ReBench - benchmarking Renoise and your CPU.


(cactoos) #1

Hi Renoisoholics!

Some time ago i had an idea about how different CPUs (mostly multi-core) can perform with Renoise, but also how good Renoise is actually running on multi-core CPUs in reality.

Being without ideas and having no inspiration since long time, i decided to do something different than music. I wanted to do something, which could help other peoples wanting to upgrade their hardware to make it easy - based on “real live Renoise performance” and also just for fun ( “i do like copy/paste” ).

Please download, test, have fun(!) and share your scores here if you like!

I am really interested how i7 4790K will perform here.

“ReBench.xrns” file starts from 1 track with 7 effects and finish at 160 stereo tracks with 960 internal effects on at once ( six effects on each channel ).

I’ve prepared version with 256 tracks and 1537 fx max, however it was taking well over 5GB memory. This one is working with my 4GB ram laptop ( running windows7 home 64bits )

No automation or devices in use - so making more complex and “real” project will affect performance probably, but even such a basic project and scenario should represent “reality”.

Please share your opinions if you think this idea could be used to create benchmarking tool as official Renoise option ( it woild be really nice to be able just to click one button and see some results :slight_smile:

Module saved under Renoise 3.01 64bits.

Download : ReBench.xrns ( or attached file )

Download :ReBench - Version extended to 256 tracks and 1537 effects( min 8 GB of ram probably required )

How to test :

Step1 : Load “ReBench.xrns” song.

Step2 : Please wait - on some configurations it will take some time ( even couple of minutes ).

Step3 : Go to your Renoise Preferences…

Step4 : Change your settings to :

Device type : DirectSound,

Sample rate : 44kHz,

Latency : 10ms

Use hardware buffers : disabled

Dithering : disabled

Step5 : Change CPU settings to “1” to test one core performance :

ReBench_CPU.jpg

Play position in pattern-sequencer should be set to the beginning :

ReBench_Seq.jpg

Step6 : Hit Play and wait till Renoise will use 90% cpu power ( as you know - you will be informed by requester, so no need for stopping it manually ) :

Then you can read position in pattern-sequencer when Renoise stopped. In this case ( above photo) it’s 26, so i would suggest to accept 25 as “last stable position”.

Last stable position is your score! In this scenario 25 RPTS ( RenoisePoints! :stuck_out_tongue: )

25 RPTS in this scenario means that your CPU can play and process : 25 separate tracks ( 25 stereo samples played together - one sample/note played on each track). Also - each track contains 6 dsp effects processing signal constantly + 1 effect on master channel.

25 RPTS in this case finally means :

your cpu is capable to play and process audio project with 25 stereo tracks with 151 internal effects on in real time ( 25*6+1 on master channel ) - using primary DirectSound device with 10ms latency and other settings as listed above.

Then you can change your CPU cores/t to maximum and test your CPU at full power!

Some tests made last night attached!


AMD FX-8320 3.5Ghz (stock) - 1CPU,latency 10ms : 24 RPTS

AMD FX-8320 3.5Ghz (stock) - 8CPU,latency 10ms : 93 RPTS

AMD FX-8320 3.5Ghz (stock) - 1CPU,latency 100ms : 30 RPTS

AMD FX-8320 3.5Ghz (stock) - 8CPU,latency 100ms : 112 RPTS

AMD FX-8320 @4.2Ghz ( OC ) - 1CPU,latency 10ms : 29 RPTS

AMD FX-8320 @4.2Ghz ( OC ) - 8CPU,latency 10ms : 104 RPTS

AMD FX-8320 @4.4Ghz ( OC ) - 1CPU,latency 10ms : 30 RPTS

AMD FX-8320 @4.4Ghz ( OC ) - 8CPU,latency 10ms : 109 RPTS


AMD FX-8350 4 Ghz (stock) - 1CPU,latency 10ms : 27 RPTS

AMD FX-8350 4 Ghz (stock) - 8CPU,latency 10ms : 100 RPTS

AMD FX-8350 4 Ghz (stock) - 1CPU,latency 100ms : 34 RPTS

AMD FX-8350 4 Ghz (stock) - 8CPU,latency 100ms : 125 RPTS

AMD FX-8350 @4.4Ghz ( OC ) - 1CPU,latency 10ms : 32 RPTS

AMD FX-8350 @4.4Ghz ( OC ) - 8CPU,latency 10ms : 115 RPTS

AMD FX-8350 @4.4Ghz ( OC ) - 1CPU,latency 100ms : 38 RPTS

AMD FX-8350 @4.4Ghz ( OC ) - 8CPU,latency 100ms : 142 RPTS


Pentium E5800 3.2 Ghz(stock) - 1CPU,latency 10ms : 15 RPTS

Pentium E5800 3.2 Ghz(stock) - 2CPU,latency 10ms : 26 RPTS

Pentium E5800 3.2 Ghz(stock) - 1CPU,latency 100ms : 28 RPTS

Pentium E5800 3.2 Ghz(stock) - 2CPU,latency 100ms : 48 RPTS


Intel i3 M330 2.13 Ghz(stock) - 1CPU,latency 10ms : 16 RPTS

Intel i3 M330 2.13 Ghz(stock) - 4CPU,latency 10ms : 30 RPTS

Intel i3 M330 2.13 Ghz(stock) - 1CPU,latency 100ms : 26 RPTS

Intel i3 M330 2.13 Ghz(stock) - 4CPU,latency 100ms : 50 RPTS



Dithering really disabled? Optimization potential?
Any really heavy rns module? [in terms of cpu load]
Any really heavy rns module? [in terms of cpu load]
(toimp) #2

Great idea. I will test this with my i7 3770k, when I’m at home.


(kopias) #3

Spec:

Notebook Lenovo Y580 (2012) i7-3610QM

http://www.notebookcheck.net/Intel-Core-i7-3610QM-Notebook-Processor.72681.0.html


1CPU,latency 10ms : 33 RPTS

8CPU,latency 10ms : 123 RPTS

1CPU,latency 100ms : 44 RPTS

8CPU,latency 100ms : Ta! Amiga Rulez!!!

P.S. cactoos: Greetings from Poland.


(toblerpone) #4

I think it’s important that the power settings are set to 100% (that will be with any ‘turbo’ functions).

As long with disabling any tools, and making sure the same visualisation settings are present.


(toimp) #5

i7-3770k, 16GB RAM, W7 64bit, Renoise 3.0.1 64bit

1 core = 38 RPTS

8 cores = 146 RPTS

My default settings: Dithering enabled + 35 ms latency =Ta! Amiga Rulez!!!


(RANSOM) #6

it was a fun thing to do

i5 2.4 ghz (macbook pro) 8gb ram

soundcard is komplete audio 6 - 44 khz, latency 28ms

os x 10.9.5

4 cores = 75 rpts


(ffx) #7

?

Renoise 3.01 32bit, Core i7 4770, OSX 10.9.4, 10ms latency, 44,1kHz, Firewire… maybe the range of test could be expanded ? :stuck_out_tongue:

1 core = 43 RPTS

8 cores = Ta! Amiga Rulez!!! / EOF

Latency at 93ms (maximum available here), 1 Core,

audio starts to stutter at ~58 RPTS, although CPU is at ~72%… Strange… A bug in the OSX version?


(danoise) #8

Here are my results.

Renoise 3.0.1 x64, onboard sound, win8.1…

Intel i7 4790S 3.20GHz - 1CPU,latency 10ms : 41 RPTS

Intel i7 4790S 3.20GHz - 8CPU,latency 10ms : 143 RPTS

Intel i7 4790S 3.20GHz - 1CPU,latency 100ms : 53 RPTS

Intel i7 4790S 3.20GHz - 2CPU,latency 100ms : Ta! Amiga Rulez!!!


(cactoos) #9

ReBench - Version extended to 256 tracks and 1537 effects


(danoise) #10

ReBench - Version extended to 256 tracks and 1537 effects

I’d keep the old version around, just in case memory usage becomes a problem :slight_smile:

Renoise 3.0.1 x64, Windows 8.1, Steinberg UR22 (direct sound, not ASIO)

Intel i7 4790S 3.20GHz - 1CPU,latency 10ms : 41 RPTS

Intel i7 4790S 3.20GHz - 8CPU,latency 10ms : 156 RPTS

Intel i7 4790S 3.20GHz - 1CPU,latency 100ms : 53 RPTS

Intel i7 4790S 3.20GHz - 8CPU,latency 100ms : Ta! Amiga Rulez!!! (63,4% CPU usage)

Same system, almost same results :smiley:


(cactoos) #11

both versions available

Download :ReBench.xrns

Download :ReBench - Version extended to 256 tracks and 1537 effects( min 8 GB of ram probably required )


(ffx) #12

Did you disable spectrum view?


(AzureKevin) #13

This is awesome, thanks for making it.

I just tested this one two of my systems:

My current dedicated music machine, an FX-8350 @ stock 4.0 ghz:

FX-8350 @ 4.0 ghz 1 core: 43

FX-8350 @ 4.0 ghz 8 cores: 144

And my gaming rig i5-2500K quad core that I have overclocked from 3.3 to 4.6 ghz:

i5-2500K @ 4.6 ghz 1 core: 51

i5-2500K @ 4.6 ghz 4 cores: 188

So yeah, there you go. I can test the 2500K at stock if you are curious, but it has no problems overclocking to 4.6.


(ABįSƧȔš ṼØȊÐƎ) #14

Great Idea. Will test when home.


(ffx) #15

Redone with Rbench x256:

Renoise 3.1 rc1 64bit, Core i7 4770, OSX 10.9.4, Firewire…

10ms,44,1kHz,1 core = 53 RPTS

10ms,44,1kHz,8 cores = 207 RPTS

75ms (max. possible),44,1kHz,8 cores = 256 / EOF /Ta! Amiga Rulez!!!


(ABįSƧȔš ṼØȊÐƎ) #16

i7-6700k @ 4.00Ghz (stock)
Renoise 3.01 Win 7(64)

1 core (10ms): 51

8 cores (10ms): 144

1 core (100ms): 65

8 Cores (100ms): Ta! Amiga Rulez!!!


(ffx) #17

Would be interesting to know what kind of audio interface you people used while testing…?


(ABįSƧȔš ṼØȊÐƎ) #18

Would be interesting to know what kind of audio interface you people used while testing…?

I just used my onboard sound so I am assuming my intereface doesn’t come into it?


(ffx) #19

I just used my onboard sound so I am assuming my intereface doesn’t come into it?

Yes, I think this is the fastest you can have, PCIe or onboard.

I just wonder that OSX seems to have a much better multicore scheduler than Windows. Or the OSX version is so much faster because of a different compiler…

To be sure here, I would like to suggest to you to ensure that you all use

  1. “desktop” / 100% power profile like toblerpone suggested

  2. The asio driver, no creepy Windows WDM or WASPI driver

  3. Write down what kind of audio interface was used


(ABįSƧȔš ṼØȊÐƎ) #20

Yes, I think this is the fastest you can have, PCIe or onboard.

I just wonder that OSX seems to have a much better multicore scheduler than Windows. Or the OSX version is so much faster because of a different compiler…

To be sure here, I would like to suggest to you to ensure that you all use

  1. “desktop” / 100% power profile like toblerpone suggested

  2. The asio driver, no creepy Windows WDM or WASPI driver

  3. Write down what kind of audio interface was used

Ok will do when I can (which may be tommorrow?) and will post results.