With a blank canvas, a resourceful, inquisitive mind, a healthy desire to be instantly creative, and chequebook at the ready - I can’t fathom why anyone would choose a Win 11 Intel-based computer, just for all-round audio production, using Steinberg’s Cubase products.
It is less than Reaper’s performance to be sure, but make sure you read the text and don’t just look at the graph because the graph itself is unintentionally misleading. More info here:
Looking at the bar graphs alone makes it seem like Reaper is 10x faster than Cubase. There is a sizeable difference between them but it isn’t nearly as big as the bar graph alone makes it seem.
1 Like
So regards the DAW’s, there is a sliding scale from Reaper to Studio One and finally to Cubase, where ASIOGuard was heavily loaded and prematurely overrun by the processing flow of the session. Studio One is better, but once placed in context against Reaper, there is still a lot of room for improvement. In both these instances, total CPU resources used in TM would have been at best 20% on the 9950X/285K, lower again on the TR.
For some additional back ground, I have been navigating numerous Cubendo clients contacting me specifically about ASIOGuard spiking and/or being prematurely overrun with 80+% of resources still available for several years now, and I really don’t have any consensus on a solution for them past changing the session logistics if possible, or the inevitable freezing/rending tracks, which feel like a cop out to be honest. But there really is only so much that can be done while the devs are in a holding pattern at best, or in a mode of denial and dismissal at worst.
Yes the text does say more..
Yes Cubase is certainly worse performing than Reaper but I’m trying to say that looking at the bar graphs alone is misleading because they are counting the number of plugins on top of a fully loaded session. So you’re not starting counting from 0, you’re starting counting from a fully loaded session.
In other words, if your data started off looking like the top, and you put just the last section in there to “zoom in” on the difference:
And you just show the graph on the bottom, the lower line looks a lot longer than the top one, but that’s because the beginning was lopped off. And that’s the way the bar graph was shown on purpose.
Now, the difference in the data is almost certainly a fair bit larger than what I show on the top, this is just an illustration of the issues with the graph format with the majority of the count removed and only the rightmost part shown. They are graphing the count of extra plugins that they can load on a session that already is fully loaded with tons of plugins.
There is still a big difference that shouldn’t be there and Cubase does need to improve and it is really important. I’m not disagreeing with any of that. Just pointing out one issue with the representation on the bar graph, where if you just show the bar graph to people it is going to be misunderstood.
When it comes to actually recording, producing, mixing, and mastering music–Cubase is better than Reaper. It’s a working DAW, not a DIY project.
MacOS is better than Windows if you’re interested in stability, performance, reliability, and low-latency pro audio. When I sit down to make some music, I don’t want to be thinking about registry-hacks, BIOS tweaks, or IT tasks–I just want it to work.
2 Likes
Eight years ago (Cubase 9.x era) I wasted a fair amount of time attempting to compare Ableton Live’s multithreading capabilities against Cubase’s on Windows. At the time, Cubase’s multithreading performance was quite good and predictable: You could progressively enable tracks in a DAWBench-style project and see additional cores get activated once the in-use cores reached max capacity. At the point when the last available core was maxxed out in the Windows performance meter, you would start hearing audio dropouts. It was easily observable and repeatable. (Ableton Live 9, on the other hand, would start to fail much sooner- and worse than causing dropouts, it would just choose to not activate plugins on unmuted tracks; they would look enabled visually, but generated no sound and made no impact on the CPU load.)
Something has definitely changed for the worse in Cubase on Windows. When I first started hearing these complaints with recent versions, I assumed it was because of the new mixed-core architectures some PC platforms have now, but this thread disproved that theory. Cubase has actually gotten worse at thread distribution.
I wonder what changed?
3 Likes
It depends on what you need and want. For a moderately powered gig machine, or for production ‘on the go’ there are several models of Macs that pack a nice punch for the money. If 99% of what you do is ‘all software in the box’, and you’re also in the market for a new interface that’ll work with a Mac (or don’t even need one and the built in audio is good enough), even the lowest end modern Macs (Base Mini, Air, or even midrange iPads) are great little ‘creative’ workstations.
In the studio, film-set, stadium-arena, theatre, theme-park venue, or with larger on-stage setups there can be scores of reasons why a PC is the go-to choice.
Racks full of perfectly good equipment exist that simply do not work with the new processors (need the Intel/AMD instruction sets, which can be emulated on other processors, but Apple is highly likely to yank Rosetta support well before it reaches its max potential, and make it difficult if not impossible to get it on future OS releases), and probably never will. For what it’s worth, you can take a brand new PC with the latest Windows Release, and still run pretty much any MSDOS or Windows app EVER written (and then some). Options even exist to get ancient 16bit devices/drivers working. Legacy support on a modern Mac is simply ‘non-existent’, period.
Perspective…replacing some of the higher end interfaces, storage and backup systems, mixing consoles, and video switchers out there would be quite hefty investments for relatively minor improvements, or in some cases it can even be a ‘downgrade’ to kick out the old and get something new. Quite a few tasks still exist where going ‘legacy’ is the only option, as the ‘new and improved’ systems have stripped it all out with no intention of ever putting it back, and whatever ‘new thing’ is supposed to take over that job simply doesn’t work yet, and might never work well.
Even if one can afford to replace everything in the rack to take advantage of Apple M and ARM architecture, unless you spend a bloody fortune for the highest end Macs, you only get a few places to plug anything in!
A mid-range PC will get you all kinds of PCIe slots (which have no problem running older ISA/VME and more with simple and inexpensive riser/daughter/converter boards), scads of USB/Thunderbolt ports, plenty of space to pop in high end internal SSD, and also RAM. Add endless choices for graphics cards, outboard DSP systems, and more. None of these things can be done on most Mac models. When getting into Mac models that do have plenty of ports and can do this kind of stuff, the options for kit that’ll work with the Mac are still very limited and outrageously expensive.
In many ways it pays to have both in the studio. There are always various niche bits of software for PC that don’t exist on Mac, and vice verse.
Again, if you do everything in the box through software, and don’t need much in terms of ability to hook up external kit, a few Mac models out there really do pack a great punch, and are worth the price. For everything else….there is the PC.
3 Likes
It isn’t that Cubase has changed for the worse in this way. They built a lot of programming infrastructure to work around deficiencies in Windows when it comes to audio handling and thread scheduling for important audio tasks. The problem is that Windows, and particularly Windows 11, has gotten a lot “smarter” since then and handles audio much better as well as thread scheduling, so these workarounds are no longer needed - and worse, are likely making the performance much worse now than if they hadn’t been put in place from the start. So the very changes that made Cubase the best performing DAW on Windows before could be the same things that are negatively impacting it now.
The problem is that they can’t just rip out what is probably 20 years worth of core handling logic in one shot, since it probably isn’t all in one place and may be spread throughout the code in various places. Yanking all that out in a hurried careless YOLO fashion could create new problems like instabilities or even tank the performance. They need to change things gradually so that they don’t risk things suddenly getting worse for people.
1 Like
That’s a plausible theory. I know that Microsoft devoted resources to improving both general performance and audio handling in both Windows 10 and 11. (Even ASIO, still the standard on Windows DAWs, was originally a workaround for Windows’ limited audio implementation from waaaayyyy back.) I also know that Intel and AMD are always implementing and refining technologies in their platforms. For example, all the “DAW optimization” guides used to recommend disabling hyperthreading (SMT) and C-states (power management) in BIOS because when those technologies were newer, they were not optimized and there wasn’t wide software/OS support for them- and they did often result in performance problems for certain workloads. But now both of those features are much more robust and performant, and you are more likely to hurt (or at least not improve) performance by disabling them.
But like you said, if Steinberg spent years building a reliable system DESPITE Windows’ earlier drawbacks, they now might have to do some unspooling to let the DAW make maximum use of the features modern operating systems currently offer.
4 Likes
For producing music, one doesn’t need endless choices of GPUs. As for outboard DSPs–UAD is good and works with Mac over Thunderbolt with very low latency–and it’s stable.
That’s not really accurate. You can connect anything you want to an external audio interface, and you can connect any MIDI devices you need to your Mac with a Thunderbolt docking station or a USB hub. If you need to run a bunch of unsupported software, or directly connect legacy devices to your machine–a PC is nice. But it’s not true that Macs are only for in the box music production–it doesn’t even make sense.
Prices for a used Mac Mini Studio M3 Ultra are getting very keen indeed.
I’ve looked into switching to Mac. I’ll need around $2,000 of new audio/MIDI interfaces to make the switch and match the I/O I currently have.
Some of my buddies that do big live audio setups still feel chained to the PC world as well. They have $20,000 mixing consoles that flat out don’t work with a modern Mac. Yes, it’s sometimes older stuff, but as long as they can keep it working, it stays in the road earning coin (they have newer ones too, and those stay in the road…if yer going to invest half a mil in racks and stacks, you crew it and keep it in service for as long as possible). Those guys need machines that can run 4 to 8 ethernet ports, a dozen or so HDMI video ports, specialized cards for lighting rigs, and a lot of other weird things that Macs simply aren’t up to doing (or cost 10 times more with little to no advantage if it can be done). Yes, some form of Cubendo is often a critical part of the system.
With a new Mac…20+ years of archives won’t work anymore. I can’t roll back to any setups that used 32bit VST2, DX, etc.
The PC in the rack has over 20 storage docks (An assortment of SATA and SAS drives. Swapping a drive is as easy as pushing a button, and all but the current system drive is hot-swappable). I can keep using this same chassis with the nice storage docking for many ‘future builds’ if I so desire. I never have to run ‘backup software’, since the array has built in hardware raid (optionally plug in two drives instead of one, and get two copies of everything at all times).
All I need to do this with the PC to get at my archives is release any cloud based software keys, turn off the machine, pop out some hard drives, pop in a different set (old system drive with the OS and all included), ‘maybe’ do a bios tweak or two, turn it back on, re-register my cloud based software keys, and I’m right where I left off with that session, exactly as it was the last time it was touched. No way I can do that with a modern Mac (Maybe the highest end stuff, maybe? The low and mid tier stuff has the system drive SOLDERED IN).
Since day 1, Apple has made it a POINT to destroy any kind of legacy and/or cross-platform support. I got burned pretty bad with the PPC (great machines, and just when the software was finally making use of RISC architecture, they abandoned it. Good move for the company swapping to Intel, but it sucked for any of us who threw money at PPC setups).
It’s not just hardware and OSes with Apple. I made the mistake of investing in Quicktime some years ago, and it cost me big time (it was sold with promises for full cross platform compatibility, access to iTunes and such by most of the other OSes, and more). Never again!
The SAS raid array that automatically handles backups for the session data, the hot-swappable bays for media that’s all raided and ready to go, the ability to swap out the system drive at the punch of a button….as far as I can tell such equipment will be useless on the Mac Platform with anything but the highest end models, and still might not work. In contrast, it still works with just about any ATX or larger PC build that one can buy off the shelf, brand new, today. It doesn’t even have to be a high end build, a moderately priced rig can handle every bit of it, no sweat.
Don’t get me wrong, I’d love to own a Mac workstation, especially for on the road, and at some point I intend to get one, but there are many reasons why a PC, or even two or three of them chained together (VSTLink, VEP/AudioGridder/Etc) still makes sense, and still doesn’t cost as much to run/maintain as a Mac based option. Particularly if you already have systems and well oiled implementation plans for keeping it all going.
Swapping the main platform is not trivial. If starting from the ground up is in order, or working it in through small stages is practical for somebody, that’s great, but for everybody else, there is still the PC.
That’s kinda the point of this thread. The PC hardware isn’t even breaking a sweat before the ‘software’ chokes. The PC architecture is ‘part’ of the reason, but tests are showing that Cubase can and should perform a lot better on Windows and modern PC builds than is is right now. It might take some time to sort it out, but in theory, performance should be much better than it is right now.
For me personally, Cubase runs fine on a modest and fairly old 12 core platform (3900XT). I don’t need piles of demanding plugins all running at the same time (a project might have 100+ plugin ‘templates’ out of the gate, but it is RARE that more than dozen or so are actually ‘in use and making sound’ at any given moment). 98% of the time, 4 solid cores locked to run wide open all the time is plenty for my needs. For the 3% that it’s not quite getting it done, I can freeze or disable some things, or move some of the instrument hosting to a server (Audio Gridder, VEP, Bidule) that uses the other cores, or even to another machine if I happen to have an old one laying around unused.
At the same time, I understand why people with beefy modern PCs are making noise. 12-30% of their processing resources used, and the DAW chokes? That could be infuriating to say the least. They just want and hope that this aspect of Cubendo starts showing ‘improvements’ when it comes to running on the latest and best PC processors and motherboards, and I don’t blame them. Yes, in the PC world the 3900XT/AM4, being near the top for a consumer level build back when I got it was indeed ‘modest’ considering the Thread Rippers and Xeons (with motherboards that can run several of those in tandem at that), and yeah, the higher end Macs are still babies compared to the true enterprise class stuff in the PC world (Xeon and Thread Ripper platforms), yet when looking at the highest end Macs, they can cost as much as doing an enterprise class PC build decked out with Xeons or Thread Rippers!
You might not need GPU options, but others do.
Spectral Layers and similar competing products make extensive use of a GPU (or even multiple GPUs) if it’s there, and has enough onboard memory to be worth-while.
How many video screens can you run with a Mac Studio? I’ve seen PC rigs loaded with cheap video cards that cost half as much as a bare Mac Studio that can run a dozen different HDMI screens at once, and yes, with different streams going to each of them. Yes, Cubendo or VST Live is sometimes the master clock and streaming engine for it all.
Sure, if you got a UAD rig in the past few years since the M1 release, or you go buy a new one today, you may well have a box that’ll run their plugins outboard. That doesn’t account for RACKS full of stuff that people already have in their studios that didn’t get Apple Silicon drivers, and probably never will. Sure, some of it will work through things like AES/EBU, ADAT, MADI, SPDIF, etc….but if your existing interfaces that support those standards won’t work on a Mac due to a ‘lack of drivers’, you still have something pretty expensive to swap out to make the platform switch. Or….I guess you can sync up Macs to the PC and ‘be creative’ on the Mac while the PC runs the studio (hence why I said it’s nice if you can have BOTH).
Again, I’m not saying Macs are bad. There is certainly a niche that they currently fill very well, and price/performance ratio works out nicely. There are still quite a LOT of things where investment into a PC running Windows just makes more sense.
Most of us would ‘like’ to have and use both, but at the end of the day we have to decide if the session singer, piano technician and the guitar luthier gets paid, or we get new toys.
3 Likes
That’s your opinion. Yet the proof is in the pudding (the data). Too many of you apologists for Steinberg work against us because they never put Steinberg’s feet to the fire to get the software to work any better.
Some of us do more than production work, in an environment where a Windows based system is more ideal. I don’t want multiple systems for multiple things. I absolutely hate the Apple UI and infrastructure. I tried it over the summer and it’s just an awful experience in my personal opinion. I understand though, you’re in love with Mac. That’s fine. Everyone should be able to pick and choose. Yet the Software should work more similarly than they do between Apple and Windows systems. That’s the point here. Because, right now, they don’t. Same hardware. Someone on a GS thread claimed they get 211 instances on their Mac in Windows yet get over 300 in MacOS. It’s clearly a threading issue in Cubase while running on Windows at play.
I’ll also say you’re quite misinformed if you think we’re all doing registry hacks, BiOS tweaks or the like. My Cubase 14 install is rock solid in Windows as is Windows itself. No registry hacks, no Bios tweaks, no nerding out with settings. Just install and run. It’s just the threading performance of Cubase is substandard, as the documented evidence shows.
There’s no honest way to look at those charts on the same hardware and not come to the same conclusion about that. I don’t really care if its 2 or 3 or 10x difference. Just there is a difference. I have so much headroom available on my CPU yet Cubase is choking. I can install Audiogridder and get more performance (as explained well above). So I know more is available. Cubase just doesn’t make use of it on it’s own. So real world, the threading in Cubase is not where it should be. And using MacOS isn’t the solution. Full stop.
4 Likes
I’m not so sure it’s completely unintentional. This graph has been doing the rounds all over the web without the appropriate context, but it’s quite difficult to find the baseline that this is associated with (or at least to find it clearly explained).
I agree that Steinberg has some work to do and there has been acknowledgement of that from company representatives, but anyone looking at that graph alone (as some people here are) will come away with a distorted view of what’s going on.
Didn’t I just cover this and clarify in further detail in the other thread, why are you attempting to misrepresent this ?!
That graph you created has zero correlation to what the data represents , you are attempting to show there being only a small variance when taken into context when the preload resources are taken into account.
So to clarify this again, the MIX session had a large number of various/mixed plugins across tracks/groups/busses, its not something that can be empirically placed on a graph. Its a session that is designed to test the thread management under a typical mixing environment. The utilization of the remaining overhead over and above, is what is detailed in the graphs, and its important, relevant and not easily dismissed.
80+ additional resource heavy plugins is significant , which ever way you want to cut it !
Like I noted in the other thread , the main focus was to show lack of scaling from 32 to 64 threads across all the tested DAW’s, the comparative performance between the DAW’s was not the main focus of the article.
Also, as I noted in the other thread, we have results of a more standardized/empirical core balance/saturation test across multiple DAW’s where Cubendo is 60-70% down on some of the other DAW’s, even against some DAW’s without an extended/hybrid playback buffer. That report will be posted when all the data across the wider system architecture pool is completed, but I already know its not going to be well received by Cubendo users.
3 Likes
I’m not trying to misrepresent this. My graph is not to try to show that you exaggerated the difference - I’m just showing the problem with only graphing over and above the “mix session” with what that has in it already. I’m not trying to show a small variance - I said the difference is probably a fair bit bigger than my crude line chart indicated - it was more to convey a point. You have said yourself, your bar graph isn’t a count from 0 load (the system doing nothing) to full load, which is what people assume it represents when looking at it without reading the text, making people think that Reaper can take 10x the plugin load of Cubase on the same system. Is Reaper actually 10 times faster than Cubase?
Really , do tell ?!
How is the baseline difficult to find, there are links to the article posted by the person who posted the graph.
I presented a fully detailed report where everything is clearly explained and placed in context, and you are accusing me of purposely presenting the data in a misleading manner , because you haven’t bothered to read the report ?!
2 Likes
And you are missing , or more precisely, ignoring the point that Cubase was unable to utilize the remaining resources that were available over and above the baseline mix session, which is what the charts represent.
Yeh, the variance is 80+ plugins !
As I noted the baseline preload has a variety of plugins, it can’t be represented as a number, but that preload was a lot closer to exhausting the Cubendo audio engine than Reaper , which wasn’t the main focus of the report ( I am sounding like a broken record), but is a clear indicator of the variance in being able to utilize the remaining available resources
I can’t control what people who skim the report assume the charts represent, all I can do is present the data accurately in my reports.
1 Like
I’m not accusing you of anything, whoever you are. I’m referring to individuals who post and repost this particular graph on various forums without providing any context.
Please point me to the baseline data that this graph is an extension of so we can get the complete picture.
Just to add: I’m not disputing that Reaper performs better than Cubase. I’ve seen other examples of it that has nothing to do wih you. What is important is that the published information is complete so that people have a good understanding of what they’re seeing.
2 Likes
I’m not suggesting otherwise at all - but it is human nature unfortunately to try to just interpret such a graph by itself, or as @KT66 said, people are sometimes just sharing the graph by itself. People love visuals and if they see a visual that seems to capture a big block of text, they are likely to just skip the text and look at the visual, and then just sharing the graph by itself with others compounds the issue. None of that being your fault in any way.
For my new system I built a couple months ago to replace my 6 year old Ryzen 3900X that was starting to fall over in Cubase, I was originally going to get a 9950X CPU. But I carefully read your reports and researched from mutiple sources that confirmed that the 285k was a much better processor for DAWs in general and Cubase specifically. There were people who tried to say that the 9950X was only slightly lower in your graphs and so that it was not that much slower, but your full explanation provided additional context in that case that made it clear that the 285k was a much better choice than the 9950X (this is a case again of people seeing your graph and reading something into it that didn’t quite jive with the explanation given). I’m personally quite grateful for the work you did on that, because it helped me to make a better purchase, and the 285k has performed well for me. The projects that fell over before are running just fine now and I have some headroom too. I’m not sure if that would have been true if I had just listened to the YouTube reviews telling people to buy AMD Ryzen which were mostly from gamer channels.