simple ASIO performance test: 3538j vs. 3538k - no difference

JGSF

Ok, here it goes:

- winxp pro SP2, no apic (only 16 IRQ)
- Audigy 2 with an IRQ all for itself.
- A64@2500, msi neo2 platinum (nforce3 ultra)... bla bla bla bla... The rest doesn't matter for this.
- Guitar rig 2 demo.
- All 16 ASIO channels enabled, using default mode.

So, for 3538J and 3538K I got:

- Minimum working latency (where I can still hear the sound and not a bunch of noise): 2ms, with ocasional pops.
- Minimum realiable latency: 2.67 ms, with pops starting to occur at around 50% cpu load.

Conclusion:

With ACPI off and with applications that make no significant use of the hardrive, ASIO performance remains the same.

If there's really an improvement as some people stated, maybe it's under diferent conditions, such has ACPI on, where in my case, I can't go lower than 4ms with 3538J. I still don't want to mess with my ACPI installation that I use to work since it will be a PITA to get all those kx plugins working again, so, if you're sure that 3538K works better with ACPI just let me know...

jiiteepee

Well, maybe most people writing 'bout those matters does not have any glue what's all 'bout --> don't trust what's written by individuals on these forums ... (not even developers more than 33% ...)

jiitee

Lex Nahumury

Why should there be any ASIO perf. diff.?
Eugene made it pretty clear in;
kx 3538k release
.. that asio code is the same.

Does that include yourself ??

JGSF

I'm not the one claiming there is... I was just taking my own conclusions.

jiiteepee

Now what, didn't you forgot quote the most essential part ...?

jiitee

Lex Nahumury

No. Or are you a 'developer' now too .

Russ

AFAIK, I am the only one that has made any claims that it was better, and I specifically said that it was better for "me", and it is better for "me", period. Since there is no apparent reason for the better performance I am experiencing, I asked how it was for other people, so I appreciate the testing done by the OP.

It is hard to judge the different performance claims made by different people, because everyone has different hardware, software, and testing procedures. For this reason, the only way to really judge, is to wait until many people have tried it, and go by the overall average, so please do not assume that because one person's results are different from anyone elses, that they have no clue (or "glue" as jiitee put it).

Because of the statement made by Eugene about the ASIO code in 3538k, I did not expect any differences, and was surprised when I tested it. It is pretty simple really, with previous 3538 versions, I could not use certain settings at all, while I am able to use these same settings in 3538k. I am pretty sure that most people would agree that constitutes better performance (for me), but I made no claims that it would be better for anyone else.

Eugene Gavrilov

well... after meditating on this, I think, there -might- be some performance change:
I changed the compiler (new Server 2003 DDK instead of WinXP DDK) - this might have caused the following:

- random fix of volume bsod issue (caused mainly by the compiler, since the BSOD said 'the driver has marked pages...' etc, this might be related to compiler internals)
- random performance change, since the new compiler might have used different code/data layout and padding etc.

E.

JGSF

I'll give it a go with ACPI when I have time.