Audio Performance of different Cubase Versions compared

Audio Performance of different Cubase Versions compared

With every new version of cubase I think it gets less performant.
Especially the feel of Cubase 8 is quite lazy. It does not respond very fast
compared to the previous versions.

Steinberg promises with Cubase 8 Pro:
->Performance boost for more instruments, more tracks and shorter loading/saving times
…which was a reason to update (and the new plugin manager that I missed since many years).

Now it is time to do an objective test and compare different versions to prove
my subjective impressions.


Based on DAWbench VI Universal - 2012-R2
http://www.dawbench.com/benchmarks-old.htm
http://www.dawbench.com/downloads/dawbench-vi-2012-r2.zip
file: DAWbench-VI-C6-CV.cpr

I’ve loaded this file in Cubase 6.55-64 first, updated all
Kontakt library files (samples have to be searched) and IR files too
and saved it as DAWbench-VI-C6.55-CV.cpr.
I saved the file in each Cubase version with a new, different file name.

Usually I use 256 samples and for bigger mixing projects I turn up to 1024 samples.

You need Kontakt 4 and the Kontakt 4 library to run this DAWbench.
Run project for 5 to 6 loops at least and observe VST-power (F12).

Cubase settings: ASIO guard off,
Mutliprocessing: on
Audio priority: normal
Steinberg optimized Audio performance: on (german: Modus für optimierte Audioleistung…)

Regards, Mike

PS: results follow in next post

Results, summary:

My system:
OS: Windows 7 Pro SP1, 64bit
CPU: Intel Core2Quad Q6600 @ 2.4GHz
RAM: 8 GB
Audio: RME HDSP9632 Driver 4.05
Graphics: ATI Radeon HD2600Pro -> 2x 1920x1080 displays, AERO on
HD: RAID 0 for recording and mixing

Audio buffer size 256:

Cubase 6.55-64: Poly 20 ASIO around 50%-55%;
Cubase 6.55-64: Poly 60 ASIO around 55%-60%; with Poly 80 peaks to over when loop starts again

Cubase 7.5.30-64: Poly 20: average load 55-60; real-time peaks to 60-75%
Cubase 7.5.30-64: Poly 60: average load 65-70; real-time peaks to 80-95%

Cubase 8.05-64: Poly 20: average load 65-70; real-time peaks to 85-90%, sometime peaks to overs

Audio buffer size 512:

Cubase 6.55-64: Poly 100: ASIO around 50% peaks to 55% when loop starts again
Cubase 6.55-64: Poly 120: ASIO around 55% peaks to 60% when loop starts again
Cubase 6.55-64: Poly 140: ASIO around 70-75% peaks to over when loop starts again

Cubase 7.5.30-64: Poly 100: average load 55; real-time peaking around to 65%, sometime peaks to 90%
Cubase 7.5.30-64: Poly 140: average load 65; real-time peaking around to 75%, sometime peaks to over

Cubase 8.05-64: Poly 100: average load 60; real-time peaks to 90%, sometime peaks to over

Audio buffer size 1024:

Cubase 6.55-64: Poly 140 ASIO around 50-55%
Cubase 6.55-64: Poly 180 ASIO around 60-65%
Cubase 6.55-64: Poly 200 ASIO around 65-70%%, no overload! Poly 220 with overload when loop moves back to begin

Cubase 7.5.30-64: Poly 140: average load 55; real-time 65% peaks to 90% when loop starts again
Cubase 7.5.30-64: Poly 180: average load 65; real-time 70-75% peaks to over when loop starts again

Cubase 8.05-64: Poly 140: average load 60-65; real-time 75-90% sometimes peaks to over
Cubase 8.05-64: Poly 180: average load 70-75; real-time peaks to over

Audio buffer size 256 compared

Cubase 6.55-64: Poly 20 ASIO around 50%-55%;
256buf_C6.55_poly_20.jpg
Cubase 7.5.30-64: Poly 20: average load 55-60; real-time peaks to 60-75%
256buf_C7.5.30_poly_20.jpg
Cubase 8.05-64: Poly 20: average load 65-70; real-time peaks to 85-90%, sometime peaks to overs
256buf_C8.05_poly_20.jpg
Next post: buffer size 512

Audio buffer size 512 compared

Cubase 6.55-64: Poly 100: ASIO around 50% peaks to 55% when loop starts again
512buf_C6.55_poly_100.jpg
Cubase 7.5.30-64: Poly 100: average load 55; real-time peaking around to 65%, sometime peaks to 90%
512buf_C7.5.30_poly_100.jpg
Cubase 8.05-64: Poly 100: average load 60; real-time peaks to 90%, sometime peaks to overs
512buf_C8.05_poly_100.jpg
Next post: buffer size 1024

Audio buffer size 1024:

Cubase 6.55-64: Poly 140 ASIO around 50-55%
1024buf_C6.55_poly_140.jpg
Cubase 7.5.30-64: Poly 140: average load 55; real-time 65% peaks to 90% when loop starts again
1024buf_C7.5.30_poly_140.jpg
Cubase 8.05-64: Poly 140: average load 60-65; real-time 75-90% sometimes peaks to over
1024buf_C8.05_poly_140.jpg

Even though I’ve always been interested in benchmarking my old rig for ages, I must admit that the need of installing all sorts of trial versions for the DAWbench projects made me never try it out (total PITA). So I can’t really add any measured numbers here and I can only thank you for your efforts to show your analysis in such detail.

As my system is somewhat similar to yours (at least from a “using a core2quad perspective”) here’s my two cents:

My system:
OS: Windows 8.1 Pro 64bit
CPU: Intel Core2Quad Q6700 @ 3.0GHz (tiny bit of OC)
RAM: 8 GB
Audio: Steinberg UR28M
Graphics: ATI Radeon HD3450-> 2x 1920x1080 displays, AERO on
HD: Samsung 840 Evo (OS & Cubase), 2x 500GB Hitachi HDDs (for samples & libraries | for projects)

After installing C8 and loading some 7.5.3 projects I was quite disappointed as the asio performance meter showed roughly the same load as in Cubase 7.5.3. When I checked the sound settings, I noticed that the buffer was set to 512 samples. So I changed it to 1024 samples as this is my setting for 7.5.3 and adjusted the asio guard the way Cubase required. This resulted in some quite noticeable performance boost of about 15 percent compared to Cubase 7.5.3 which means needing less channel freezing and thus me being slightly happier. But this is not too interesting.


The really interesting thing is:

I also installed C8 on my very vintage sony vaio laptop.

The vintage system (2007, I guess):
OS: Windows 7 Ultimate 32bit
CPU: Intel T2130 @ 1.86GHz
RAM: 2 GB
Audio: m-audio fast track pro, sometime ASIO4all with onboard sound (travel)
Graphics: Intel onboard-> 1x 1280x800 display, AERO on
HD: 160GB Hitachi HDD

I still used this old machine with C5 to catch ideas when on vacation and to play around in the living room. The fact that about 3 to 4 VSTis and 5 or 6 VST FX can be enough to make it peak has been the beauty of it as limitation can really make things simple and kind of less distracting. After installing C8 the performance situation changed tremendously. I now have much more power now. 5 VSTis & 12 VST FX (EQs, compressors, Turnado, reverb) drove the performance meter to roughly 50-55% which is overwhelming for a 2007 consumer BS of a laptop.


Again, sorry for not delivering scientificly relevant numbers. But I have to say that they indeed improved Cubase’s performance (in some way). Sure, seeing the vast improvements makes me wish my desktop setup would have benefit more from C8, but I still think it’s quite impressive.

Cheers

Kind of a weird test, don’t you think?

Cubase offers additional ways of making sure ASIO stays low (Asioguard) and then you compare version with AsioGuard disabled…

It’s like testdriving a Ferrari and not using anything about the 3rd gear… Just doesn’t make sense.

If you wanna testdrive and compare, compare it to it’s fullest! Of course, keep everything as much the same as possible, but let each DAW shine at it’s best. In this case: turn AsioGuard on.

AND set the Audio Priority to ‘Boost’. Also, it’s better to use the stock plugins, rather than using 3rd party ones. That way anyone can perform the benchmark on their machines without running into compatibility issues. Just a thought :slight_smile:

Comparing performance with stock stuff seems the only way to gain a representative amount of data. This way everyone had something reliable without the nasty trial installations. We might also upload the project for direct download to make it even easier.

Any suggestions which plugins to use?

Thanks for the tests, Mike! Greatly appreciated. I knew, performance wise, 6.5 was better than 7 or 7.5, but I had “the feeling” it had improved in Cubase 8.

While it may not be the “best” test, it is certainly representative of the way a lot of us use our DAW’s. So I don’t understand what all the fuss is about. And turning on ASIO guard adds a huge amount of latency, so I would almost call it cheating. Not that I don’t like or use ASIO guard, but to achieve a fair comparison between the 3 products, they need to be on equal playing fields.

Without adding large amounts of latency to compensate for the problems introduced into the audio engine with version 7.0, Cubase would be in trouble as a product.

p.s. Forgot to thank for the time and effort :smiley:

If I have correctly understood some of this thread comments, so that Cubase 8 turns out to be the winner, I should:

-“Enable ASIO GUARD” (increasing latency playback). On 7.5 is also allowed?
-“Set Audio Priority to “Boost””. Can I do the same on 7.5 or should I leave it to Normal, so that its results are worse?
-“Use only stock plugins”, therefore forgetting about Waves, UAD, Steven Slate, Sonnox, Focusrite, Lexicon, iZotope, Softube Abbey Road,…

Excellent advices!!! :question:

However on the Steinberg website you can read:

Massive engine rebuild: The new audio engine now also means that even the most complex instruments and plug-ins run with stunning efficiency. This allows a lot lower latency while not out-powering your computer.

From the graphs the average ASIO load Cubase 7.5 is the double than Cubase 8. Is it really true?

Every time I understand less and less.:open_mouth:

I think your sarcasm was unnecessary, but I get your point regarding ASIO Guard and the Audio priority. Obviously, you would enable and set their settings the same were applicable. But I agree it would also be helpful to have a series of tests done without using these two features, in order to make it fair on C6 which doesn’t have them (at least not ASIO Guard).

However, the plugins used in this test MUST be the stock ones since not everyone has the same 3rd party plugins. Cubase already comes with a good variety of plugins that would achieve the same goal of the test (to stress your system to its max), so there’s no need to install extra stuff you may not use/need just to perform this test.

So the test is supposed to be for those who only use stock plugins. What about the 3rd party plugins users? I bet they are not a small minority. I also invite you to reread the bold lines of my post. That’s what Steinberg says on its website. Why should we go against?

By the way, I’ve been ironic, not sarcastic. The essence of sarcasm is the intention of giving pain by bitter words. It’s not my case. :wink:

I like the comparison test.
Please repeat the test using only cubase plug-ins.
and ASIO guard on and off
If you publish the test project, i can benchmark it on my system

Thnx,

By using stock plugins, ANY Cubase user can perform the benchmark test. If you start adding 3rd party plugins, then you’re alienating the people who don’t own them. Not only that but, even if you and I had Kontakt, our library collection would probably be different, which in turn makes the test inconsistent/unreliable/irrelevant. As I mentioned earlier, Cubase already comes with a variety of plugins that would give a good idea of its performance across different systems on different platforms. All you need is the test file (no extra downloads). Easy!

By the way, I’ve been ironic, not sarcastic. The essence of sarcasm is the intention of giving pain by bitter words. It’s not my case. > :wink:

By the way, your knowledge about the meaning of these words astounds me. I wonder if I was being ironic or sarcastic…

:mrgreen:

I wonder what the results would show if you upgraded your RAM memory from 8G to 16G or 32G.

Most of the Stock plugins really sound like CRAP compared to add-ons!

I’d much rather see a “real world test” where the exact same wave files, VSTi’s and 3rd party plugs (with exact same settings) are tested across the various versions.

Real world for who? Everyone uses their favorite plugins, which I assure you will be different for each individual, even if they use (insert 3rd party favorite). You could just modify the test for your own real world pleasure :slight_smile:

I hear what you’re saying, but this is what I meant…
“Real world” simply means loading plugins that are NOT designed and version optimized by Steinberg (as are the stock plugins).

So it really doesn’t matter which specific brand of plugins and VSTi’s are used as long as they are the exact same versions with the exact same settings across the tested Cubase versions.
(Of course selecting your instruments and plugins from the bigger industry brand names probably wouldn’t hurt!)

While instrument and plugin choices may or may not severely degrade performance in each individual version we will see the actual differences in the performance realized across the various versions tested.

Though for the test it would also be a good idea to really push the host. Load enough plugins and VSTi’s to get that CPU usage above 75% in the first version tested and then compare the results of each version.

Another benefit of such a test is that not only does it ensure that the same versions of the plugins are used across the tests (which would certainly not be the case using stock plugins specific to each version!) the exact same plugins and settings could then be replicated and compared across host DAWs !

CPU usage above 75% or ASIO usage over 75% real time or average ?
I think it would be easier to go over to http://www.dawbench.com/ and have a look.