48 Comments

  • YAVAMAY studio

    November 5, 2022 - 2:35 am

    Hallo Woody can you do the test also for iOS iPad synth plugins please πŸ‘

  • Ric Holland UK

    November 5, 2022 - 2:35 am

    Top video Woody 😊

  • Jim B

    November 5, 2022 - 2:35 am

    Interesting video πŸ‘
    Really noticed how thin and weak sounding both versions of Massive are in comparison with these other synths, I’m really not a fan, used it for years and I’ve never got the sound I was after with it.

  • Genuinefreewilly

    November 5, 2022 - 2:35 am

    Its embaressing to admit how old my PC is, but I am a pretty heavy user of Vital and its a newer synth, and runs without much effort. The only vst sampler synth emulator that gives me any issues is Decent Sampler and sometimes Kontakt.
    Reaper is my main DAW its already pretty efficient . A faster PC would help as far as rendering and freezing tracks but there is a work around that, to play back tracks into another 'project tab', using loopback . This was covered in a recent Kenny Goia tutorial.
    I use it for all kinds of things now, a pretty powerful feature of Reaper

  • AERI

    November 5, 2022 - 2:35 am

    Not a very scientific test to be fair. Most people who hit high CPU usage have the wrong latency settings. You also need to factor in effects. Switching on a reverb per instance will greatly increase CPU usage. I don't own any of these synths, however, the synths I have, Omnisphere, REFX Nexus and VPS Avenger don't impact my Ryzen 7 at all.

  • Fried-egg-on-head chan

    November 5, 2022 - 2:35 am

    The REAPER performance meter doesn't seem to be accurate. It seems to just total the utilization of the entire CPUβ€”including all cores and virtual threadsβ€”even though most of these synths only use a single thread (Diva is the exception, if optionally enabled). Since synths and DSP for the most part don't generalize to multiple threads (which carry their own overhead to begin with) well at all, this vastly underestimates the actual workload and how much resources are available. In reality the actual CPU usage is, assuming the instances are processed in parallel, the highest load imposed by any of the 4 instances multiplied by the thread count, which is 12 in this case. I'm sure REAPER has a proper meter for measuring this, but clearly the performance meter is not appropriate.

    As far as demanding synths, the worst ones these days are rather Chromaphone 3, Arturia's Pigments, Kilohearts PhasePlant (depending on the patch, some use 80% of my Ryzen 2700X CPU just to play a couple extended chords), U-he Repro in the Prophet-5 mode when played polyphonically and Knifonium from Brainworx. All of these can easily use more than 50% of a modern CPU with a regular amount of polyphony.

    Especially Knifonium is incredibly resource hungry. I have a track work in progress where I use it for the bass, arp and lead sounds, and those 3 instances alone exhaust my CPU resources nearly completely (80-90% usage), and in practice due to other tracks and effects used I need to bounce those tracks into wave files for the song to be playable in realtime. And that's at 44.1 kHz sampling rate, so I cannot go any lower either.

  • John Nelson

    November 5, 2022 - 2:35 am

    I think U-He's claim shouldn't be an asset or anything to brag about. I use Arturia v collection 9 exclusively. And while some of those guys can eat up some CPU more than others, there were pains put fourth for accuracy and performance from Arturia's end. However, I've noticed performance differences with the same plugin based on what DAW I'm using. Studio one seems to have higher CPU usage than Bitwig, and I find Reaper sits in between both. Logic has always been the worst.

  • Saricubra

    November 5, 2022 - 2:35 am

    Arturia's upgraded emulations of vintage analog synths are CPU intensive but sound way better now than previous versions. Roland Cloud kinda is CPU intensive too, there is the physical modelling collection of SWAM audio.

    U-he Zebra 2 and UVI Falcon when pushed hard are intensive.

    When you take a synth and you use oversampling (specially to remove aliasing from FM synths) it's CPU intensive.

    Diva is nothing compared to the beasts above.

  • Hazy J

    November 5, 2022 - 2:35 am

    Oh, BTW- For those of us who use FL Studio, we got a performance meter too in v 20. It even breaks down each VST, as does this meter here in Reaper. I mention this should anyone reading this want to take a look at their system.

  • Hazy J

    November 5, 2022 - 2:35 am

    Wow Serum used almost nothing lol. Very cool. I wish you would have tried VITAL, though!

    Enlightening video, thank you. Very relevant to me specifically, as I use a Ryzen 5 3600 @ 4.2ghz.

  • XYKO since 92

    November 5, 2022 - 2:35 am

    good stuff man whats the name of that lead in sylenth 2

  • qasderfful

    November 5, 2022 - 2:35 am

    One instance of pre-patch Diva could take up to 45% of my i5 if I play a 7-chord with a complex enough patch.

  • K-Chill

    November 5, 2022 - 2:35 am

    I think they improved Diva at some point – I remember CPU usage dropped considerably after an update a couple of years ago. However, I wish preset designers for Diva would stop using such long release times… that really is a CPU killer when sessioning sounds,.

  • Ray Subject

    November 5, 2022 - 2:35 am

    regarding sound, for me unsurprisingly Serum far away best sounding.. it's wavetables inmolementstion ans filters are really ahead of it's time of release…

  • Andrew Southworth

    November 5, 2022 - 2:35 am

    I wouldn't say that it doesn't matter, its just not as much of a bottleneck as some people think it is. You don't necessarily need to get to 100% CPU usage before you start hearing little glitches and pops. Additionally the CPU usage of a plugin is greatly dependent on how many voices and effects are being used, and then your other effects on each track will add CPU usage as well.

    Its fairly common for songs to grow to 50-150 tracks in size depending on your genre and style. If we assume a song with 100 tracks has 50 synth layers, some of which have 6X unison, effects and plenty of polyphony, you can pretty easily start getting buffering issues in your track with a modern midrange computer. 100 tracks can easily have 100+ compressors and 100+ EQ's too, some tracks have none but its not uncommon to have a synth track with multiple stages of compression and EQ.

    You can of course just freeze tracks or bounce them to audio. This is why it doesn't matter in many cases how CPU intensive a plugin is. If you use your computer in a live situation and want a Diva track, people often just make a basic multi-sample of the preset they're using and use a low CPU-usage sampler to play it back. However freezing / bouncing can be annoying since it slows down the creative process, and as a result this is why we care about the CPU usage of synths.

    Nobody should skip using Diva just because it takes more CPU, but everyone should be aware of how CPU hungry their plugins are so they can make smart decisions along the way before they end up with a 100 track song their computer can barely run without bouncing everything to audio.

  • Almark

    November 5, 2022 - 2:35 am

    I thought it was just me: showing the front of the video of Diva. That thing was eating up processes back when I was running on windows 7, dual core.

  • Doug S

    November 5, 2022 - 2:35 am

    This music is too thick and lush πŸ˜‚ – thanks for doing this – answers questions VST users often wonder about.

  • tenchu Georg

    November 5, 2022 - 2:35 am

    Do not jump to conclusions about Real Time Computing Performance by using the Percentage of the Task Scheduler!

    Measure how long it takes for an ASIO Buffer to computer – AKA FRAME TIME
    Ie. Bitwig shows that graph in realtime.

    Sample Accurate VST3 Automation, 8x Oversampling on modulated Wavetables, Zero-Feedback-Filters with many voices DO still come at a high price.
    Just 1 single Serum Instance using wavetable mods and unisono can hit Frametime on your 8core system, and then all your 5% cpu usage is worthless.

  • Lee Pride

    November 5, 2022 - 2:35 am

    Roland Sound Cloud instruments!! Why??

  • James Pingel

    November 5, 2022 - 2:35 am

    I think it's worth noting that Diva is inline with the newer VSTs like DCO 106 and Massive X, but it came out in 2011. So comparing that to things like Sylenth and original Massive, I'm not surprised it still has the reputation of being a CPU hog, especially for it's time. Almost makes you feel like it's about time for a Diva 2, we've had it 2 easy for 2 long.

  • Serge Tarasov

    November 5, 2022 - 2:35 am

    I mean, turn up unison in serum and play some arps, 4x oversampling enabled. An easy way to choke a modern cpu. I can't do this with sylenth.

  • David Lincoln Brooks

    November 5, 2022 - 2:35 am

    In terms of beauty, the humble SYLENTH can really hold its own against these other behemoths…. Such a warm and fuzzy, authentic and precise sound!

  • WilkoPiano

    November 5, 2022 - 2:35 am

    There are still issues with disk types and ram speed. Minimum ssd and not external usb drives either, has to be internal connected. Which is why macs are crap only one nvme slot

  • WilkoPiano

    November 5, 2022 - 2:35 am

    The only time cpu burden is an issue is when multiple fx buses are being used or time based fx on VSTS e.g. NIs hybrid keys or noire piano particles engine. Apples macs are so expensive for equivalent intel pc specs people often go for second hand apples with less than the optimum 32 GB ram, and only 1TB disk. When in the pc world for less money you can get 64GB ram 2TB nvme drive just for plugins, another 1TB nvme for project saves. Timed based fx over midi are the killer. Cheap apple users are advised to bounce midi to audio to save cpu and ram. No need if you use pcs or are rich enough to afford a mac studio at 5k. For reference on my i9 i run 64 midi instruments with time based fx and 16 audio channels for hardware synths all capable of recording at the same
    Time and recording video direct to OBS via nvidia GPu backed recording codecs.Combining omnisphere and komplete kontrol.

  • Ann Other

    November 5, 2022 - 2:35 am

    One major thing many people dont realise is that the Windows default is a low-power mode so theyre not getting the full power they paid for!! Bloody EU legislation!

  • Subramaniam Chandrasekar

    November 5, 2022 - 2:35 am

    I bought a VST from the sixteen company. At that time they advertised |the only synth you will ever need" True enough, this overloaded my CPU on many presets and I was unable to load any other vst. Let me wait another few years to try again.

  • busywl69

    November 5, 2022 - 2:35 am

    Freeze functions in most DAWs (except the one I use) remedy the CPU problem with thirsty VSTs.

  • Eugene Epstein

    November 5, 2022 - 2:35 am

    Hungry=less optimized

  • brettdevme

    November 5, 2022 - 2:35 am

    Reaper is amazing at VST efficiency. So well written. Faster CPU, RAM & nvme SSD helps as well. Thank you computer games

  • Dima S

    November 5, 2022 - 2:35 am

    from my perhaps not super representative experience it is group processing in things like compression that eats the most cpu. nothing gets your project as much risk of stuttering as a chain of u-he satin and presswerk on master

  • Sevenfifteen Music

    November 5, 2022 - 2:35 am

    This test is not very telling. There's a lot not taken into consideration. For example, there's no difference in CPU load between 1 and 4 notes played simultanously. Because all plugins make use of SSE, which runs four channels for the price of one cpu instruction (instead of 1 instruction per channel). Only at the 5th simultanous note will there be a jump in CPU load, then again at the 9th, etc.
    Many synths have settings like "draft", "full" and similar. Those modes affect the way calculations are done, saving instuctions on the CPU (in case of "draft") and vice-versa. Since some don't have such settings, they should all be set to their highest quality. The mentioned "100%" are just hypothetically. Most DAWs will cut way earlier, around 80%. Because DAWs need CPU load as well. To test limits, you should make use of iZotope's Iris 2. It is not as old and its spectral filtering is pretty demanding. 3 years ago on an i3 I could barely run one instance. On my mid-range Ryzen 5 5600G it's clearly better, but still hungry for CPU. And lastly, some plugins make use of multiple cores (most do not). Those who do, have an adavantage over the others, so shouldn't be directly compared.
    But those tests never cover a real world scenario. My projects typically have like 7 or 8 synth tracks, 1 or 2 drum machines, and on average 2 fx per track. With such projects the occasional peaks (which are more important than the average usage) go as high as 50% on my PC.
    And that brings us to a point, that is neglected, but shouldn't. The DAW used. Reaper is very optimized CPU-wise, and is therefore a rather bad test environment. Try the same tests with different DAWs, and you will be surprised.

  • wrigna1

    November 5, 2022 - 2:35 am

    It's absolutely 100% an issue. No need to waste time making or watching a video on the topic.

  • Marklar3

    November 5, 2022 - 2:35 am

    I've done a little experimentation of my own (on my i5 4-core @3.8GHz), mostly between SurgeXT and Vital (both free, but Surge is open source).
    I've found that the amount of notes you're playing and the specific patch you're using have a huge influence on performance. If I play one note on the Surge patch called "CPU hog" by Vospi, it's 2.5% cpu usage, but if I play 10 notes, the CPU meter hits 12-14%. You have to also keep in mind that it's not really that hard to exceed 10 note polyphony if you're using a slow release time or a sustain pedal. If I move from one 10 note chord to another, it's processing 20 notes at once, and it starts to glitch, hitting 20-25%.

    That is obviously an extreme case though, considering the patch is called CPU hog. Most of the pads on Surge run at about 1-2% on single notes and 3-4% with 10 note polyphony. For Vital, most pads run at 3-4% on single notes, and 7-8% on 10 note polyphony.
    It seems like layers and modulation have a feedback effect with polyphony. If you increase both at the same time you'll use exponential amounts of CPU.

  • Rua Whitepaw

    November 5, 2022 - 2:35 am

    In my experience, Roland plugins are by far the worst offenders when it comes to CPU usage. Adding just a few instances of one is enough to cause my sound to stutter, which kinda defeats the point.

  • Mad Mohawk Films

    November 5, 2022 - 2:35 am

    Thanks this is quite illuminating.

  • Old Guy and his library

    November 5, 2022 - 2:35 am

    Interesting video, but somehow, I miss an explanation of cpu measurements in DAWs and perhaps a comparison between what the DAW displays vs the actual system load. THings like usage of multiple cores, dependency between ram transfer rate and cpu load, ASIO driver issues are just being neglected here. So, for me , this is an interesting viewpoint, but ultimatively falls short.

  • Sound Author

    November 5, 2022 - 2:35 am

    Diva can still eat up a lot of CPU (yes, even on my M1 Mac mini) depending on several things: "Accuracy" modes such as "Divine", whether or not you're using the "Multicore" mode, the number of stack voices selected, and one of the most overlooked things: the BITE highpass filter, which is actually THE most CPU intensive module in Diva. But there's good news! U-he's new CLAP plugin format takes advantage of something called "thread pooling", which makes the Multicore mode SO much more efficient now. But there are still a few bugs in the beta versions of u-he's CLAP plugins, so we gotta wait for a bit 'till they are ready for primetime.

  • oldskoolfunkandsoul1

    November 5, 2022 - 2:35 am

    also the ram is just as important,..16gb ram i minimum.. πŸ™‚

  • oldskoolfunkandsoul1

    November 5, 2022 - 2:35 am

    What is the model and year of your ryzen pc?/..still very low cpu levels..similar to my 4 core itel fro 4ears ago in studio one.

  • Prod.3xnd1t

    November 5, 2022 - 2:35 am

    Maybe get a better rig.

  • John A. DiMeglio

    November 5, 2022 - 2:35 am

    you could run these on a cell phone then?? HAHA! Triumph of the dream!

  • KozmykJ

    November 5, 2022 - 2:35 am

    Watching people's CPU capacity running out makes me think otherwise.
    e.g. Data Broth, Polarity Music, Venus Theory.
    I'm inclined to regard your testing as inconclusive.
    Granted, a lot can be taken up by FX rather than synths on their own.
    And, Yes, one can apply strategies to circumvent CPU overload.

    If you want to try a Real guzzler take a look at Odin 2.

  • thesrabbit

    November 5, 2022 - 2:35 am

    Wonder how the ACB Roland plug-ins compare

  • Saar Dean

    November 5, 2022 - 2:35 am

    Wow i think i am falling behind. Not that i "need" a new laptop but my 2019 Mbp 6 core has a lot (a lot) more Cpu load when i use similar amount of plugins.
    I remember once i loaded the Memorymoog emulation by Cherry Audio and with some presets i saw 20-30% Cpu load with 1 instance….
    My settings are most time 48khz and Buffer 1024.Β 
    Also before this i had a Mbp of 2015 with an i7 Quadcore and 1 instance of Multiplicity was almost sufficient in order to bring it to its knees. On the 6 core it works just fine.
    I wonder if it has to do also with the Sowtware Platform. Your Pc seems to be flying. Nice
    However,, as i said. The times i used 100+ tracks in a project are long gone and now i achieve the same level of "satisfaction" with 8-10 tracks.
    As time goes by Computers get faster and faster but if you are lucky thank God so is your Mind`s efficiency

  • Agents of Rush

    November 5, 2022 - 2:35 am

    Used to have to bounce everything but got a new PC last year with an AMD Ryzen 9 5900X 12-Core Processor and can now run Diva etc fine.

  • piano stuff

    November 5, 2022 - 2:35 am

    I donβ€˜t like this video. It showed me how old my computer is πŸ˜‚πŸ˜‚πŸ˜‚

  • vis cuz

    November 5, 2022 - 2:35 am

    I think it depends on whether you are working on a desktop or a laptop. CPU is of no consideration if you make music on a desktop, but it definitely matters if you're a bedroom producer working on a laptop.

  • 1000sofusernames

    November 5, 2022 - 2:35 am

    Every time I upgrade there's still a DIVA preset that starts crackling and popping. One day I'll be able to use every thing in it. One day. I normally end up bouncing to audio clips. I've also got a decent PC as I sim race and need a decent CPU.

Comments are closed.