Performance Issues with AMD64 3000+, 1.5GB RAM, FreeBSD 5.4-RELEASE

[Mark Kane]

Kris Kennaway wrote:
> On Thu, Aug 25, 2005 at 12:48:17PM -0500, Mark Kane wrote:
> 
>>Roland Smith wrote:
>>
>>>Another thing to look at might be the scheduler. I'm using SCHED_4BSD.
>>
>>Hmm, I'm using just a GENERIC kernel with support added in for my sound 
>>driver and atapicam for K3b. SCHED_4BSD looks default in GENERIC:
>>
>>options         SCHED_4BSD              # 4BSD scheduler
>>
>>I've been trying this morning with no real disk I/O and just XMMS open. 
>>It's much better than if I did have a download/upload going using disk 
>>I/O or trying to read from a file on the drive, but still nowhere near 
>>perfect. Every minute I'll still hear a little skip or stutter in the 
>>audio, and still notice the mouse locking up a bit at the exact moment 
>>that happens.
> 
> 
> Did you not read my mail?  I explained this and the solution.
> 
> Kris

Yes, I read your email and responded to it yesterday. I said:

-------------------------
I found the sound skipping thread by Jeff. Looks like a great 
explanation, however my problem even occurs even with light disk I/O 
such as reading an mp3/Ogg/video file off any of the drives. I did some 
more testing with actual files off the drives last night instead of 
streams, and it seems worse with actual files.

I'm not sure if the proposed patch will fix it or not, and trying 
6.0-BETA2 for my other hardware/hard drive problem just introduced new 
problems (panics, network stuff not recognized, etc). I do have a spare 
hard drive where I could put test anything though if you think trying 
-CURRENT or another BETA would be worthwhile.
-------------------------

So, in other words, even when there is NOT heavy disk I/O it still has 
problems, although they do get worse with more disk I/O. I don't know if 
you saw my previous replies, but trying 5.4-STABLE on my Athlon XP 2000+ 
instead of 5.4-RELEASE seemed to make it much better on that machine 
(however still not perfect).

I was wondering if the code that fixed it in 6 was merged to 5-STABLE or 
if the patch was already committed to 5-STABLE? I haven't tried -STABLE 
on my amd64 yet though.

-Mark