kX ASIO Performance: Attention Lex N.

Toadly_Ossum · May 17, 2006, 03:31 PM

Im running P4 3Ghz with HT. Hyper Threading needs to be disabled for use with 3534 versions. Disabling HT is done in the BIOS setup.

Toad

Russ · Aug 27, 2006, 05:14 PM

Have any of you (that went back to 3534f) tried 3538k?
I have definitely noticed better ASIO performance with 3538k, than with previous 3538 versions, and I wonder how it compares to 3534f.

(I did not think Eugene changed anything in 3538k as far as ASIO goes, yet I do get better performance...)

Doug W · Aug 28, 2006, 08:13 PM

Russ,

I have not tried it, but I fear change! Have you experienced any crackling on playback?

Doug

**dj_stick** · Aug 28, 2006, 08:44 PM

Quote:

Originally Posted by Russ

Have any of you (that went back to 3534f) tried 3538k?
I have definitely noticed better ASIO performance with 3538k, than with previous 3538 versions, and I wonder how it compares to 3534f.

(I did not think Eugene changed anything in 3538k as far as ASIO goes, yet I do get better performance...)

it could be one of the many minor changes that eugene makes to the code, and forgets to add to the changelog

Russ · Aug 28, 2006, 09:16 PM

@Doug W,
Honestly, I do not use it like some of you guys do (and my system is not the best for ASIO performance (maybe the VIA chipset)), so maybe I am not the best person to judge. But, I had mentioned previously, that I absolutely could not use the 2ms setting at all, with previous 3538 versions (the sound would sorta of freeze up (instantly), hard to explain), but with 3538k (as with 3534f (which I only did minimal testing with)), I am able to use a 2ms setting (and even lower for just playback (not sure if ever ever tested just playback with 3534f)). You may want to wait and see what other people say about it (and maybe until the new ProFX are available), but it is definitely better for me (I still have some issues, but I did with 3534f too). BTW: Most of my testing is on the recording side of things (i.e. actual recording, or using VST effects (i.e. using both the ASIO inputs and outputs)).

@dj_stick
That is possible, but he did specifically say: "asio code is the same. I will probably check 3534f-3538 difference soon", so I cannot explain it, other than maybe something else (unrelated) had an effect on it (or maybe just some code optimizations, etc.). In any case, it would be good to hear from other people, to see if it is better for them as well.

Toadly_Ossum · Aug 28, 2006, 10:29 PM

I switched to 3534f some months ago specifically because of crackling/poping in the recorded audio tracks and VST instrument tracks. I banged my head on this problem for some time before seeing this thread. 3534f solved my problems completely, and produced profound feelings of euphoria and well being. With Pro FX for 3534f I have my system working great. I do however dislike disabling Hyper Threading (not sure if it even matters for VST performance) Maybe when I finish with the Blues recordings Im working on, Ill try the newest.

Toad

Russ · Aug 28, 2006, 11:40 PM

Yeah, I have a feeling that most people who went back to 3534f, also depend on being able to use ProFX, so we may have to wait until it is available for 3538k, before people will be willing to give it a try, and let us know how it compares.

Lex Nahumury · Aug 29, 2006, 02:54 AM

I haven't tried '38k' yet, but if it turns out to have the same ASIO performance as 34f(which I doubt) that would benefit Profx3.x users.

/Lex

**Eugene Gavrilov** · Sep 7, 2006, 05:09 PM

can anyone identify when exactly has the performance become worse?
was it 3534f - 3535, or 3534f - 3535/36/37?..
--
3534f and 3535 asio code is identical, however, kernel-level driver uses different technique to be hyperthreading-compatible...

E.

Doug W · Sep 7, 2006, 08:21 PM

I was not aware of problems when I had 3535 or 3536 installed but I was still a bit confused by kX at that point and not doing much recording. I noticed playback problems, (pops and clicks), with maybe 8 tracks when using 3537. At first I attributed the problems to an aging computer but I switced to 3534f after reading this thread and the problems went away.

Personally, I am happy with 3534f but I am sure there are others with newer cards than mine who can't use it. I don't do anything very tricky as far as recording so I have no complaints unless you want to hear about my back. Hope you can figure it all out.

Thanks,
Doug

Lex Nahumury · Sep 8, 2006, 01:13 AM

Quote:

Originally Posted by Eugene Gavrilov

can anyone identify when exactly has the performance become worse?
was it 3534f - 3535, or 3534f - 3535/36/37?..

E.

As mentioned/discussed before;
Every release >34f, so since 35.

/Lex.

Toadly_Ossum · Sep 8, 2006, 10:02 AM

Quote:

Originally Posted by Lex Nahumury

As mentioned/discussed before;
Every release >34f, so since 35.

/Lex.

I concur though be it by memories from over a year

**Eugene Gavrilov** · Sep 11, 2006, 10:50 AM

> Every release >34f, so since 35.

I have a special build with a lot of additional sync options (a2-specfic, timer, hw interrupts etc.), I doubt it is a good idea to provide a list (more than 20 entries) of sync options for end-users... anyone willing to test this? (SMP, uniprocessor, HT, Athlon, Intel, XP/2k etc.) this way we could probably limit the number of options

E.

Russ · Sep 13, 2006, 05:52 PM

How about leaving all the options in, but making some of them only available via advanced methods (using kxctrl, or maybe a registry key modification, or something (i.e. maybe a kxctrl command, that when enabled, the ASIO Control panel shows all of the options, otherwise it shows the smaller list)), this way people who want to test them can, and those who don't, will not see them.

Just a thought...

In any case, I would be interested in testing what I can (my mobo/processor does not support a lot of the newer stuff).

-Russ

Doug W · Sep 15, 2006, 08:13 PM

I am willing to do some testing even if my computer is older than Russ's.

Russ · Sep 18, 2006, 04:49 PM

I thought that I would add that I did some preliminary testing with 3538L, and it seems (to me) to offer the best performance thus far (even exceeding the performance that I got when testing out 3534f).

I decided to post this here, and not in the 3538l ASIO performance thread, because it is difficult to judge how the different sync options effect the performance (which one works best seems to depend on what latency setting I choose and the CPU utilization level (and only 4 of the 11 sync options apply to my system)). In any case, I need to do more testing, but so far, so good.

JGSF · Sep 20, 2006, 05:31 AM

Yep! 3538l is better than previous releases for (I never tested 3534f, so I'm just comparing with the previous 3538 releases)!

Maddogg6 · Sep 20, 2006, 10:04 AM

Quote:

Originally Posted by Russ

I thought that I would add that I did some preliminary testing with 3538L, and it seems (to me) to offer the best performance thus far (even exceeding the performance that I got when testing out 3534f).

I decided to post this here, and not in the 3538l ASIO performance thread, because it is difficult to judge how the different sync options effect the performance (which one works best seems to depend on what latency setting I choose and the CPU utilization level (and only 4 of the 11 sync options apply to my system)). In any case, I need to do more testing, but so far, so good.

Im glad I found this - I too find modes 'inconsistant' - but I couldnt place my finger on how it was changing...
But yes - I see improvement with 3538L too.
When I find mode(s) that work best for me - I will report it in the proper thread.

nakamichi · Sep 20, 2006, 02:11 PM

If you have a 10k2 card,try the (10k2) modes.
With the last mode (for single-core CPUs).I get the same *great* performance as I had before with 'Thread mode',but without the 100% CPU usage.
Try Irq/SMP (10k2) with a Dual core CPU.