Audio Performance of different Cubase Versions compared

[jose7822]

Fretthefret,

As Peakae noted, there are already such a tests, created a while back, found on the website linked in his post above. They’re called the “Dawbench DSP Universal 2014” and “Dawbench VI Universal 2014”. The former focuses on audio projects while the latter is VSTi oriented. However, you need to download and install the plugins specific for each test, as noted under each test description. That’s the part I personally find inconvenient only because I wouldn’t want to install software I’ll never use, except on these benchmark tests. I don’t want to tinker with the stability I enjoy right now, especially since I just recently formatted my system. Also, I think that more people would be inclined to run tests on their systems if they included the stock plugins. It’s more practical.

When creating the test, you would obviously stick to the stock plugins that are common across the various versions of Cubase. For example, you wouldn’t include the new Quadrafuzz plugin, that is only available for Cubase 8 users, in the test. As I’ve said several times, Cubase has plenty of plugins of different types and CPU usage to choose from, the majority common accros several versions. They can illustrate real world performance just as well as any 3rd party plugin.

Anyway, I think I’ve made my point already so it’s redundant for me to keep discussing this matter. Plus, I think I’m on the loosing side, lol.

Peace!

[fretthefret]

yes Jose, I don’t completely disagree!

I was just noting that by using the plugins included with each version there is an inconsistency with the testing procedure.

Each plugins/effect used is DIFFERENT (same in name only!) - It is tailored and optimized to its respective version and thus NOT a common across the testing.
Plugins and effects included with version 6 may be different versions (and thusly optimized differently) than those included with 8.0.5 for example.

So to keep a common performance environment and testing procedure it would be far more accurate to test with 3rd party plugins and effects… like those from WAVES and Native Instruments.
This way the exact same versions and settings could be replicated across the tests.

Like you, I’m also not into the idea of installing plugins I will never use in my daily operations of Cubase Pro 8.
(Since I’m on widows every install adds to the Registry)

But since there are those that do these test and publish results… the only way to be “real world” accurate is to be consistent with the testing procedure across versions!
By using the version-specific “stock” plugins/effects (which as I said are the same only in name and not program version!) there’s a glaring inconsistency introduced.

[MikeWK]

Comments:

Interesting, many comments to read, discussion is active. That’s what I intended to do.

The test I used fits very well to my needs because I use Kontakt often. If someone has time
he could replace Kontakt with Halion Sonic SE (that runs in 6.55 too). But not me, I’m too busy.

The ASIO Guard thing: This adds additional latency to your system. ASIO Guard does not help
if you play a big stack of VSTi live. Some people discuss about U-HE Diva and I tested myself
the update of Arturias Analog Lab including presets from the new Matrix 12 synth. Heavy CPU
consumption indeed…

I made some additional tests (results follow soon) with ASIO Guard enabled. And I can say,
yes, Steinberg increased the audio performance in Cubase 8.0 compared to 7.5 but with ASIO
Guard enabled. They improved the Guard performance. But not without.

Remember: ASIO Guard enabled adds latency, the feel is more and more lazy if you increase
buffer size. Tanant wrote:

And turning on ASIO guard adds a huge amount of latency,
so I would almost call it cheating.

And I wrote: "

Especially the feel of Cubase 8 is quite lazy.

BTW: I had some crashes too… fine, if your customer sits beside you.

Latency values follow…

[MikeWK]

My latency values (RME HDSP9632):

Audio buffer size 1024:
Cubase 6.55, 7.5.30, 8.05 w/o ASIO Guard
Input Latency: 24.240ms
Output Latency: 25.011ms
Sample-Rate: 44.100 kHz
no pulldown

Cubase 7.5.30-64 with ASIO Guard
ASIO-Guard-Latency: 46.440ms

Cubase 8.0.5-64 with ASIO Guard
low - ASIO-Guard-Latency: N/A, ASIO Guard deactivated !
normal - ASIO-Guard-Latency: N/A, ASIO Guard deactivated !
high - ASIO-Guard-Latency: 92.880ms

Audio buffer size 512:

Cubase 6.55, 7.5.30, 8.05 w/o ASIO Guard
Input Latency: 12.630ms
Output Latency: 13.401ms
Sample-Rate: 44.100 kHz
no pulldown

Cubase 7.5.30-64 with ASIO Guard
ASIO-Guard-Latency: 23.220ms

Cubase 8.0.5-64 with ASIO Guard
low - ASIO-Guard-Latency: 23.220ms
normal - ASIO-Guard-Latency: 23.220ms
high - ASIO-Guard-Latency: 92.880ms

Audio buffer size 256:

Cubase 6.55, 7.5.30, 8.05 w/o ASIO Guard
Input Latency: 6.825ms
Output Latency: 7.596ms
Sample-Rate: 44.100 kHz
no pulldown

Cubase 7.5.30-64 with ASIO Guard
ASIO-Guard-Latency: 23.220ms

Cubase 8.0.5-64 with ASIO Guard
low - ASIO-Guard-Latency: 17.415ms
normal - ASIO-Guard-Latency: 23.220ms
high - ASIO-Guard-Latency: 92.880ms

PS: you have to distinguish between Latency and ASIO-Guard-Latency! It’s
not the same, please have a look into the manual.

[MikeWK]

Results, summary, ASIO Guard enabled:

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 performance still normal and not set to boost, because it did not improve
the performance. Perhaps I will try that sometime again…

Test with ASIO Guard as requested:

Audio buffer size 256: ASIO Guard low

Cubase 8.05-64: Poly 140: average load 80-90; real-time 5-10%

Audio buffer size 256: ASIO Guard normal

Cubase 8.05-64: Poly 140: average load 70-75%; real-time 5-10%
Cubase 8.05-64: Poly 180: average load 80-90%; real-time 5-10%

Audio buffer size 256: ASIO Guard high

Cubase 8.05-64: Poly 140: average load 60-65; real-time 5-10%
Cubase 8.05-64: Poly 180: average load 70-75; real-time 5-10%
Cubase 8.05-64: Poly 280: average load 80-95; real-time 10-35%

Audio buffer size 512: ASIO Guard low
setting is not different to normal

Audio buffer size 512: ASIO Guard normal

Cubase 8.05-64: Poly 140: average load 70-75; real-time 5%
Cubase 8.05-64: Poly 180: average load 75-85; real-time 5%

Audio buffer size 512: ASIO Guard high

Cubase 8.05-64: Poly 140: average load 60-65; real-time 5%
Cubase 8.05-64: Poly 180: average load 70±; real-time 5%
Cubase 8.05-64: Poly 280: average load 80-95; real-time 5-10% peaks to 15%

Audio buffer size 1024: ASIO Guard low
N/A, ASIO Guard disabled

Audio buffer size 1024: ASIO Guard normal
N/A, ASIO Guard disabled

Audio buffer size 1024: ASIO Guard high

Cubase 8.05-64: Poly 140: average load 55-60; real-time <5%
Cubase 8.05-64: Poly 180: average load 65-70; real-time <5%
Cubase 8.05-64: Poly 280: average load 80-95; real-time <5%


Audio buffer size 256: ASIO Guard ON

Cubase 7.5-30-64: Poly 80: average load 70-75%; real-time 80-95%, somtimes OVER
Cubase 7.5-30-64: Poly 140: average load 80-90%; real-time OVER

Audio buffer size 512: ASIO Guard ON

Cubase 7.5-30-64: Poly 80: average load 50-55%; real-time 55-65%
Cubase 7.5-30-64: Poly 140: average load 60-70%; real-time 70-75% OVER when loop restarts

Audio buffer size 1024: ASIO Guard ON

Cubase 7.5-30-64: Poly 140: average load 50-60%; real-time 70-85%
Cubase 7.5-30-64: Poly 180: average load 60-70%; real-time 70-80% OVER when loop restarts

PS: if a Poly value is missing that means OVERs. Sometimes I’ve noticed when the OVERs
are close to the limit or slightly above.

[MikeWK]

I can post the pictures too, if requested.

A table comparing all values would be fine, perhaps someone could do this.
So keep on discussing and most important: keep on making music!

That’s it for now, regards, Mike

[peakae]

Interesting read, and thanks for taking the time. It’s impossible to come up with benchmarks that cover every aspect and usage of a DAW. But it can be useful, even if it only is to compare ones own system to others.

[MikeWK]

@peakae: you’re right. But this test fits very well to me. I did not use the latest one that is using
Kontakt 5, because I made the test with Cubase 6.55 a year ago or so and only DAW test version 2012
was available.

[sonic74]

How is it even possible that they screwed up CPU usage? Isn’t it supposed to be optimized with each version?
My system halts unexpectedly when ASIO guard is on. When I add a new audio track for example, Cubase halts for about a minute. When ASIO guard is off, everything is normal.

[Patanjali]

Of course, ASIO Guard adds latency, but it is ONLY to the stuff that is NOT on the critical path. Basically, it gives some CPU breathing space to process the playback channels.

It really blows out the playback latency, so it can optimally bias CPU to providing smaller recording latency. It does not matter if the playback latency is 100ms, it just means that there is much less chance that the recording channels will suffer.

But, if the playback channels are fairly constantly full on, with no chances for the CPU processing of them to get well ahead, ASIO Guard is just delaying the inevitable crunch point.

In general, it is not as if ASIO Guard’s latency is adding it to EVERYTHING, otherwise it is just extra buffers.

However, as pointed out, it is not good for live, as there, EVERYTHING IS on the critical path, though it would help with backing tracks.

[MikeWK]

Some of you criticized that the DAW bench is not a Cubase bench.

OK folks, I’ve made a padshop benchmark for cubase 6.55 in January.
I created 30 tracks all equal with a padshop in each track.
Increasing from one note to seven notes played each with a pause in between.
You can use the markers 1…7 to navigate. The more notes padshop is playing
the load increases. You can mute tracks too.

I resaved it in 7.5.40 (yes, latest) and 8.05 with a 2x1920x1080 screen layout. And
my padshop was used as VST3.

Play along with those. But be careful: if you overload cubase it becomes
unresponsive and I had several crashes too (8.05).

NI Kontakt is somehow excluded from ASIO Guard. Can’t remember if it is the same with
padshop. May be, it’s using samples too.

Please let me know your results. Regards, Mike
Padshop_bench_MWK_2015.zip (522 KB)

[MikeWK]

Did a quick test with Cubase 8.0.10, 256 samples buffer and ASIO guard normal using
the DAW-Bench-VI 2012 I used before.

Unfortunately the VST2 performance (here using Kontakt) was not improved compared to
8.0.5. which I hoped. For my opinion ASIO load even increased a little bit.