CPU Performance vs. Real-Time Performance in Digital Audio

I took the topic title from this 9 years old video:

despite the video's age all the new comments are like these:
"2024 and this 8 years old video got more quality than 90% of todays' videos"
"Great video. A must watch if you have a DAW."
"this video is a gem, thank you!"

So, what is this video about? You have a power CPU, a lot of RAM, SSD and bla-bla-bla, but despite this you still have spikes, cracles and audio dropouts...
so, watch this video to know whay it all may happened,
and share your experience

 

With all due respect to the video but he has no clue into the depths of the issue.

1- Consumer grade CPU's are not meant for real time audio
2. Consumer grade OS from Microsoft and Apple are not meant for real time audio
( Although Mac OS internal audio runs more tight with the kernel)
3. Go back to number 1

We do not have a Real time OS for Audio work running on our consumer computers, so this means the OS itself
must be setup for extremely running with barely anything on.

With consumer grade OS we have the issue that it's a daily "all in" machine so that means it has to run as many services
in order for all your games, office applications, network, system tools to function properly.

This means that we need to disable almost everything that is not needed for audio as the extra services run cpu cycles and also in ram memory as well. This is not a concern on the memory side since we have lot's of it today but cpu spikes YES it is a concern.

If our computer is not a ZERO CPU during idle ( with no applications running) in task manager you need to do some work on that OS
and that includes Mac OS as well not just Windows. All computers period end of story should be at zero CPU and ZERO drive hits during idling this is not a option it must be that way.

You needs to disable dozens of useless services which are not needed when working with a DAW

Windows 10 and 11 include more services than previous windows versions which now are tied to others this means that some services will be a challenge to disable since one service might take down an important system service. This also applies to Mac OS as well as it now has a stack of services a page long.


The bottom line is CPU spikes from non essential services must be stopped this is what leads to pops when running lower latencies.
There are limits to what our machines can do after all they are not really dedicated for Audio Only, since they also play games, office suites, online connections etc.

Learn the Services in Mac OS and Windows that is the key to a great setup.

 

What is a consumer grade CPU? Are you implying only Ryzen EPYC and Intel Xeon CPUs can do audio work? Because I can assure you they don't. I used to set up computers (servers) that cost $2000+ an hour to run. Servers with 128 cores, more RAM and storage you could ever find a way to use. They definitely don't perform audio any better than a consumer grade Ryzen 9800X3D.

 

 

Last edited: Jun 21, 2025

Spoiler: Troubleshooting DPC latency

Troubleshooting DPC latency

Overview

DPC (Deferred Procedure Call) is the operation that Windows uses to assign a priority to processes/drivers that run at the same time in the same system. If processes that aren't involved in streaming audio take up processing time for longer than they really should then this can prevent audio drivers from being able to send/receive data in an appropriate amount of time, leading to audible issues (e.g. pops/clicks, dropouts, 'glitchy' audio and sometimes even device disconnections).

A common cause for DPC latency is out of date device drivers and Windows processes that are not optimised correctly. Many processes/drivers are involved in streaming audio, and many other processes/drivers can cause interruptions in the audio stream.


Analysis
To analyse whether DPC latency could be the cause for any pops, clicks, or disconnections you might be experiencing, you can run the following software tool: Latency Mon (Windows 7 and later).

To run the tool, click the Play button and then play audio from any application through your audio interface for a couple of minutes. If there are no DPC problems, the scan will report this:



If there are problems, the reporting text will be black/red and the drivers/processes that are the likely cause of the problem will be displayed:



Solving DPC Latency issues

Updates for some processes/drivers are often delivered via Windows Update. It's recommended that you ensure that your version of Windows is fully up-to-date should you encounter any DPC problems, you can update Windows by following these steps:

• Windows 7: Go to Start > Programs > Windows Update > click 'Check for updates'
• Windows 8: Go to Control Panel > Windows Update > click 'Check for updates'
• Windows 10/11: Click the Windows icon in the bottom left corner of the screen > Settings > Updates and Security

If you use an Intel-based system, you can use the Driver Update Support Assistant to find driver updates for your system. This tool can be downloaded here.

If the DPC problems persist after installing all Windows updates, then the next step is to try to deduce the device that's causing the problem. Common problematic areas are:

• Network/Wi-Fi adapters
• Card readers
• Other sound devices that aren't in use
• Bluetooth adapters
• Graphics card

To narrow down which device causing the issue, you can try disabling the above components in Control Panel > Device Manager (only disable your graphics card if you have on-board graphics as well) and then run the DPC test again. If this fixes the problem then this would be a strong indicator that this particular device is the cause of the issue - you can then either leave this disabled when working with audio or check for any updates from the component manufacturer.

Source: https://support.focusrite.com/hc/en-gb/articles/208360865-Troubleshooting-DPC-latency

How to FIX High DPC and ISR Latency in 60 Seconds

 

With all due respect for your answer, but...
1) You said that consumer CPUs aren't designed for music production, which isn't true (it used to be that way). Consumer CPUs are designed to handle all tasks, and with features like SSE and AVX, consumer CPUs are even better suited to running plugins/DAWs (which, of course, must also be used by the plugin/DAW). A good audio interface is recommended for this, but not essential.
2) You say that consumer operating systems from Microsoft and Apple aren't designed for real-time audio. Steinberg, for example, invented ASIO for Windows PCs, and ASIO Guard in Cubase. It's worth noting that today's CPUs have a fairly large Level 3 cache, which also provides an additional boost. Some CPUs have the 3D V-Cache, which also adds a significant amount of cache, but slightly lowers the clock speed. And, as we all know, clock speed is more important than cache for some DAWs, so you'll have to look at it.
In addition, today's CPUs are so fast that you can hardly use them to their full capacity with the DAW and plugins.

I use a translator from German to English.

To show what audio power the Ryzen CPUs have, here is a short video.

 

I never disable any "useless" services in MacOS for audio, the way people do for Windows. I would be very skeptical of anyone claiming you "need to disable" any services in MacOS for audio production. Anything I even remove after a fresh MacOS install is just apps I never use, but only because they waste space on the system disk.

Out of every process running, only 22 of them are even registering higher than 0.1% cpu use; and those are things I installed. In fact, other than Logic; the instance of Activity Monitor I am using to check this is the most CPU usage-heavy process I have running.

Someone trying to tell you that you cannot just buy a new mac, install Logic and your audio interface, and be up and running without doing a bunch of system tweaks has some kind of their own agenda. Any Apple Silicon mac will not even blink due to "useless services".

 

1.Did you see the video?
2. yes, it's 9 years old, and you state "To show what audio power the Ryzen CPUs have"...hm, ok, just few days ago we had a topic exactly with this build:
OS: Windows 11 Home
DAW: Cubase 14 Pro
I/O: SSL 2+
CPU: AMD Ryzen 9 5950X 16-Core 3.4GHz
MOBO: ASRock X570 Taichi
GPU: NVIDIA GeForce RTX2080Ti
RAM: 128GB 2667Mhz (4 x 32Gb Corsair Vengeance PRO DDR4 360)
Cooler: Be Quiet Dark Rock PRO4
C Drive: Samsung SSD 980 PRO 1TB
Project Drive: Samsung SSD 850 PRO 128GB
VST Drive: Samsung SSD 970 EVO 1TB
And the owner still complained about crackling in Cubase:
https://audiosex.pro/threads/good-system-on-paper-but-ongoing-cubase-performance-issue.80349/

 

A PC consists of thousands of components, dozens of programs, and then there's the operating system. If just one of these components is faulty or something is misconfigured, there will be errors or, as in your example, crackling.

I'd say that PCs even run and function as they do is actually a miracle. But if a user makes mistakes, or there are program errors or hardware failures, finding and fixing them is sometimes very easy or very difficult.

I think we've come a long way technologically, and videos like the one above from 9 years ago should be updated to reflect today's state of the art. A lot has happened in 9 years, especially with CPUs, RAM, etc., and it's for everyone's benefit.

 

This whole thing is funny, even on the Windows side. My production PC runs a 5600X that is technically very entry level for today's standards, it also has a Windows installation that is over ten years old, is full of services and shit running because I'm also a developer and is of course, fully online. I don't have any issues working with audio whatsoever. I can compose with low buffers, and I mix with high buffers. Maybe I'm just too good with computers without not even trying to.

 

And one huge step: you can freeze as many tracks you wish and it will be less time/storage spending than recording on tape. You can overdub the things a zilion times with high quality processing going on without Nasa computers. For a long time people didn't understand how Skrillex could generate that sort of sounds and modulations and still had any space to work with (computers were weaker back then). The short and rude answer: he was printing stems of everything. Plus, you will lose all the worry about plugins backward compability: you will only need any DAW and you can open your project normally, without huge losses.

 

Agree, most of those background processes are dumped on efficiency cores and DAW's fully use performance cores. Also agree on your other point, Mac/macOS is ready for audio from the get go, it's designed for that target audience, even if it has bells&whistles for ordinary folks.

Not saying you must tweak Windows to work good for audio, last Ryzen build I made worked perfectly for audio with official Windows version without any services disabled, without messing with BIOS and just setting power plan, but it's not like I threw most random parts in it and hoped for the best or Windows isn't really capable of anything if you don't disable telemetry and mess with whatever. It can work even better if you nail hardware and software aspect, especially if vendors have proper drivers in place, some solutions seek tweaking, some combo's are more suitable then others, some setups and DAW's too. It's not uncommon that someone have USB problems with some boards, things causing DPC, vendors bloated drivers, ultimately some solutions aren't targeted for most optimal low latency audio folks in mind.

Also some folks would rather DIY then pay HP, Dell, SCAN or Apple for computers that are targeted to them, so each to their own... it's also bumpier ride when you are also gamer and whatever, privacy concerned on top of everything, so expect one machine to be everything without any compromises while you even try to cut the corners as much as you can on the way.

 

Last edited: Jun 22, 2025

My experience is I don't, and I'll tell you why I think I don't:

• First, I switched to Mac (just as I'd switched to PC from Atari) when Logic switched platforms. No more turning off services, and digging in the registry for obscure or arcane keys to gain 0.5% more computing power. Macs may seem like consumer bla-bla-bla to you, but Core Audio is a mature, and continually evolving API, designed specifically for audio. No ASIO, no drivers, needed. True plug and play. I don't even think about it. And I've no problem working with PCs if need be - I just don't set them up for audio.
  
  "Back in the day, it felt as though you had to be an elite hacker just to route audio from one program into another, or install a soundcard driver. That began to change back in 2003, when macOS 10.3 quietly introduced a new API for sound handling. Core Audio has since undergone more than 20 years of continuous development, and is now a mature technology. It’s a remarkable and underappreciated achievement, taking care of everything behind the scenes with effortless grace."
  - Source

• Second, I max out my Mac. Max RAM, max HD or SSD. Simple rule of physics I learned from flying: the higher the altitude, the farther you can glide - same applies to computers, but also, the roomier your SSD (say, 8-16TB), the faster your computer.

• Third, and most importantly, I'm old school - analog school. I started out with just a couple of reverbs and a digital delay, one compressor, one EQ. I learned to squeeze every electron from the little gear I could afford.
  
  In plain 2025 DAW English, it means I seldom slap on plugins on individual tracks (only on bass, and vocals - sometimes on piano).
  
  Everything else is bussed or handled at the source. Hall reverb? Bussed. Plate reverb? Bussed. EQ? Angle/move the damn mic, edit your synth patch parameters, layer on a brighter or fatter sample. Get it right at the source.
  
  What this does is free your computer to compute (duh), and run the instruments, audio tracks, and bussing - comfortably. I don't even remember the last time I've had to freeze tracks - I think Logic still had a dongle (remember that white dongle?).

• My point being: modern machines are ultra powerful, and plenty cap/able - if spec'ed accordingly - to run most real life projects. Learn how to buss and architect your bussing commensurately. And leave 9 year old videos in the past - where they belong.

 

Last edited: Jun 22, 2025

That! Build one and see, PC.
That's a period.

 

Check this https://www.kvraudio.com/forum/viewtopic.php?t=621326

 

Way too many people tweak (or rather: anti-tweak) way too many knobs on Windows (or their hardware for that matter). Then they just re-install every 5 months, because inevitably, their installation ends up breaking in random places, shows unexpected behaviour or some update turns their system into a 2004 Pentium IV, because suddenly there are a million 'is this service running yet?' queries per second.

But that service we're querying for was disabled months ago, because some guy, who doesn't know what he's doing, believed some other guy, who also doesn't know what he's doing, heard from guy #3 that disabling the service makes your computer go 0.1% faster. So he put it in his 'perfect-windows-hacks.ps1' script, which we blindly believed would make our Windows system go vroom.

"It's not the tweaks, bro. The tweaks are good. There's like 3 guys on reddit saying they worked for them. Really, bro, it's Windows, you're just supposed to re-install it every now and then."

Don't believe the stuff Pictus posts either. He collects "tweaks" from random hardware forums on the internet and dumps them in his posts. His hands-on experience: almost always lacking. Useful as a reference for random stuff you could try - I guess.

Nothing against tweaking. But if you don't know what you're doing, don't trust random scripts or "Windows distros" to magically fix your problems - most of this stuff is misguided.

You gotta be realistic about the level of agency you have and the impact you make when tweaking in an age where we exponentially lump on billions more transistors every chip generation, but application software complexity hasn't changed much at all since the late 2000s. (Other than PowerToys)

That Windows tweak from 2012, that actually did make your stuff go 4% faster, how much will that do with a 2025 CPU of the same tier? 0.2% - 0.4%.

Scales.

 

I still remember Tune Utilities – I used them once and the problems started. That's why I haven't used tuning tools again for 25 years. Disabling services doesn't help either; it just creates problems.

The fewer programs you have, the more stable Windows will run. Once you finally have a fully functioning Windows, make a full backup of your C drive to a second hard drive so you can restore your entire system in an emergency.

I bought an update stop program last year so that Windows would stop generating errors and changing things on my PC.
After endless discussions, I also killed Defender. If I get infected, my rescue disc goes into the CD/DVD drive.

 

Asides from giving false positive on k's, Windows Defender doesn't affect my performance that much. I just turn off while scanning plugins because it slows and some times can mess things up. Asides from that, no fuss.

 

While I was fervently researching which components to build my last PC with, I got a little frustrated with how most hardware "shootouts" were focused on gaming/video-editing/3d-rendering performance while any discussion of audio latency or DAW usage was just an afterthought at best, I did actually stumble across one german-language forum post where someone had posted a massive comparison of all the then-current AMD and Intel cpus and chipsets specifically focused on this topic.

I wish I still had those bookmarks so this could be something more than mere anecdotal paraphrasing but his main conclusion was that you wanted a Ryzen with all the cores on a single die. Obviously the clock speed and ram speed matters but what he saw was that once your number of cores jumped above 8 there started happening the spikes in latency. The constrained resource, where the bottleneck is for audio, is the cache that is shared by all cores. Each core may have its tiny amount of local cache but there is the (L1?L2?L3?idk) cache which is shared by all cores and that is the important resource being consumed as each thread containing the results of a stream of output from each chain of vsts needs to be merged into the thread that mixes all the streams into one final audio stream which gets sent to the audio output. Each piece of Infinity Fabric (they call it something different now) could hold up 8 cores and he was getting the best performance on 8-core Ryzen 5s and 7s but for the 12-core, 16-core, 18-core, 24-core systems there were these occasional spikes in latency when a lot of activity was happening across the fabric interconnect, presumably synchronizing what was in core-local cache with whole-cpu cache. Keep in mind I'm trying to summarize this from memory of something that was google-translated and is probably somewhat out of date so make of it what you will.

 

a lot of sense already spoken on the thread. I used to build my own PC's when I was younger and short of money and tweaking the things that affected the DPC did help to some degree but I spent more time trying to improve performance than I did making music in the end. Fast forward many years and I decided to try SCAN custom built PC which was the best thing I ever did. Great machine, dead fast and just worked right out of the factory. IF you can afford it and want to spend your time recording and mixing, rather than messing around with shit that probably wont work well then just go custom built. Then I finally took the plunge to try mac, mainly because I needed one to test Genesis as we were in the early beta stages so I opted for a M3 Pro. It took a little while to get used to the "Where the F**k is this, or that"", but once I could navigate freely again then I discovered a whole new world of simplicity and well... everything just worked really well, straight out of the box, power for days and really good solid low latency in my Daws.