Stereo Tool

As DJ-DOGGY said - Hans we would be very grateful if you gave us some guidelines for the performance tests. In order to do fair comparison I think there is a need for some standarization. We have to take into consideration, that general difference in performance between various generations of CPUs is huge. Hence we can't do all the stuff on slower CPUs due to their poor overall performance. As an example, I was not able to run Bojcha's King's Fire on Athlon II x4 (4096 samples), due to severe hiccups (buffer loss), yet I could run several other presets with similar options enabled effortlessly.
For this reason some "UNIVERSAL TEST PRESET" or even more than one, would come in handy.

I will try to do the tests today on my laptop (i5-3210m), HTPC (i3-3220T), PC1 (Athlon II x4 620) and PC2 (Pentium E2220). The OS will be Win7Pro x64 SP1. Can try it also on 1st gen mobile i3.
Bojcha, could you please give us details how exactly you measure processing time of the file using ST (and is it just ST or in combination with something else)? Unfortunately I'm not familiar with this method.

Regards,
Slawomir

As i understand Bojcha right , he use the DSP in some software like Adobe audition and with some preset in the ST he process a track while measure the time of processing in seconds. May be this is not the best way to test , because again in the different CPUs , there will be different preformance. What we need here is to consider the different classes CPU power and to make some different optimizations for them ( may be Athlon vs Intel , may be let`s say K9 vs K10 , or the analog families in the Intel CPUs ) . I think this is what Hans wants to do. To make some clasification , so he can make the optimizations for different classes.
As i see for now the Athlons are not that "fickle" to the compilers or libraries as reported by the user with the 9550 !?

The most effect in the difference of optimizations will be available with a slower CPUs . there will be the most visual effect , because in the new ones the power is so big and the effect will be as little as possible, am i right for this ?

I have some old Sempron 3200+ at 1.8Ghz real clock and version 6 (as i remember 6.20) was running with around 85% CPu load on it . Now with almost the same settings , this computer is totally "block" when i start ST

Yep I was wrong

Phenom 9550 (runing on Winamp): 32%(30%) | 31%(27%) | 30%(25%)

The diference is small but still is a diference

_________________
Very proud user of Stereo Tool since 2010. Thanks Hans.

My radio station
http://amradio.ddns.net

Simple..
- Prepare all 3 DLLs in winamp's plugin folder. Rename dll's as stereotooldsp1.dll.. stereotooldsp2.dll.. etc, just to know what is what.
- Set winamp output to DiskWriter.
- load one by one DSP and process same track with each.
- measure time.

I did, what Bojcha said:

6:11 track 192 kbps 44.1 mp3 stereo into 16 bit PCM 44.1 wav:

Loaded preset with default settings (Bojcha B-Town 4/Bojcha King's Fire - both FM), minutes:

ST1: 1:59/1:49
ST2: 1:46/1:41
ST3: 1:42/1:34

1 = Old/Old, 2 = Old/New, 3 = New/New

This is for i5 3210m (Instruction set: MMX, SSE1, 2, 3, 3s, 4.1, 4.2, AVX, AES, EM64T, VT-x) running Win7 Home Premium x64 SP1

I will update this shortly for other CPUs that I have

EDIT: Athlon II x4 620 (Instruction set: MMX, 3DNOW!, SSE1, 2, 3, 4a, AMD-V, x86-x64) running Win7 Pro x64 SP1

ST1: 3:20/3:03
ST2: 3:23/3:06
ST3: 3:13/2:59

EDIT: E2220 (Instruction set: MMX, SSE1, 2, 3, 3s EM64T) running Win7 Pro x64 SP1

ST1: 3:45/3:25
ST2: 3:44/3:20
ST3: 3:40/3:16

EDIT: i3-370m (Instruction set: MMX, SSE1, 2, 3, 3s, 4.1, 4.2, EM64T, VT-x) running Win7 Home Premium x64 SP1

ST1: 3:02/2:49
ST2: 2:59/2:50
ST3: 3:00/2:35

I got similar results (only slightly different) twice for this one, which points out, that preset might affect the readings.

I didn't test i3-3220T as it is basically the same chip as i5-3210m (dual core, hyperthreaded, has SSE4.1, 4.2, AVX, the only difference is that it doesn't have turbo and AES support).

@Slawomir: 1 = Old/Old, 2 = Old/New, 3 = New/New? In that case your measurements resemble mine (but you have a 3rd gen i5 and I use a 2nd and 4th gen i7, so that makes sense).
Edit: Ow... The Athlon gets slower due to the new libraries, but faster again due to the new compiler. Interesting. (This differs from Mpex2006km's Athlon 3500 measurement). This again might mean that new compiler + old libraries would work best. For AMD. Not sure if it's worth the extra effort though (compiling and worse, getting people to install the correct versions).

@Mpex2006km: Your AMD results seem to resemble those of the latest Intel i-series. But the results are very different this time from your previous measurement...

Which leaves the older Intel's. Although there's not THAT much difference between Bojcha's measurements (2% or so). It would be nice if I could figure out where that difference comes from though, based on Bojcha's numbers it seems to be the new compiler (not the new libraries).

I also have an old E5200 here somewhere (I used it at the IBC and it's still stashed away in a box), will try to find it tomorrow and try what happens on that system.

Yes, correct. Currently I'm testin E2220. Will report afterwards. I will also test 1st gen i3-370m, which has SSE 4.1 and 4.2, but no AVX support.

EDIT: Testing done, results in my previous post.

Wait a minute ,
Like Bojcha said , i`m getting a real time processing . Time is running as normal .
What settings have this Disk writer ? . Some screen shot ? I click the song and it starts as normal (yeah it writes a WAV at the selected folder) , which means that the time will be the time of the song . May be i`m doing something wrong ?

I mean that i saw Slawomir is processing 6:11 minutes song for 3:22 minutes ? So it must be some way faster than the real time running , right ? On my winamp the time of the song is as normal ? ST load for the "B-Town 4 HVz mod" is as normal too.

I don't know why the results very that much but i think it goes faster. Not so easy shattering and timeouts....

(BTW I had that problem in the past. I could run ST FINE restart the system and have problems and then restart again and it was fine strange..... But it is an old install of windows 7. I am getting ready for moving out of 7 and run a clean install of 8.1 on a ssd and do more testing of corse)

_________________
Very proud user of Stereo Tool since 2010. Thanks Hans.

My radio station
http://amradio.ddns.net

Just like that:

https://drive.google.com/file/d/0B1oKe7 ... sp=sharing

And yes, it was faster proportionally more/less to CPU speed.