C6 Performance 32-bit/64-bit. Tests and results !

Hi all.

Everybody’s been wondering about whether 64-bit Cubase (6) is any faster than 32-bit version. So read this article and you’ll get some guidelines about it. Of course these tests have been executed with only my own PC, but the results are comparable in a general sense to other systems also. Certainly the results will not be vice versa on other systems anyway – at least I don’t know how they would.

I divided the tests in 2 categories:

  1. Passmark Performance Tests (done with Passmark’s Performance Test v7.0 x86 and x64 versions)
  2. Cubase Tests with both x86 and x64 versions

The details of the tests are as follows:

1) PASSMARK TESTS:

A. CPU test:

  • Integer Math
  • Floating Point Math
  • Find Prime Numbers
  • SSE
  • Compression
  • Encryption
  • Physics
  • String Sorting

B. Memory test:

  • Allocate Small Block
  • Read Cached
  • Read Uncached
  • Write
  • Large RAM

C. 2D graphics test:

  • Solid Vectors
  • Transparent Vectors
  • Complex Vectors
  • Fonts and Text
  • Windows Interface
  • Image Filters
  • Image Rendering

D. 3D graphics test:
(not relevant)

E. Hard drive test:

  • Sequential Read
  • Sequential Write
  • Random Seek + RW

2) CUBASE 6 (x86 and x64) TESTS:

A. Opening time for Cubase 6 with the following plugins installed:

  • Waves 7 plugins pack (200+ plugins – jBridged in x64 Cubase 6)
  • Line 6 audio plugins (~10 large plugins)
  • Halion Symphonic Orchestra 1.5
  • Groove Agent 3
  • IK Multimedia T-Racks 3 mastering bundle

B. Opening time for large project files:

  • This project utilizes plugins from ALL sections from the section a) except for the T-Racks 3

C. Downmix time of the same project:

  • The plugins (some) are present, like mentioned in section b) + there’s a lot of automation, MIDI + AUDIO events, Inserts/Sends/EQ present also

D. Mastering time of the same project:

  • This project utilizes T-Racks 3 (demanding on the CPU), Waves 7 and Steinberg’s native plugins

E. Time-stretching time of the downmixed file of the same project:

  • Time-stretching was done with Poly Complex algorithm and the file was expanded 200% of the original length

ONE NOTE: Of course the 3D measurements + results have got nothing to do with Cubase 6 performance whatsoever but I had to run them also to have a Passmark Rating Value done by the Passmark’s software; It demands that all the sub-tests be performed in order to give you a general Passmark rating.

TEST SETUP:
First of all, I spent some 8 hours yesterday tweaking my PC’s BIOS to maximum performance values and I finally found those. This was done so that other Windows services would have the minimal effect on the test results; on a slower PC the relative effect on the Background services + the rest would be bigger so the results wouldn’t be accurate. Of course there are no “accurate” or “absolute” results or truth about the Cubase 6 performance to begin with. It always depends on the system.


Hardware:

  • ASUS P7P55D Motherboard with the latest BIOS (revision 2003, 25.1.2011)
  • 8GB DDR3 RAM @ 1540MHz
  • Intel i7 core 860 processor @ 3,67GHz (4 physical cores / 8 hyper-threaded)
  • ATI Radeon HD 5700 1GB Dual [over-clocked with ATI Overdrive to the maximum values (CPU: 960MHz (default: 400)/ RAM: 1445MHz (default: 600))]
  • Apacer A7202 SSD 64GB (w/ latest firmware) dedicated for the Operating System. WRITE-CACHING is ENABLED. R/W-average speed: ~240MB/s (Advanced System Optimizer 3 benchmark)
  • 2TB RAID 0 Stripe-set (hardware) for the audio files and applications (like Cubase itself) (4*500GB Western Digital Black Series SATA2 hard-drives. WRITE-CACHING is ENABLED. R/W-average speed: ~1,3GB/s (Advanced System Optimizer 3 benchmark))

Software:

  • Windows 7 64-bit SP 1
    • All the latest drivers installed (checked manually + with Advanced System Optimizer 3 & DriverFinder 2.04 & Driver Genius 10)
    • All the updates & patches installed via Microsoft Update
    • Steinberg Power-scheme is ENABLED.
    • Audio priority: BOOST
    • Windows Pagefile (12,3 GB static) + all temporary files set on RAID 0-drive (D-drive)
  • Cubase 6 32-bit & 64-bit + plugins (like described above)

BIOS:

  • Intel SpeedStep: ENABLED
  • Intel Turbo-mode: ENABLED

OK, there’s a setup configuration.

And at this point I’d like to emphasize that my meaning is NOT to be showing off my PC!! I’m trying to be of help to others here, so if someone is annoyed out of his head at this point, I’d suggest you to stop reading this and do something else instead. No inappropriate or irrelevant replies needed also. Thank you.



Then to the interesting part: The results.

PASSMARK TESTS (which tell the overall performance of x86 and x64-systems)

Performance Test 7 32-bit:

CPU Mark: 5613.6
2D Graphics Mark: 742.7
Memory Mark: 1231.6
Disk Mark: 2777.7
3D Graphics Mark: not relevant
PassMark Rating: 2598.5

Featured and interesting sub-scores:
CPU - Integer Math: 674.6
CPU - Floating Point Math: 2756.9
Memory - Read Cached: 2635.2
Memory - Write: 1640.9
Memory - Large RAM: 2158.1


Performance Test 7 64-bit:
CPU Mark: 7608.4
2D Graphics Mark: 791.1
Memory Mark: 2173.8
Disk Mark: 2828.4
3D Graphics Mark: not relevant
PassMark Rating: 3333.0

Featured and interesting sub-scores:
CPU - Integer Math: 2885.8
CPU - Floating Point Math: 3405.6
Memory - Read Cached: 2993.5
Memory - Write: 1742.8
Memory - Large RAM: 5254.1

CONCLUSION:
Performance advance for 64-bit score over 32-bit score (formula: [[x64-value / x86-value] * 100] – 100):


28.3%

=====================================================

CUBASE 6 TESTS

The start-up times (for Cubase itself nor the project start-up time) didn’t vary namely at all on my system, so I won’t present any numbers for no reason.


Exporting the MIXDOWN of a large project to 96kHz/32-bit .WAV-file:

Cubase 6 32-bit: 196 seconds
Cubase 6 64-bit: 80 seconds

Performance advance for 64-bit score over 32-bit score (formula: [[x86-value / x64-value] * 100] – 100):


145%


Mastering project to the corresponding downmixed file to 320kbps .MP3-file:

Cubase 6 32-bit: 251 seconds
Cubase 6 64-bit: 172 seconds

CONCLUSION:

Performance advance for 64-bit score over 32-bit score (formula: [[x86-value / x64-value] * 100] – 100):


45.9%


Time-stretching the corresponding downmixed file (algorithm: MPEX - Poly-Complex, time value: expand, 200%):

Cubase 6 32-bit: 1655 seconds
Cubase 6 64-bit: 845 seconds

CONCLUSION:

Performance advance for 64-bit score over 32-bit score (formula: [[x86-value / x64-value] * 100] – 100):


95.9%


So the answer to the ultimate question: “HOW MUCH FASTER IS CUBASE 6 64-BIT OVER 32-BIT???” IS:


~89% faster


This is a weighted average of the sub-results with the following weighted amounts:

Passmark test: 10%
Cubase 6 (mixdown): 30%
Cubase 6 (mastering): 30%
Cubase 6 (time-stretch): 30%

So for me, Cubase 6 64-bit is ALMOST TWICE AS FAST as Cubase 6 32-bit.

The conclusion:

For me the result is pretty clear; Cubase 64-bit is way more efficient than 32-bit (or x86) version and I’m going to use the 64-bit version only from now on. And of course the results would vary on other systems but this test was merely meant to be a baseline for others.

Also please notify that the mastering project is heavily loaded by the 32-bit T-Racks 3 plugin (+ jBridged Waves’ plugins) and the time-stretching uses natively 64-bit Steinberg’s own time-stretching plugin ONLY! Imagine when ALL the plugins that we use are natively 64-bit. Then the performance advance for 64-bit Cubase 6 is even higher!
I hope I could shed some light for this performance issue. For me the performance difference actually was a bit of surprise – but a nice one! Steinberg is proven to have coded their 64-bit version very well indeed and not just with the attitude “well, we have to do this too, so let’s do it quickly with our left hands”.

This document can be downloaded in .DOC format here:

http://www.pc-professor.fi/Cubase6/performance/report.doc

© 13.3.2011 Tommy Dee

Thanks Tommy! Very interesting indeed. I must admit I had thought 64bit Cubase was faster than the 32bit version, but I’d never stopped long enough to think of doing tests.

Interesting that it is nearly twice as fast! And as you say, once more plugins have gone natively 64bit and we don’t need to bridge, things will only get better!

It’s interesting that you chose to use a project with so many 32bit plugins (Waves, T-Racks). I bet the results would favor x64 even more if you used native x64 plugins.

Did you do a full reboot of windows in between each test pass? I suspect that the major benefit is the accessibility of more RAM resulting in less thrashing to disk. But also remember that Windows7 uses excess RAM as a disk cache. If I open a project with a large sample set, like Sampeltekk’s The Big One piano, it takes more than a minute to open the first time, but subsequent times I open the project without rebooting and it opens in seconds. Because all of the samples are loaded from the RAM cache instead of disk.

Aloha Tommy,

Great work. Major Mahalo.

You mentioned ‘other systems’.
Perhaps someone will do a test similar to your test to see if Mac 'puters
have similar results.

Thanx again.
{‘-’}

Positive stuff. I’m an inveterate early adopter and I’ve paid the price in the past, but my leap of faith into 64-bit has been a VERY positive experience, having expected the worst. Cubase/Steinberg have left the rest standing. Although I use “proper” guitar amplifiers (valve/tube of course) I know what an impact the VST guitar stack would have made on me when I was working in a bedsit through headphones. Wow!

Workflow is great with only a couple of little niggles. I just did a five hour session which was smooth, uncomplicated, plugin heavy - including Jbridged BFD2 - flawless, rewarding.

I’m happy. I’m smiling! :slight_smile: :slight_smile: :slight_smile:

The downside seems to be the increased CPU usage while bridging 32Bit plug-ins (after my own testing). So it would also very helpful to see your values the VST performance/CPU usage in C6 32/64Bit.

Very interesting reading, thank you for sharing this.

Sincerely
LarsErik

Tommy,

Thanks for putting in the work and the feedback.

Dave

Nice!

What’s the Steinberg Power scheme?

nice1 tommy, interesting stuff…

very interesting.

I’d be interested to see something similar without all the 32bit plugs - to eliminate jbridge/vstbridge. In fact to get a true representative test maybe you should elminate all 3rd party stuff.

I’d also like to see the speed difference between 32bit on a 32bit OS compared to 64bit on a 64bit os. It ‘could’ just be that cubase32 doesn’t run well on a 64bit os ?

Yes and all-native (making sure you don’t use any of the old 32bit included plugs that have a “II” symbol next to them) test with plugs that are both native x86 and x64 is the best way to go.

And you also MUST make sure you’re rebooting between each test so you’re not getting free RAM caching of disk content done by Win7. I’m an x64 fan but I’m cautious about these numbers because I’m not sure what the methodology was.

Yes,64Bit native plug-ins have some performance boost, but the problem is that we have not all plug-ins 64Bit and so without a Bitbridge most plug-ins are not usable. Also the well known UAD plug-ins, Waves etc.
If I compare a project used with Cubase32Bit the VST performance meter shows something about 50%. The same project with Cubase 64Bit and some plug-ins bridged (jBridge) needed VST performance is about 55%.

If a BitBridge is needed, the performance is not better then with a 32Bit Cubase. Not to mention that some features are missing (32Bit VST3 plug-ins cannot be bridged with jBridge).

I’m not sure if I mentioned, but I DON’T HAVE the x64-versions of all my plugins. OF COURSE I would have done the tests natively with corresponding plugins, but LIKE I MENTIONED in the test document “If the plugins were all native x64 versions, the performance benefit for 64-bit Cubase would be even bigger”.

And yes, I rebooted my PC between EVERY SINGLE ONE test with the similar setup always on.

Thanks for your idea. I’ll do a similar test with ONLY Steinberg’s own plugins. Gotta amke a project with all the Steiny’s native plugs included and rip the last bit of performance out of my PC :slight_smile:

Good to know! As an x64 user, I’m actually SHOCKED at how much better your tests showed x64 runs.

On the plugins–Cubase comes with a ton of built-in native x64 plugs, so you could easily test with those instead of 3rd party plugs even if you don’t own anything else. For example, DAWBench uses the multiband limiter built into Cubase for its cubendo tests.