Proposed Benchmark testing for T2

[zeoy]

That's odd; it shouldn't matter what you're recording, even if it's empty space, because as far as the computer's concerned, it's sampling at 44.1, with a bandwidth of 24bits.

Something about this test ain't right. I gotta think.

Don't know if it shouldn't but in practice it does make a difference for sure. Maybe the CPU load is heavier when a reverb plugin process sum busy input rather a silent wav. Anyway, I did the test using a 44.1 16bit drum loop and I managed to run smoothly 27 tracks. With the 28th the CPU indicator went occasionally red but it continued to play flawlessly and the GUI response was not so bad. When I did the same with the empty wav that you provided it played more than 40 tracks without going into the red zone. Thus the wav does matter. Now the specs:

Windows XP SP1 untweaked
M-Audio 2496 PCI @ 44100- 256 samples latency
Tracktion 1.6 (will try the same with the demo of T2)
CPU P4 Prescott @3.0GHz (overclocked to 3.6GHz = 15 x 240 FSB)
RAM Samsung 1 Giga DDR 3200 (1:1 @480MHz, timings by SPD)
Mobo ASUS P4P800 E Deluxe - chipset intel 865
Tracktion audio-data folder on Maxtor SATA 10 - 200 GB

That's it for now (not bad me thinks )

PS @polaris: we should standardize the wav and maybe we could have an additional test focusing on some of the included VSTi's (e.g. mda DX10 or JX10)

[FoxV]

My sense (given my ignorance of all this) is that when actual sound is present, the plugs struggle more mightily to process the audio. When the audio is silent, the plugs struggle less.

...that's my simpleton explanation.

Let's standardize the wave. Post one somewhere!

[doffenheim]

An easy option might be to render 60 secs of DX10 playing a sound (horn?) at C3.
It makes a nice uniform wave and everybody has it.
Also, wouldn't latency setting also have to be standardised?

[larsby]

why not take a sample from one of the tutorial/sample projects, then we'd know its the same.

[polaris20]

why not take a sample from one of the tutorial/sample projects, then we'd know its the same.

I am on board with this idea. Which one would be a good one?

[zeoy]

It doesn't matter as long as it's the same for all. Choose one with a reasonable size.
If you want you can create an archive (export->create an archive of this edit) of lowest quality compression (aprox. 16:1). This would make things easier.

[bear]

How about "Announcements" from Gare du Nord? This is a 12 second clip with lots of dynamics.


On my system I can do 12 tracks with screen zoomed out all the way so no scrolling necessary, 13 hits red. Hyperthreading is turned off. With windows task manager, CPU is hovering between 90 and 95% with 12 tracks


ABIT IS7 Mobo with 2.4ghz P4, I think this is i865 chipset, hyperthreading off
800 FSB
1 G RAM DDR400
Nvidia 128m dual head video
Maudio Firewire410 with only outs 1+2 enabled
Maxtor 80g 7200 drive EIDE audio and seperate 80G for system
WinXP pro with tweaks for audio as discussed here and elsewhere

Interestingly, turning 64bit mode on and off has very little effect....

[polaris20]

How about "Announcements" from Gare du Nord? This is a 12 second clip with lots of dynamics.


On my system I can do 12 tracks with screen zoomed out all the way so no scrolling necessary, 13 hits red. Hyperthreading is turned off. With windows task manager, CPU is hovering between 90 and 95% with 12 tracks


ABIT IS7 Mobo with 2.4ghz P4, I think this is i865 chipset, hyperthreading off
800 FSB
1 G RAM DDR400
Nvidia 128m dual head video
Maudio Firewire410 with only outs 1+2 enabled
Maxtor 80g 7200 drive EIDE audio and seperate 80G for system
WinXP pro with tweaks for audio as discussed here and elsewhere

Interestingly, turning 64bit mode on and off has very little effect....

Announcements works. I'll do it later when I get home from work.

[FoxV]

Reconducted the test using Announcements.

Got same results:

8 tracks (with recording) with no redlining. (Changing latency did matter, I used a latency of 256 samples; a latency of 2048 allowed me to add 1 more track.)

Using (again) this system:

Celeron 1.2 GHz
640MB Ram (don't know the kind)
80gig WesternDigital EIDE HD for system,
80gig Hitachi/IBM EIDE audio
VIA Apollo Pro133T chipset
Audiophile 2496 soundcard

[polaris20]

With "Announcements" clip, I got 12 tracks before red cpu.

This is with the A64 3000+ at work, running 512MB DDR, 80gig SATA. ATI Xpress 200 chipset, XP Pro (regular, not x64).

Anything above that and I got red, at 16 tracks I got a CPU overload warning pop up.

[will.record.for.food]

GEEZ!!

Its killing me not to have my PC running. I want so badly to contribute to your study... *eyes tear up*

[polaris20]

As much as I love Tracktion, 12 tracks is pretty bad for a state of the art CPU like the A64.

Why does it get so little? It just doesn't sound right.

[ModuLR]

Sony Vaio Laptop - Pentium-M 1.7, 1G ram, echo indigo @ 256 sampels... using the "anouncements" clip. I could get 13 tracks without it dipping into the red. I could run 15 tracks in the red with no glitches in audio.

[dealwithit]

To be truly subjective, wouldn't a standardized archive be the way to go? This seems to be very haphazard as it stands. I'd love to contribute, but I don't think that the results are very useful without something standardized.
Also, wouldn't it be better to use non-Tracktion freeware VST's to allow people to compare across sequencers? I have Cubase and would love to try it out against T with a standard test (in fact, I think I will).

[mandolarian]

Oh, come on, dealwithit, don't get all logical and methodical on us.

We're just here to prove that no matter how fast or slow your hardware, 12 tracks is all you need to produce some really phat barbershop tracks. That and the bowtie and stripes colour scheme.