You load one pad, play a chord, and the meter jumps to 70 percent. If that sounds familiar, you’re not alone. High Serum 2 CPU usage is the most common complaint I’ve heard since the update landed, and it’s a fair one.
I make Serum presets for a living at Monosounds. Every patch we ship has to pass a CPU check on a mid-range laptop before it goes into a pack. So I’ve spent an unreasonable number of hours making heavy patches light without killing the sound.
Here are 12 fixes, ranked. The first four solve most problems. The last one fixes a weird GPU bug that looks like a CPU problem but isn’t.
Why Serum 2 CPU Usage Runs Higher Than Serum 1
Let’s be honest first: Serum 2 is heavier than Serum 1 on some patches. Not all of them. A plain wavetable patch with sane settings costs about the same as it did in 2020.
But the new stuff adds up fast. You now have 3 oscillators instead of 2, each switchable between 5 engines. Granular can run up to 256 grains at once. Spectral resynthesis rebuilds sound from individual partials in real time. And the FX section went from one chain to multiple buses with per-effect modulation. That’s a lot more DSP than the old engine ever asked for. I broke down the full differences in Serum 1 vs Serum 2.
Xfer knows this. They publish official CPU guidelines on their support page, and the fixes below follow the same logic, plus what I’ve learned building presets on the new engine since launch.
The Big Four: Voices Are Everything
1. Cut Unison to 3-7 Voices
Unison is fix number one because voices multiply everything. A patch with unison 16 runs the oscillator, the filter and every modulation source 16 times per note. Play a 4-note chord and you’re at 64 voices before the FX even start.
Drop it to 3 on bass, 5 on leads, 7 max on huge supersaws. With detune around 0.10 to 0.15 and Width pushed up, unison 5 sounds about 90 percent as wide as 16 at a third of the cost. Almost none of the bass presets we ship use more than 3.
2. Fake the Width With Chorus Instead
Here’s the trade most people miss: chorus is cheaper than unison. A chorus on an FX bus processes the summed signal once, no matter how many notes and voices you play.
So run unison 3, then add chorus for the smear and stereo spread. On our supersaw patches this one swap cut CPU by 40 to 50 percent, and in a blind test nobody on the team could reliably pick the unison 16 version.
3. Cap Polyphony
Set Serum’s poly limit to what the part actually needs. A bass needs 1 or 2 voices. Keys need 6 to 8. Pads maybe 8 to 12. The default limit is far higher than most parts ever use, and every allowed voice is a voice that can pile up.
Released notes count against polyphony too, which brings us to fix four.
4. Shorten Release Times
Every note you let go keeps eating CPU until the amp envelope fully closes. An 8-second release means every played note lives 8 extra seconds. Stack a few chord changes and you’ve got 30 invisible voices running behind the music.
Cut the amp envelope release under 2 seconds and let a reverb supply the tail instead. Reverb cost is fixed. Voice tails scale with your playing.
Engine and Quality Settings
5. Oversampling: 2x, Not 4x
Oversampling raises the internal processing rate to reduce aliasing, and it multiplies CPU right along with it. 2x is the sweet spot. 4x roughly doubles the cost again for a difference you’ll mostly hear on bright, high-pitched patches in isolation.
My rule: play at 2x, and if a lead sounds gritty up high, render that one part at 4x. Nobody has ever emailed me about aliasing in a finished track.
6. Watch the Granular Grain Count
The granular engine can run up to 256 grains, and every grain is a tiny playback voice with its own envelope. 256 grains under a unison stack is a CPU fire.
Most granular textures sound the same with the grain count and density kept modest. And if you’re going granular, keep unison at 1. The engine already gives you thickness for free.
7. Bounce Spectral Patches When You’re Done
Spectral is the heaviest engine in Serum 2. It’s resynthesizing hundreds of partials in real time, which is amazing for sound design and expensive for playback.
Once the sound is right, use Resample to Oscillator to bake it into a wavetable, or bounce the track to audio. Same sound, a fraction of the cost.
8. Turn Off What You’re Not Using
Every active module costs something, even when its contribution is subtle. Second filter barely doing anything? Off. Third oscillator mixed at 5 percent? Mute it and see if you actually miss it. An effect sitting at 3 percent mix? Delete it.
This boring pass reliably shaves 10 to 20 percent off bloated patches. It’s the first thing I do when tearing presets apart, and skipping it is one of the habits I call out in Serum 2 beginner mistakes.
Project-Level Fixes
9. Simplify the FX Buses
Multiple FX buses are great until each one carries its own reverb. Reverbs and the Bode frequency shifter are among the priciest effects, and per-effect modulation adds more on top.
Keep one reverb per patch. And if three patches in your project all want the same big hall, put one reverb on a DAW send instead of one inside every Serum instance.
10. Raise Your Buffer Size
This one is free. While recording, 128 or 256 samples keeps latency playable. When mixing, push the buffer to 512 or 1024. Serum 2 CPU usage per patch stays the same, but the crackles disappear because your computer gets more time per block.
11. Freeze and Bounce
The oldest trick still wins. Once a Serum part is final, freeze the track or bounce it to audio. An audio file costs near zero no matter how insane the patch behind it was.
I bounce anything that survives two sessions unchanged, and I keep the MIDI plus the preset name in the track notes so I can always go back.
12. The Serum.cfg Fix for GPU Glitches
Some CPU problems are actually graphics problems. If the interface stutters, your meter spikes only while the plugin window is open, or the visuals glitch, try disabling OpenGL.
In Serum 2, go to the Menu tab and choose Open Serum 2 Presets folder. Inside, open the System folder and edit Serum.cfg in a text editor: set UseOpenGL = 0, save, and restart your DAW. On machines with flaky GPU drivers this kills spikes that no synth setting could touch.
And if the plugin ever vanishes from your plugin list after system tweaks like this, here’s how to fix Serum 2 not showing up in your DAW.
Heavy vs Lean: The Cheat Sheet
This is the checklist I run when a patch fails our CPU test:
| Setting | Heavy patch | Lean patch |
|---|---|---|
| Unison | 16 voices | 3-7 voices |
| Width source | Stacked unison | Chorus on an FX bus |
| Polyphony | Default, uncapped | Matched to the part (1-12) |
| Amp release | 6-8 seconds | Under 2 s, reverb carries the tail |
| Oversampling | 4x everywhere | 2x, render 4x if needed |
| Grains | 200+ | Modest count, unison 1 |
| Reverbs | One per FX bus | One total, or a DAW send |
Notice that none of these changes touch the character of the sound much. The CPU lives in voices, tails and oversampling. The character lives in the wavetable, the filter and the FX choices.
This is exactly how we build at Monosounds. Every patch in our Serum 2 preset bundles has to pass on a 2019 laptop, and you can inspect the approach for free: open a few patches from our 250+ free Serum 2 presets and check the unison and poly settings for yourself.
Serum 2 CPU Usage: Common Questions
Is Serum 2 really heavier than Serum 1?
On some patches, yes. Anything using granular, spectral or several FX buses is doing more math, so it costs more. Plain wavetable patches are close to Serum 1 territory. Since the upgrade is free for Serum 1 owners, the CPU tax is the only real price of admission.
Does 2x oversampling sound worse than 4x?
In a full mix, you will almost never hear the difference. Aliasing shows up on bright, high-pitched material played solo. If a specific lead bothers you, bounce that one track at 4x and leave everything else at 2x.
Why does one patch use 40 percent CPU on its own?
It’s almost always voices: unison times polyphony times release time. Check those three before blaming the synth. A unison 16 patch with a 6-second release can quietly run 80+ voices from a simple chord progression.
Will a faster computer fix Serum 2 CPU usage?
It raises the ceiling, but voice discipline scales better than hardware. I’ve watched a capped, well-built project run 40 tracks on an old laptop while an undisciplined 12-track session choked a new one. Fix the patches first, then upgrade if you still need to.
Start With Unison, End With Bounce
Work the list from the top. Unison, polyphony, release and oversampling fix 80 percent of cases in about five minutes. Freeze handles the rest, and Serum.cfg handles the machines where it was never a CPU problem at all.
If you’d rather start from patches that already pass the test, grab our 250+ free Serum 2 presets. No strings attached. Tear them apart, check the voice counts, and reuse whatever tricks you find.
I run Monosounds.studio and make the packs myself. Ten-plus years of producing, 97+ releases, 16,000+ downloads. Every Serum 2 preset we ship passes a CPU check on a mid-range laptop and gets used in a real track before release. If it does not survive a mix, it does not go in the pack.
More about Maxim & the studio →
