Ableton vs. SONAR - CPU Shock!!

[headquest]

Looking forward to Ableton Live 6 including support for dual core processors, I have been wondering how much of a performance increase I can expect to see on my Intel Core Duo 2.16Ghz laptop. When I upgraded to this computer I noticed major performance improvements with Sonar 5 Producer, which has Multicore support.

Thinking this through it occurred to me that it would be interesting to devise a simple test to compare how much CPU Live 5 and Sonar 5 each use… My expectation was that Ableton would be far less efficient, not simply because it does not support the dual core architecture, but also because users have regularly complained that Live is a "CPU hog". Sonar of course has a reputation for being highly optomised to run efficiently on Windows systems, and I expected to see very impressive results.

However, the actual results are the opposite of my expectations, leaving me somewhat shocked!

Here's the test I did, followed by the results:

I decided to check out using a single MIDI file playing through Green Oak Crystal (which is known to be CPU intensive), all processed through the equally CPU intensive Ambience reverb. That's it! The MIDI file was about 2 minutes long, and I measured changes in CPU throughout that duration using the Windows Task Manager. The Performance tab shows both CPU cores as a graph as well as giving the overall (average) CPU use. Also included in the Page File Usage, although I confess that I don't understand the significance of this - hopefully somebody can enlighten me!

And now the results:

SONAR 5 PRODUCER:
Both CPU cores were clearly in use, with a roughly even spread.
The overall CPU usage ranged from 18% - 33%
PF Usage was shown as a constant 386.

ABLETON LIVE 5:
One CPU Core was predominantly in use, although there were flickers from the other core, presumably due to Windows
The overall CPU usage ranged from 15% - 29%
PF Usage was shown as a constant 426.

These results are cumulative - I repeated the test several times to verify the figures given.

Obviously I am shocked to see from these figures that Ableton Live - in spite of its reputation and even without dual core support - performed measurably better than Sonar. This completely confounded my expectations. Quite possibly this is a flawed test of course, due to my limited knowledge of how CPU usage works. I would be grateful for any constructive feedback that might explain why the results are so surprising and help me interpret them from a different perspective if necessary.

Presumably when Live 6 appears in the Autumn, with dual core support, it will take an even more marked lead?

Cheers!

[Hunter]

I think there were lots of complaints on the sonar forum about the cpu use with 5.02 update being really high

[whyterabbyt]

I think there were lots of complaints on the sonar forum about the cpu use with 5.02 update being really high

Yup, but if I recall, it turned out to be something slightly different; seemed that earlier versions werent doing dithering on-the-fly, even when it was switched on, so their CPU usage was artificially low. Turn off dithering in 5.02 and the CPU drops to 'normal'.

BTW, headquest, I dont think the Windows Task Manager is the best test for CPU usage. Im sure some hosts grab more CPU cycles than they're necessarily using so that their own in-program CPU reporting is more accurate...

[WilliamK]

I could be wrong, so please correct me if I am, but doesn't Sonar have more bits on the Mixer part, and also more quality there, that could be related to more CPU usage versus Live

Wk

[whyterabbyt]

Aha; its actually a sticky in the Sonar Forum:

http://forum.cakewalk.com/tm.asp?m=750994

[fandango]

It's definately the "dithering on by default" problem, as whyterabbit says.

There was a sticky notice in the Sonar forums that says something like "CPU problems with 5.2?" or something similar.

[edit - boooo... slipped in there ]

[WillieJenkins]

headquest,

just some things to consider.

1) Are the latencies the exact same? Along the same lines, are you using ASIO drivers with Sonar?

2) As whyterabbyt stated, you'd be better off checking in-program CPU numbers, as I know Sonar will tend to grab more cycles then it needs as a buffering device (buffering for cpu spikes, not for audio )

3) Was delay compensation enabled in Live? Sonar has full PDC at all times, so Live should have it on to make the comparison more equal.

Another major thing to consider is that Ableton Live starts falling apart when the cpu hits about 60%, meaning crackles, dropouts, and general weirdness. Sonar on the other hand can be pushed into the 90%'s before it start to act weird.

I'm not trying to push Sonar over Live here because they're both my main hosts, I'm only posting this stuff, because I know from real world usage that Live has way less runway to play with compared to Sonar. I often start projects in Live, then re-create the exact same project in Sonar and go back and forth. 100% of the time The project will become unusable in Live before Sonar.

Just my .02

-Eric

[WillieJenkins]

ahh yes, and the dither

[kp]

Each program's CPU meter is much worse to use than Windows Task Manager: you have no idea what it's measuring so you can't compare readings from 2 different applications.

If I remember rightly, for example, Cakewalk slightly changed what the SONAR CPU meter measured between v4 and v5 (one included some MIDI monitoring, one didn't)...that's 2 versions of the same software measuring differently!

[opia]

another thing to consider

maybe they both run rather close in the cpu dept

and perhaps
just perhaps

some claims are exaggerated on every side

[Stupid American Pig]

I think that there are so many configuration options on every host that a direct comparison is damn near impossible. I agree you may want to look at the task managers CPU meter, or install process explorer so you can get a better look at what is going on with each

[kp]

Or even set up a performance monitor so you can see CPU usage throughout the song. Process Explorer is indeed a good tool to use too.

[headquest]

Thanks to everyone for the replies, advice and tips

As a rare visitor to the Sonar forum I had not spotted the sticky. It's a slightly difficult one to compare of course because I believe that Ableton Live also has dither "enabled" (but not as a selectable option).

Anyway, in a moment I will rerun the test in Sonar and:

1) Disble the dither and
2) Switch off the 64-bit mix option as well.

I should point out that the comparisson is in no way an attempt to suggest one program is better then the other by the way - I am a registered owner/user of both and love 'em both!

Some specifics:

1) Are the latencies the exact same? Along the same lines, are you using ASIO drivers with Sonar?

Yes, and Yes. Soundcard settings using ASIO are identical for both.

2) As whyterabbyt stated, you'd be better off checking in-program CPU numbers, as I know Sonar will tend to grab more cycles then it needs as a buffering device (buffering for cpu spikes, not for audio )

I've read various bits of "opinion" about Live's CPU meters, which is why I used the Windows Task Manager. No idea about Sonar's CPU meter, but the Ableton one is "controlversial" .

That said, I will compare the WTM meter with the Sonar one when I redo the test.

3) Was delay compensation enabled in Live? Sonar has full PDC at all times, so Live should have it on to make the comparison more equal.

Yes, I keep PDC on by default in Ableton, because I usually drop in the Perfect Space convolution reverb during the mix phase of a song.

Another major thing to consider is that Ableton Live starts falling apart when the cpu hits about 60%, meaning crackles, dropouts, and general weirdness. Sonar on the other hand can be pushed into the 90%'s before it start to act weird.

On my new system I have pushed Ableton much further than 60% before the deterioration became too noticeable, but I agree with your point there - Sonar holds up better with large projects under strain!

I'm not trying to push Sonar over Live here because they're both my main hosts

Ditto!

another thing to consider
maybe they both run rather close in the cpu dept
and perhaps, just perhaps, some claims are exaggerated on every side

That may be the truth of it yes. But as I said at the start, my assumption was that the Dual Core processor support in Sonar would make a bigger difference.

I have certianly found that Sonar 5 runs far better on the Dual Core system than it did on my previous P4 computer. But Live 5 has no dual core support at this point... which suggests that when Live 6 is out the performance should be even better than in this little test.

Can anyone advise what the difference in PF Usage signifies (if anything)?

Cheers everyone!

[nollock]

Also, unless you compare *all* the common features you wont get fair picture as they quite likely vary in relative performance across different features. It's also probable that one host may have a higher idling usage. Just testing a simple midi->synth--> fx will exagerate that and will not be a test of perormance in a typical situation. So..

The only practical way to test would be to set up a fairly *typical* project that uses a range of features, and do it as close to the same in each host. So somthing that uses HD recording/midi/synths/fx/beatstretching/ of course any synths/fx used should be the same plugins in each host.

[nollock]

Another thought... if you use the task manager to measure the cpu usage i'd sugest minimizing the apps while testing so that any screen redrawing code doesnt affect the measurements. Cause the task manager will be measuring the whole CPU usage, the GUI and Audio threads.

PF usage is just how much memory they are using.