Why is Cubase so slow to load projects even on fast hardware?

Cubase
1

Until recently I was doing mockups on a Mac Studio M1 Ultra with 64 GB of RAM. I first started with Logic Pro X. Every project loaded pretty fast, but to be fair, I didn’t have as many virtual libraries then. Still, I used to work in it with the Eastwest Hollywood Orchestra Opus Edition, which can eat several gigs of RAM on some instruments.

But as some or most of you may know, 64 GB is not a lot when you work with virtual libraries, especially OT and the like, unless you work in sections, and I didn’t want to waste time with that. So I built my own PC with 192 GB of RAM and an Intel 14900 KF CPU, plus 4 Samsung NVMe drives (4 TB each) to make sure I had space for all my libraries and that they would load fast. These drives benchmark at 7 GBps when empty and when almost full, around 6 GBps. In the Mac I only had 4 TB in the internal drive, so most of the libraries had to go on external USB-C drives, and one Thunderbolt.

So when I built this beast of a PC with such an insane amount of RAM and SSDs that gave me between 6 and 7 GBps, I was sure loading projects was going to be much faster and as soon as it was done loading, I could press the spacebar and play them just fine, something that I could hardly do in the Mac Studio, especially on projects with lots of tracks.

But the thing is, it doesn’t load projects faster. Or maybe it does by a bit, I haven’t timed both on the same project to have an exact number. What I know is that I load a large project and it’s minutes of the “Loading MixConsole” until I see the actual mixer appear on my second monitor and the main interface on my main monitor. Depending on the project it can take 3-6 minutes, and if it’s a gigantic project, let’s say over 150 tracks, it takes an eternity. Those projects are not that common for me, but even in projects with around 50 tracks, loading takes several minutes.

In the Mac I used Cubase Pro 12, 13 and Nuendo 13, in the PC Cubase Pro 14, always up to date, so 14.0.20 as of now. Still slow. And it’s not like it takes forever to load but at least once it shows me the project, it’s smooth sailing from there. Even if at the beginning there are a small handful of tracks with instruments playing, it will start to hiccup, because I noticed more than once that once the project shows all the tracks, instruments are still loading in the background. So several minutes after the project shows all the tracks, only then I can press play and have it play smoothly. If the project is gigantic, it can be well over ten, even fifteen minutes after it loads for smooth playback.

I benchmarked my PC with a bunch of different programs, and stress tested all the components. Everything runs at the speed it’s supposed to run, and passes the stress tests without fail, even after running all night long.

So I don’t get it, is Cubase supposed to be this slow normally, or is there some setting I’m missing?

2

Maybe it’s not Cubase but Kontakt?

3

For comparison, I loaded an EW Opus “template” file I used for testing, and at 373 tracks it loaded up and was ready to play in ~24 seconds. Not all tracks are “enabled,” but most of the actual instrument tracks are, and are supported by midi tracks routed to the instrument tracks.

I actually purchased a “Cubase Template” from Composing Tutorials and though it was $50, it’s really been worth it to me.

You may want to examine the structure of your projects themselves. Even if your system sliced 20% load times off the top, your 15 minute project would still lake 12 minutes to load. So your returns probably are not going to be what you are expecting.

The reason I paid for the template was to see exactly how a professional composer approached the technical aspects of creating the tracks and routing everything “properly.” The template is only midi and libraries, but I’ve actually learned quite a bit regarding the organizational optimizations for managing large orchestral projects.

The reason I say that is because simply “buying a faster machine” doesn’t necessarily equate to “n faster loading times.” Little things could matter on your home-made rig, not the least of which is the correct chipset drivers and choices for “on-board RAID configurations” if you went that route.

Just as an aside, on my M3 MBP Pro with 128g LPDDR5 RAM and 8tb NVMe SSD, I did a test running a Win11 arm VM after loading the entire VM onto a 32g RAM DISK. The VM also had 32g RAM. One would think that running a VM from LPDDR5 RAM would have been an “instant boot,” but it wasn’t. It actually wasn’t appreciably faster at all, and the reason is because all the standard IO processes en masse could only run so fast. In hindsight, I should have just presented the RAM disk as a system drive directly to the Windows VM, but I’m not sure how much that would have mattered. The point is that gross raw power does not equate to gross raw “processing speed” gains when you’re talking about tiny gains at the micro-process level. It’s kind of like speeding in between red lights in your car. Sure, you’re going faster, but those little gains you make still encounter bottlenecks. Again, just an aside.

The variable here is how your project is structured, presuming you have every driver properly installed and that your rig is perfect. A 373 track Opus project loading and playing in 24 seconds at least points to the fact that you may want to look at how your project is structured, and how/where you can bounce to audio instead of “live” MIDI-based sample libraries on all tracks (if that’s what you’re doing). I think that would not only help you now, but it could help you in the future, particularly if you are going to be sharing that project with others who don’t have bleedin rig such as yours.

4

Two things to consider here beyond what you have already mentioned.

  1. There are unfortunately some plugins out there that are licensed with an online system that from time to time needs to refresh the authorization and depending on the instant speed of your internet connection (ie how many neighbors are watching 4K movies on Netflix at that instant), how many other processes on your machine are using the network adapter, how many cows are within a 20 mile radius of the authentication server, etc… your plugins may take a minute to fire up when you load your project.

I also have some other plugins that check for updates when instantiated (not when Cubase first scans the plugins at startup) so that is another potential delay waiting for the network.

  1. your NVME drives may have fast transfer rates but doesn’t mean their random seek rate is any better than any other device. Usually they are, but it isn’t a given. And if the NVME drive is connected via SATA vs PCIE connection your seek times are going to be way lower.

If each of those tracks in your project is loading up some different sound in the library that is a lot of drive seeking at project load. Its probably not the data transfer slowing things down, your drives could probably transfer the whole library in a matter of seconds.

You will probably find your project starts loading faster if you start bouncing/freezing some tracks because that will then become a simple disk read task that your drives should handle with ease.

5

Well, I hardly use Kontakt, only when it’s absolutely necessary. I have lots of libraries that are Kontakt based, but I just hate Kontakt because it’s an old GUI (at least the part where you interact with the instrument itself, even if they made changes to the cataloguing GUI), that usually is blurry or too small, and it’s slow to load.

6

That sounds interesting. I think you’re referring to this one, correct?

If so, you’re saying that from Cubase at the welcome screen, without this project having been loaded since the most recent bootup, it takes about 24 seconds?

The thing is that running anything from a VM, especially when we’re talking about Windows 11 ARM, which is still rather new software, doesn’t really matter that much because, in reality, why would you run Cubase inside a VM in a Mac? I mean, Windows 11 ARM is probably not meant for Macs, despite being ARM machines, it’s a very new OS and Cubase for Windows ARM is also very new.

I assume that the template you say opens in 24 seconds is in that MBP M3 with 128 GB of RAM, correct?

7

That’s the one!

Indeed - I tested the time when I saw your post, and I haven’t loaded that guy for a couple of weeks now. I’ve also got my EW libraries on an external TB3, but granted that’s a SanDisk PRO-G40, and incredibly fast.

That’s what I meant by an “aside.” I’m NOT running Cubase in the VM. I was just letting you know how reality is different from expectations, using the example of running an entire Windows 11 VM host disk from a RAM disk, and the VM itself had 32 gig or RAM. And even then, it wasn’t any faster using the windows session (booting, rebooting, using) than just running the VM from an SSD. It was a completely different example of how even running a VM from a RAM disk didn’t speed things up as one may have expected.

Yes sir, that is correct :slight_smile:

8

I think the number of cows within the 20 mile radius is the most logical explanation :rofl:

But yeah, good point. However, the VSTis that I have that need online authorization via iLok are the VSL ones, which I hardly use, because I only have the Epic Orchestra, Synchron Prime Edition and a couple of choirs. The rest that are iLok based are tied to the machine, not online.

Good point, but they’re all PCIe Gen 4. The motherboard has one slot for Gen 5, but I didn’t use it because the drives are Gen 4, and the Gen 5 slot shares the PCIe lane with the graphics card or something like that. So I used all four PCIe slots, four Samsung NVMes, one of which is the 990 Pro, and the other three are the 990 EVO Plus. The Pro is obviously used as the C: drive.

Now, the slowness could be attributed in part to the OT libraries, whether it’s the SINE engine that is slow or the libraries are too large. I know in the case of the Berlin series libraries, those are gigantic and when it comes to the strings, they go past 5 GB of RAM very easily.

But I also wonder if there’s something in the Cubase code that is just not optimized well enough for fast machines. Meaning, that it can only go so far because it doesn’t know how to take advantage of the newest hardware. Please don’t take this as me bashing Cubase, it is my favorite software ever, and I will keep using it, even if the project loading is slow, because that’s just one disadvantage while it’s full of amazing features.

But there’s something weird with Cubase since I started using it on version 12. Like I mentioned, loading a large project can take several minutes, and even when it loads the project GUI and the mixer, it can’t play the song without hiccups unless you wait several more minutes until it finishes loading all the instruments in the background.

However, if as soon as you load the project you do an export mixdown, it exports that file in less than half the real time of the project, and the rendered is perfect. No glitches, hiccups, or anything bad at all. So that puzzles me, because it would seem like when doing a mixdown, Cubase accesses the full capabilities of the hardware, while when working in normal mode it slows down considerably.

Now, when I posted about this before, some people suggested that when doing a mixdown, Cubase freezes the tracks. Not really. Because I just selected the 47 tracks in this project, which is almost 14 minutes long, and it took almost five minutes to freeze just six tracks until I cancelled it. But clearly it was going to take about 40 minutes to freeze all tracks. And when you do a mixdown, the rendering starts in a couple of seconds, and goes really fast. The mixdown, which I did right before I tried to freeze all the tracks, took about the same time as freezing one track, two at best.

So if Cubase can load a project, however long that part might take, but once it shows the tracks, it can do a mixdown of a 13 minute project in 2-3 minutes without hiccups, why is it that it can’t play the song without hiccups when you press the spacebar? That’s the part that I can’t find the logic for.

9

Since C12, SB built native support for Intel hybrids, and they also take full advantage of both performance and efficiency cores on Apple Silicon, so that’s simply not accurate.

1 Like
10

OK, then what’s the logical explanation for the scenario I just mentioned?

11

Is Windows malware scanning disabled for the relevant locations and/or file-types?

12

OK, so I bought the template. Thanks for the tip, it’s a really nice template and while I don’t own SD2 or Spaces II, I own the rest, so it’s a great template to learn. There’s a track that has a “Hang Drum” from SD2, and it’s not on SD3, but I looked up what it was and added a metal drum from SD3 temporarily until I find something more accurate, which I probably have in one of the gazillion other libraries I have collected in the last two years (Yes, I had an addiction, but it stopped).

As for Spaces II, I replaced all the instances with Cinematic Rooms Pro. This was for the project called “EW OPUS PURGED DEMO VST3.cpr”. I’m not sure if you were talking about that one or the empty one. But this one, from the Cubase Pro 14 welcome screen, loads in 33.35 seconds. That’s not bad, but it doesn’t seem to me like it’s such a large project, considering that most of the 373 tracks are disabled, and the length is less than a minute and a half.

My most gigantic mockup was I think like close to 200 tracks and used about 178 GB of RAM (which I only realized when I loaded it in the new PC, since I worked on it in the Mac Studio months before I built the PC), and it takes ages to load to a point where it can be worked on, but my regular projects are more like 100 tracks or less. I don’t know, it seems to me that based on the hardware I have, they should load way faster. Maybe it’s something in the way I approach them, but I don’t want to start freezing tracks all over the place, for that I would’ve kept working on my Mac Studio. But freezing and then unfreezing tracks adds up to a lot of time in the end.

13

That it’s your project that isn’t optimized, not the code.

14

Which is what I said, and they’re MIDI tracks, so length is somewhat irrelevant in that case. The INSTRUMENT tracks are enabled, as stated. It’s great you got the same project, since that’s a 1:1 comparison - it takes nearly 50% longer on your Win11 system than my M3, so that’s something to look at.

It seem like you’ve already answered your own question here. You’ve got massive projects, and you can optimize them if you would like, or not and just wait for them to load. Nothing here speaks to “the code isn’t written to take advantage of new hardware.” Again, even if all other things being equal you build some system that was a full 30% faster in every way, you’re only saving a few minutes if you leave things as they are. The most valuable thing for you to do, in my opinion of course, is to optimize your projects. Buying new hardware every time your project gets too big when you just leave all tracks enabled with “live” midi rather than freezing/bouncing doesn’t seem very effective no matter what software you’re using.

15

P.S. Do you have VE Pro? I purchased it for this explicit purpose, and while not something I’ve used enough to get my money’s worth, using your old Studio with 64gig may be a really good use case for that. That way the libraries are already loaded on the Studio “server.” I think they have a trial, and that may be something else to look at from an optimization perspective.

16

Its the only thing I can think of. I have some stuff from this one company and the authorizations were on this little blue thing that you plugged into a USB port and they just fire up instantly. Things would open before you could even take your finger off the mouse. I don’t understand why everybody doesn’t do things that way.

I have experienced this as well with projects far simpler than yours, and I started seeing it more often after C10. I am going to go out on a limb here and completely speculate … but I suspect a lot of the random “delays” random people are having with random projects is being caused by VST plugin GUIs.

I hate to say it but a lot of VST stuff is basically “done” being 30+ year old technology. Now its about repackaging that same 1997 algorithm and making it look pretty to get people to re-buy stuff. And I’ve probably purchased too many of these myself, so I am contributing to the problem.

But with the system you have (better than mine, my 4th NVMe is using a SATA bus) you would be able to load hundreds of .wav files in cubase and play them back simultaneously with zero issues. I think my channel max count for a project is near 100 … but cubase has no problem with 100 96 64 wav files playing back at same time… it doesn’t even break a sweat.

And when rendering, cubase for the most part disables VST GUI interfaces although they will periodically update their display to make it look like they are alive … they aren’t responding to user input while rendering.

So just guessing, but I suspect that is where the lag is coming from and with hundreds of tracks that your computer should handle no problem … its only got one keyboard/mouse/etc in that each of them are taking turns monitoring… or its the Cows.

17

But I’m talking about the same exact project, not an optimized project to mixdown and another one to edit. Let’s say that my projects are poorly developed, that because I’m fairly new to this, I make every mistake there is to be made. And that could totally be the case, since I learned from hands on, some advice here and there, YouTube videos and so on, so it’s not like I’m a professional composer and sound engineer that has been working with Cubase for years.

So let’s say my project is as bad as it can get when it comes to optimization. Then why does Cubase renders a mixdown in less than half the duration of the timeline, even when the project is just loaded, but can’t play it back right away?

It’s what I said before, it’s like Cubase, when doing a mixdown, goes into Turbo Formula 1 mode, and when it’s not in mixdown mode, it’s much slower. And the interesting part is, when it’s rendering the mixdown, it doesn’t even torture the CPU. I’m not saying I can run another demanding app at the same time like DaVinci Resolve editing 4K video, but I can do normal stuff in Windows while it’s rendering that mixdown and barely feel it.

That’s the part I really want to know, what does Cubase does when it goes into mixdown mode? I know it doesn’t freeze the tracks, that takes ages, so what else is there?

18

Keep in mind that macOS is also a far better OS than Windows 11, it knows how to use the hardware to the best it can, and it’s not surprising when both hardware and OS are coming from the same company. Honestly, when I got the Mac Studio, I still hadn’t discovered the world of VSTis, and 64 GB of RAM back in 2022 seemed to me like more than I could ever use.

In January of this year when I decided to build this PC, a Mac Studio with 192 of RAM (the max back then before the recent M3 Ultra), would’ve cost me $8,000. I think this PC cost me half as much, but if I had the $8,000 for the Mac Studio M3 Ultra, I would definitely go for it, especially after 3 months of using Windows 11. :face_with_spiral_eyes:

LOL that’s definitely not the plan. If only there was a way to replace my Mac Studio M1 Ultra 64 GB with 192 or even 128 GB of RAM, I would be happy with that machine for many years to come.

But 192 GB of RAM should be all I need, I doubt I will ever need more than that for any mockup, and the one I mentioned was more like “let’s see how much can I put in there before the machine explodes” :rofl:

Which if you care to listen what a project that takes 178 GB of RAM is like, this is it: Stream Bear McCreary - Prelude To War MIDI Mockup by Sebsaghini | Listen online for free on SoundCloud

I do actually, but only because it came with the Epic Orchestra 2.0, which seemed a good purchase at the time and hardly ever use anymore. But recently I was very glad I bought it, because I have Moog Model D, Moog Model 15 and Animoog Z, all three of which are Apple OSes only and don’t have a Windows version, so I can load them in the VePro Server on the Mac Studio and access them on my PC.

Now, doing the same but the other way around, having most of the instruments loaded on the PC and working on the Mac, I tried that and I wasn’t too crazy about it. The VePro GUI doesn’t zoom and it’s a bit too small for me.

Still, it’s a great thing to have, if I ever run out of RAM in my PC (fat chance), I know I have a little extra room on the Mac.

19

Because when you are rendering it, the project is already loaded; and unless you’re doing real-time rendering, then it will render the tracks “at speed.” Processing existing, loaded tracks is different than having to load those tracks in the first place. Two very different processes.

20

Hahah. For a second there, I thought you were saying YOU were Bear McCreary - I was like, “oh, cool.” :smiley:

You know, I’m right there with you. I got it because I’ve got an iMac Pro 8-core with 64g and 2tb drive just SITTING THERE and figured I’d put it to work - but it’s a bit clunky to work with sometimes. But I’ve seen the way some folks approach their workflow with it, and if I get to that point, I’ll probably do the same.