Stereo Tool

Mee to agree that ST is Intel optimized because Hans have only Intels and work, write it , optimize it and compile it on Intel.It`s the most clear thing here. We all know that Hans works only with Intel Cpus. I`m not saying that this is worse , just show you some "compare" That`s it.

If you really want to know the differences in speed between Cpu's,
Then read this: http://www.agner.org/optimize/instruction_tables.pdf

Ok, here are 3 versions for comparison. Let me know the results!

OLD COMPILER, OLD LIBRARY: http://www.stereotool.com/download/dsp_ ... 51-014.exe
OLD COMPILER, NEW LIBRARY: http://www.stereotool.com/download/dsp_ ... NEWLIB.exe
NEW COMPILER, NEW LIBRARY: http://www.stereotool.com/download/dsp_stereo_tool.dll (not an installer)

Note: The 1st and 3rd were posted before in the thread, only the 2nd is new.

If you compare these, please tell me the results, and which OS and CPU you have.

I tested ST on my phenom 9550 (4 cores,2.5Ghz overclocked), my athlon 3500 (single core,2.2Ghz) and my intel pentium D (single Core, Hyperthreading,3.4Ghz) and the result are very strange(tested with stereo tool in its last stable version ( 7.50))... The Pentium D runs ST as fast as Athlon 3500 if not faster. The passmark benchmark shows that the amd should be faster but that was not the case.

Now if i try to run st on my phenom i can't get more than 65-70% per core (even when i set affinity to 1 core only) withou sound shattering. So i have three PCs that have several years and computing power difference but no one can really run ST even with quality set to 20%.

What i mean is that ST is clearly optimized for intel and amd proccessors can't run it as well as evn older intek ones. I really compare my Pentium D and my Phenom when running ST because they are very close performance wise.

Thats what I see from my testing so far.... And i am really thinking about buing a new Intel Pc just for running ST.

Thank You All . (I am sorry for 1)my english and 2)maybe i am off topic)

EDIT: I will test the three new ones and post my results

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

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

The only thing that i dont understand is why we test with the DSP plugin ? Is there any reason or it`s picked as one of the versions ?
I think we can "see" any difference best with the Stand alone as it don`t use any additional software ( like winamp here ) . It`s just a self alone program. That is what i think.
Nevermind I`ll see what others will report , because my CPU is noted with up to 70% better preformance in the multimedia apps ( what ever this means ) . May be that is why i have half load compared with my previous CPU. It looks like i`ll not be the best reporter for the Load . Will see others what will say.

And guys ! There is one important thing . i think that the tests must be with some preset and only that preset on different CPUs . After that, if you want, make your tests with your own preset . But results must be on equal parameters, am i right ?

Testing Done...

TO BE CLEAR : The test was to see the diference from the the "older" version to the newer. Each CPU was tested with each own preset (three diferent preset). These are PEAK cpu values and these (are average)


OLD,OLD NEW,OLD NEW,NEW
Athlon 3500 (runing on RadioDj): 58%(52%) | 56%(52%) | 55%(50%)
Phenom 9550 (runing on Winamp): 32%(30%) | 39%(33%) | 34%(29%)
Pentium D (runing on Winamp): 56%(54%) |59%(53%) |56%(54%)

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

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

Processed track with 3 versions on my Core2Duo E8400 CPU gives this times.

1. 72.4 sec
2. 72.6 sec
3. 73.4 sec

Hm, your results are opposite of each other. In all cases the difference is small. But in Bojcha's measurement the new compiler's code is slower, and in mpex' measurments it's equal or faster. More specifically, on the Athlon and Phenom (both AMD) the new compiler + new library is faster than old compiler + old library, while on the Pentium D it's equal and on the E8400 (both Intel) it's slower.

Mpex: Your Phenom 9550 measurement seems odd, the average for old/new is higher while the final result (new/new) is lower again... Could you repeat that test to make sure? If this is indeed correct the combination new compiler/old libraries might be interesting to test. How long did you measure to get these average values?? Because the difference is quite small so it's easy to draw the wrong conclusions.

Bojcha: Your differences are also very very small... have you done each measurement at least twice to rule out some hiccups on the system, drive, ...?

Would like to get more measurements from other people. And Bojcha's way of measuring (processing a file and looking at how long it takes) is generally far more reliable than looking at TaskManager. That's one of the reasons to use Winamp for this test...


Btw - the good news is that the new Intel compiler is finally not or not much slower anymore than the old one. When I tried to move from 10.1 to 11.0, I had a performance degradation of 25%. When I tried 13.1, it was 6%. Now, it's apparently very close to 0% (and on newer hardware it runs faster). On my laptop and my new 4770S test system, I see a considerable drop in CPU load between Old/Old, Old/New and New/New (where New/New is about 20% more efficient than Old/Old - the biggest drop comes from the new libraries though).

I disabled all things around that can use CPU, even turned off network.

With DSP that is best way to test. Thing is that i needed to use stopwatch to measure time, wich is odd.
It would be easier with VST. I can put it in dbPoweramp and measure exact time. For some reason dbPoweramp does not recognize ST DSP (shows blank settings).

Some versions ago i said that plugins are load the CPU with about 10% more. .It was the measure with the VST and Stand alone . After some optimizations this difference was lowered . That is why i tell you that the best way is to measure the "speed" of the Stand alone. I don`t know in which way you test it . But in the beggining of this change of libraries and compilers you said there is some "faster preformance" . How you measure it ? In the Task manager or some special utility . Or may be the compiler is measuring it ?
Hans this is a question to you .