@A3000T

Quote:

When any player is active and the network is started, the sound starts to crackle a bit. If the network is started first and the player second, this doesn't happen.

I experienced something similar. I have two little scripts to start and stop the network. Yesterday I was playing MP3s with mplayer and then I started the network and suddently I heard a short beep. Then the playback continued regularly so I leave my A1 downloading 4 ISOs to do the testing for the problem we are talking about in this thread.
After more than 1 and a half hour I decided it was enough and I shut down the network: mplayer suddenly stopped (I think it exited, but I'm not sure because I started it from the GUI; I received no GR and the system kept on going on fine).

Just a little detail; I don't know if it could be useful.

Ciao
Gabriele

_________________
--
In a world without walls and fences
we won't need windows and gates

@DaveAE


13 minutes (it appears that crash ocurred 3 minutes after I posted the prev. time and went home).

3 simultaneous loops of d/l with wget, local server, 4.5 MBytes/s of traffic in total).
Jack


Edit (an important one!):

Got hands on sb vibra128 and the ahi drivers form os4depot (thanks Ross!) and guess what?
The sblive is sitting on the desk, the vibra is rocking while downloading like mad, already the 4-th
cycle of 3 simultaneous d/l. It is already 5-th one in fact and makes it >1200MBytes of d/l at 4.5 MBytes per second.

The sound didn's skip a bit (amigaamp). Will swith to tunenet now (it used to suffer from skipping sound much more than amigaamp when both were tried with sblive).


Edit #2:
tunenet still skips, back to amigaamp, 9-th cycle of d/l!!!
In the past 1 cycle was enough to provoke the problem!!!
10th cycle is on the way...(2.7 Gb so far)

Jack

Last edited by jack on 12-Jul-2005 at 09:23 AM.
Last edited by jack on 12-Jul-2005 at 08:25 AM.
Last edited by jack on 12-Jul-2005 at 08:20 AM.

_________________

"the expression, 'atonal music,' is most unfortunate--it is on a par with calling flying 'the art of not falling,' or swimming 'the art of not drowning.'. A. Schoenberg

@all
No i don't have the chance of check articia temperature. Anyway with this big mod fan, my motherboard is literally cold

The air blow on cpu and near, so also on the articia.

I really don't know what to think.

To check if the problem is articia over heating, take a..... i don't know the right word in english.... a big fan which you use to "cool" yourself and your room.... like this:
http://www.uffa.it/uploads/ventilatore.jpggKAvfs.jpg
And put it in front of your mb (open the case lol). if you don't get the lockups then, maybe the problem is really a chip which gets too hot in some configurations.

_________________
ikir

@Chris_Y

Quote:

So, the MIDI is completely disabled under normal circumstances anyway?

Yes, but it seems clear now that the emu10kx is still the culprit, because people who are changing cards have no more problems. I'll have to look at those 3 PCI values you posted, but I doubt they will solve it. Anyway, I'll have to reproduce it myself here too first. Perhaps I'll have time on Friday but there are other things I have to do as well. Note that it doesn't matter if you set these values in U-boot or at driver init time. Are the 3 PCI values for the VIA686B or SB Live?

_________________
Audio Evolution
http://www.audio-evolution.com

@Yogi27

Hi, Yogi27.

I've just read your reply, regarding your experiences with the A1XE. Your answer actaully underlines my arguements towards the CPU/Chipset heat problem. Let me try to explain.
When running an internet server, much off the taskload resides on the network card and chipset. The workhorses of the system are doing the work, and the CPU has fairly supervisory task job, with a low CPU overhead. You wil not run into problems running a system like that. You can compare it to having a Ferrari, where you never run into mechanical problems, when you drive within the legal speed limits. But when you arrive at the Autobahn(hooraay!), you find to your big surprise, that you Ferrari "only" has a measly 250Kph topspeed, when you where promise a 330Kph Ferrari. And this is my arguement, why it could be badly fitted CPU cooler or another heat problem.
You need to go to the "Autobahn", and performance test your A1, just like Wegster did. See his massive task list. It involves heavy use of the NIC CPU, IDE CPU, RAM fetches, heavy DMA usage, CPU intense realtime tasks, chipsets(Articia, bridges, PCI... Etc.) are crosstransfering a lot of data. Now you are running flatout on the Autobahn, and any problems will show up under these extreme test conditions. It's uner thes conditions that Wegster has run into problems with his A1XE, and it seems very connected to my CPU/chipset heat problems.

If you take a look at a "modern" soundcard, you'll find that most of them are fairly "dumb", when compared to modern network, 3D Gfx, IDE controller card. The soundcards rely on heavy usage of the CPU to perform realtime tasks.
Other modern hardware components has usually enough CPU power onboard, so they "get by themself". The usage of the main cpu is low, when compare to a "dumb" soundcard. The soundcard is basicly a Digital-to-Analogue-Converter(DAC), with a "hardware mixer" and PCI(IRQ, INT, adress) and DMA stuff. But if you take an Audigy type soundcard, you'll find them more intelligent with many times more onboard CPU/DSP power. A SB128 type doesn't help the main CPU that much, resulting in intense CPU/chipset usage. An Audigy type would decode on MP3 onboard at very high quality(if drivers permit), with very low main CPU usage. A SB128 makes heavy use of DMA and the main CPU.
Even a modern xxxxMHzCPU will be affected by a resource eating SB128, when running many other realtime tasks. But on a stable well cooled system, things runs just fine, except the DVD-player sometimes skips a few frames. But your system doesn't freeze or stop working like Wegster's.

If the CPU core gets "hot spots/thermic bubbles" in the ArticSilver layer, due to a slightly misfitted CPU cooler. Then something like this could happen...

The cache, pipeline, and inner core areas of has proper cooling, but the FPU part hasn't. When you start your MP3 player, the player makes heavy usage the FPU. The temperature DIfFFERENCE is simply too high between the FPU and the other parts of the CPU core. Too much heat in a CPU core has some nasty effects. Noise levels starts to rise, the electron flow slows, signal levels drops, and a lot of other quantum related mindblowing problems. The system works fine in the beginning, but in time all the quantum nasty parameters will keep rising! Slowly, but surely until the system freezes due a a small timing error in the chipset/CPU/memory system.
Some bit errors can be detected and corrected, in HD's, RAM, net, modem and a lot of other types of transmissions. Not all lowlevel chipset timing signals are error corrected. Designers lay out a "safe area of operation", and then up next set of designers to keep things within specifications and limitations. If some computer component gets outside it's "safe area of operation", just about anything can happen to these CRITICAL lowlevel timing signals. When this situation happens, the system goes into "undefined area of operation" I.E. the system will freeze sooner or later. With no watchdog to reset the system, pressing the master reset button is the only solution.
Anybody who has tried to change the BIOS chip timing settings in a wintelbox, has properly to use too aggresive chip timing settings, resulting ín a nonbooting PeeCee. Reset BIOS to default, and the system should boot again(Yeahh, right Bill).

To Wegster I just wanna say please get yourCPU Cooler checked. It might involve your local PeeCee dealer(Pffffttzz), getting a pro($) to refit the cooler. Get him to fit a northbridge cooler on Articia, to keep the motherboard signal quality as high as possible. These factors keeps my A1XE running along rockstable.
So please get your local Ferrari mechanic to check that "250Kph Ferrari engine". The mechanic thinks the Ferrari has the latest MaxEngineTempThres firmware, that activates the LowerEngineTemp software. He can also see from the log, that FANwaterCoolerRPM parameters is 0. You need the get a new cooling fan, then your monster machine will do 330Kph easily .. and much more

Will return tomorrow

!8O)
/Cipher

@CiPheR

Quote:

The soundcards rely on heavy usage of the CPU to perform realtime tasks. A SB128 makes heavy use of DMA and the main CPU.

Oh, come on. The PCI bus can handle 133MB/sec whereas the soundcards use 176kB/sec. And on the A1, the emu10kx uses the same amount of CPU and PCI power than any other card.

_________________
Audio Evolution
http://www.audio-evolution.com

The VIA686A/686B vs. SBLive bug is known since a long time. One of the last PC's i owned, around 1999-2000 had an AMD K6-2/450 along with a VIA MVP3 chipset mainboard, which also had a VIA686B southbridge. That one also had the SB Live crash/lockup bug. In fact, both the EMU10K and the VIA southbridge fails to be 100% conform to the PCI standard, and these tiny differences makes the problem appear. There were some partial software workarounds for it, but neither of them was a complete solution. The best solution was to replace the soundcard and/or the mainboard.

For me it wasn't a problem since i had no SB Live, but a Gravis Ultrasound PnP 8MB, which i used under Linux and OS/2 Warp. But once i tested an SB Live in that mainboard, and the whole system went nuts, just similar what you're talking about. Remember, this was on a PC, cheap & widespread hardware by that time. IIRC, the problem was fixed on newer VIA southbridges, and maybe also on newer revisions of EMU10K. Man, talking about these old stuff really takes me back...

_________________
MorphOS, classic Amiga, demoscene, and stuff
"When a bridge is not enough, build a Viaduct!"
"Strip the Amiga community of speculation and we can fit every forum on a 720k floppy" (by resle)

@A3000T

Quote:

When any player is active and the network is started, the sound starts to crackle a bit. If the network is started first and the player second, this doesn't happen.

Yes, that can happen. I just looked at the eth3com sources. When it starts, it uses Forbid() and then Wait()'s for a signal before Permit(), which can lead to jerky playback if something was already playing.

_________________
Audio Evolution
http://www.audio-evolution.com

@wegster

In the via686b docs, these are actually the defaults:
PCI Delay Transaction: 0 or off
PCI Latency: any value from 0 to 32

and this register is not in the docs:
PCI Master Read Caching: 0 or off

Not surprising since:
"This is an AMD-specific BIOS feature. It determines if the processor's L2 cache will be used to cache PCI bus master reads."

I can fiddle a bit with the latencies to see if it solves anything, like maximizing the max latency for PCI grant on the emu10kx and setting it to zero on the VIA and see what happens.

_________________
Audio Evolution
http://www.audio-evolution.com

@wegster

AmigaOne XE-G4 (7445 @ 800 Mhz) (unfixed)
512 Mb Kingston registered SDRAM (memtester - ok)
Soundblaster Live! 5.1

When running The Test, the system is stable from anywhere between 10 minutes and 2 hours. I can't tell you how many times this has happened to me: as one of the original posters, it probably numbers in the hundreds . When running The Test without passing anything to audio my system remains stable throughout.

My V-Core is set to 1.84, and I have replaced the default fan with a Swiftech MCX159-A AMD Chipset Heatsink and Fan. The Lian-Li case my AmigaOne calls home has two fans in the front sucking in air, one in the back and one on the top blowing air out, and a PCI double-fan to bring in more cool air.

@DaveAE
I ran your test together with MPlayer, downloading ISOs and DNET - total uptime: 7 minutes.

_________________
After a decade away from the scene, I am back!

@jack

Quote:

tunenet still skips, back to amigaamp

This is a buffer issue in TuneNet and won't be a problem in the next TuneNet update. It isn't related to the sound bugs you are, or were getting.

Cheers,

Bean.

_________________
OS4.1 + SAM Flex
RIP my A1XE.. that used to have an appetite for batteries!

@CiPheR
Running Dnetc alone creates more heat than playing an mp3.
Just streaming internet radio alone bearly increases the temp, even with a massive network load.

@DaveAE

Just tried running your audio test file while downloading 3 large (600MB each) files. Only took 6 minutes before the computer froze and junked the audio.

_________________


AmigaOne XE-G4, 512MB RAM (AmigaOS 4.1)
Radeon 9250, Terratec Aureon 5.1
CDRW, DVDRW, SII0680

@wegster

> Any chance you've got an SB Live somewhere you can swap in temporarily? If
> not, I can send you one, seriously, if you don't have or can't borrow one,
> and are willing to? (I've got a Terratec coming this week.)

I don't have a SoundBlaster Live myself, and most people near me who know what sound card they have hate SoundBlasters. The rest thinks "SoundBlaster" and "sound card" is the same thing.

I don't mind doing the test again with a soundblaster installed, but don't send me the SB live yet. Parcels from the United States can take a long time to arrive here. I'll ask around a bit more to see if someone near me has a soundblaster live.

Kind regards,

Dennis

@DaveAE


Quote:
Oh, come on. The PCI bus can handle 133MB/sec whereas the soundcards use 176kB/sec. And on the A1, the emu10kx uses the same amount of CPU and PCI power than any other card.

AFAIK PCI modem and soundcards cannot share IRQ properly with other devices and this is only limitation, what I know. Thats why normally in pc-world modem cards are specified to last PCI slot and soundcards second last slot. Last is bottom most in this case :)

@DaveAE

Quote:

Yes, that can happen. I just looked at the eth3com sources. When it starts, it uses Forbid() and then Wait()'s for a signal before Permit(), which can lead to jerky playback if something was already playing.

Oh, yuck! why on earth is it doing that? I'm sure there must be a more system-friendly way of doing whatever it's trying to do. Please, enter that as a bug in the OS4 bugserver, that sounds like a horrible hack to me.

@CiPheR

Welcome aboard and thx for the info, but I REALLY think we are beyond heat induced lockups for the moment.

The supplied fan and heatsink has been a subject for discussion since the A1's were first delivered and most people have either solved this to their own requirements or have little need to bother more for it's own sake.

Most people here have dedicated active northbridge coolers and fans and still have lockups while in my case a simple change from the SB live! to an Envy24HT chip based soundcard has solved all apparent problems.

@Thread

Have been playing mp3's two days in a row now from a 128mb USB thumbdrive (with DNetC running as usual) without the slightest problem. Haven't stress tested this setup in any other way yet as I haven't had time to do anything (really up to my bejeezes in work atm) but I'll throw in some ISO downloads now as well as the aiff file from DaveAE and see if things can't get any worse.

@wegster

If you want I can put the SB live!(yeeuch) back into my machine and do an instant replay if you think it's needed?



Will download ISO's now. Managed to get hold of 10, so that should work if anything does!



@mlehto

Quote:

AFAIK PCI modem and soundcards cannot share IRQ properly with other devices and this is only limitation, what I know. Thats why normally in pc-world modem cards are specified to last PCI slot and soundcards second last slot. Last is bottom most in this case :)

That sounds more like a hack than a solution?

Last edited by Steff on 12-Jul-2005 at 10:38 PM.
Last edited by Steff on 12-Jul-2005 at 10:32 PM.

_________________
Fixed A1G4XE 7455 RX933PC with fried CPU
Sapphire Radeon 9100 128mb
ESI Juli@ 24bit 192kHz Envy24HT
Sil 680 Ultra Ata 133 E-ide
SeaGate Barracuda 120gb 8mb cache

@CiPheR
I've been building systems for a _very_ long time. Thanks for the comments, and interesting note about the 'hot spots' on the CPU thermal pastes. It would be a _very_ cold day in hell before I'd let someone at a local shop touch my system though

Anyways, you're right about the relative lack of onboard power on older (and some modern?) sound cards, but note that CPU utilization was never > 25%, versus I can peg the CPU at 100% for 24x7 without problems, as long as I don't use sound. Again, that does NOT sound like a heat induced or cooling problem to me, at least.

_________________
Are we not done with the same silly arguments and flames yet??!

@DaveAE
Quote:

Yes, that can happen. I just looked at the eth3com sources. When it starts, it uses Forbid() and then Wait()'s for a signal before Permit(), which can lead to jerky playback if something was already playing.

So, knowing almost zero at this point on signals under AOS, but knowing how async io works on other OSes..is this an issue in the current code that can be corrected with the current API for OS4, or something that's going to be what it is until some underlying code in Wait() and friends are overhauled?

NOTE- the above question is NOT in relation to the lockup problem this thread is discussing, please don't confuse them

_________________
Are we not done with the same silly arguments and flames yet??!

@mlehto


Quote:

mlehto wrote: @DaveAE AFAIK PCI modem and soundcards cannot share IRQ properly with other devices and this is only limitation, what I know. Thats why normally in pc-world modem cards are specified to last PCI slot and soundcards second last slot. Last is bottom most in this case :)

Moving the card does not change the problem - tried it, no change.

_________________
Are we not done with the same silly arguments and flames yet??!