default as well.

[kraster]

Very weird they didnt test the code on AMD cpus till they released the plugin. Now its probably way to late to update /optimize anything or it will take a lot of time till we see some improvements.

Is there an actual source for this.

The closest I can get to something like this is Dave Gamble on gearspace saying that it was mostly profiled on Intel machines and that he'd check it out on his own new Ryzen machine and that he wasn't too worried about it.

That's not the same as the claim that it was never tested on Ryzen.

[jrwaltb]

Very weird they didnt test the code on AMD cpus till they released the plugin. Now its probably way to late to update /optimize anything or it will take a lot of time till we see some improvements.

Is there an actual source for this.

The closest I can get to something like this is Dave Gamble on gearspace saying that it was mostly profiled on Intel machines and that he'd check it out on his own new Ryzen machine and that he wasn't too worried about it.

That's not the same as the claim that it was never tested on Ryzen.

Each processor has different mico-code features that it optimizes and for some reason either AMD has not optimized a proc feature that Serum 2 uses or Serum 2 has not been able to optimize for the AMD specific mico-code. The Dave Gamble quote is true as I have read it and he probably means that whatever the CPU issue is (if it actually exists with AMD I don't know personally) its worth it because of the resampling abilities that everyone should be doing regardless of CPU usage. On my M1 with 16 gb ram its just like phaseplant but has MUCH better filters and sampler/synth engine.

[nusound mind]

fwiw reiterating what I said earlier in the thread; one of my cpu's uses an older Ryzen chip & Serum 2 is not noticeably different relative to other resource heavy plugins. I use it regularly, just tweak this and that here and there as/if called for.

[vitocorleone123]

Very weird they didnt test the code on AMD cpus till they released the plugin. Now its probably way to late to update /optimize anything or it will take a lot of time till we see some improvements.

Is there an actual source for this.

The closest I can get to something like this is Dave Gamble on gearspace saying that it was mostly profiled on Intel machines and that he'd check it out on his own new Ryzen machine and that he wasn't too worried about it.

That's not the same as the claim that it was never tested on Ryzen.

It's not the same that it's claimed that it was. In fact, it implies it was not since, presumably, the other platform it was profiled on was Apple. Finally, even if it was tested on AMD, it mostly was not.

Never mind the performance of the software....

[tumface]

Finally, even if it was tested on AMD, it mostly was not.

Never mind the performance of the software....

It works fine on AMD. AMD uses the same instruction set as Intel. They are compatible. AMD is as popular or more popular than Intel for new PCs now. (At least higher-end stuff.) Serum 2 had a semi-public beta test. It's silly to claim that Serum 2 was never tested on AMD. A bunch of people tested the Serum 2 beta, and many of them have AMD systems.

[vitocorleone123]

Finally, even if it was tested on AMD, it mostly was not.

Never mind the performance of the software....

It works fine on AMD. AMD uses the same instruction set as Intel. They are compatible. AMD is as popular or more popular than Intel for new PCs now. (At least higher-end stuff.) Serum 2 had a semi-public beta test. It's silly to claim that Serum 2 was never tested on AMD. A bunch of people tested the Serum 2 beta, and many of them have AMD systems.

Do you have performance comparison data? I don’t. So we just don’t know.

However, there’s anecdotal information (not data) that

1) Apple may perform better than Intel or AMD
2) Intel may perform better than AMD

Until someone does a proper, rigorous test, no one knows.

Anyway, I bought it and am using it. I hope they optimize it further, like uhe did for Diva etc.

[kraster]

Very weird they didnt test the code on AMD cpus till they released the plugin. Now its probably way to late to update /optimize anything or it will take a lot of time till we see some improvements.

Is there an actual source for this.

The closest I can get to something like this is Dave Gamble on gearspace saying that it was mostly profiled on Intel machines and that he'd check it out on his own new Ryzen machine and that he wasn't too worried about it.

That's not the same as the claim that it was never tested on Ryzen.

It's not the same that it's claimed that it was. In fact, it implies it was not since, presumably, the other platform it was profiled on was Apple. Finally, even if it was tested on AMD, it mostly was not.

Never mind the performance of the software....

The fact that it was mostly not tested on AMD makes sense since intel is still more dominant and would have been more so throughout development. Even Intel Macs are still in use today. SO that's not a surprise.

Definitively claiming something was "never" tested on AMD based on one vague forum from one developer is a bit strong, imho.

He also said "profiled" as opposed to tested which is more to do with analysing how the code performs on a given architecture. Testing also includes beta testing which is more about real world user experience.

We don't have even have anything definitive about performance. Lots of anecdotal information sure but that includes lots of edge cases like someone using 16 times Unison on a granular oscillator and complaining that it's eating up their CPU!

You'd need a database of like for like benchmarks to get a clearer picture.

From my own usage, performance is about where I'd expect it given the complexity of certain patches.

Having said that, I do feel like the granular oscillator is heavier than the equivalents in other synths.

[tumface]

Do you have performance comparison data? I don’t. So we just don’t know.

Yes. I posted some in this thread and elsewhere. And I've done other testing as well. Serum 2 performs 1.5x-3x worse on my Zen 1 Threadripper for the same presets as Serum 1, for presets which don't use unison. (Serum 2 has some kind of unison optimization that Serum 1 doesn't.) On my M1 Pro Mac, Serum 2 uses about the same or slightly less CPU for the same presets as Serum 1.