DAWs and E cores

[kvotchin]

So I watched a video recently that went through a bunch of well known DAWs, testing cores utilisation (P and E).

Based on that video at least, it seems Reaper is the best at this, followed fairly closely by Cubase, with all others quite some way behind those two. From memory, FL Studio could’ve been the absolute worst.

Is there any other evidence that backs this up, or contradicts it? Also, have any of the tested DAWs got better - or worse, I suppose - since then? And has anyone tested DAWs that were not included?

Possibly relevant for anyone using Apple silicon, or on the other hand, very recent Intel CPUs.

[Scoops]

Point of reference would be helpful.
Whats the video?
Whats the time frame/when was the video put out?

[kvotchin]

Apologies. Thought I’d linked it already. Here we go:



It’s from November ‘24. Note: Timestamps are in the video description @ YT.

edit: And oops, I seemed to recall Intel P and E cores testing, but no, clearly not the case. Well, if anyone can provide a link to those tests, assuming they’ve been done, that would be most appreciated.

[Kaine]

I did some quick testing on Intel around the same time the video came out and the results were fairly similar if you need a rough guide, although I didn't get around to publishing anything back then. It's something I've been meaning to pick up and do a proper look at for a few months to see how it's changed since, I should get around to that.

[PAK]

The answer is a bit more awkward than the video implies, though the bottom line is correct - If a host can use the Efficiency cores then that’s better than not being able to use them at all.

How much difference it makes depends on both the amount of E-cores on your CPU and what you run. EG A CPU heavy synth plugin might not “fit” on an E-core at all. So E-cores can help out, but are mainly of use on plugins with light requirements. This is why you always want as many P cores as you can get.

In terms of CPU heavy synths you also want to leave multi-core support off when you can, as this tends to carry an overhead of around 30% Vs leaving it on single-core.

In terms of hosts and performance differences, this is also not quite as simple as raw numbers imply. EG Some hosts will attempt to properly isolate plugins so they have a harder time crashing the host. That extra layer might add a performance penalty, but many users might want that stability over running a few more plugins.

Another example is Cubase attempts to keep some CPU in reserve. This is so you can interact with its interface, whilst under high CPU load, without causing pops and crackles. Another host might not care about that, and will allow its interface to become laggy to squeeze that extra few percent.

[kvotchin]

How much difference it makes depends on both the amount of E-cores on your CPU and what you run. EG A CPU heavy synth plugin might not “fit” on an E-core at all. So E-cores can help out, but are mainly of use on plugins with light requirements. This is why you always want as many P cores as you can get.

Definitely, re: P cores preference. Which is why I’m very likely going to upgrade to one of the current AMD high end CPUs, later this year perhaps. 16+ P cores should do the trick.

And thanks for the other info in your comment - I don’t know enough, honestly, to respond to it, beyond that. But I’ll look into whatever aspects I can, out of curiosity.

[uOpt]

The key here is that CPU scheduling has become very complicated. And the CPU schedulers in common OSes, especially Windows 11, are highly developed. The naive, underdeveloped scheduling schemes that some DAWs use aren't suitable anymore.

Reaper doesn't mess with CPU scheduling at all and let's the OS' CPU scheduling software do its job instead. That's why Reaper comes out on top on complex CPU architectures.