Computer crashes _ALL_THE_TIME when encoding with XviD...

[MoritzT]

Hi,

as I have already mentioned in the subject, my problemm is that my computer crashes whenever I try to encode a file with XviD.
First of all, I'm going to give you my config, then some of my thoughts about this.

My Specs:

--------------------------------------------------------------------

AMD Athlon Thunderbird 1,4 GHz
512 MB Apacer DDR RAM
ECS K7VTA3 Mainboard
GeForce 2 GTS Pro 64MB
Enermax power supply

Windows 2000 SP3
NVidia Detonator 44.03
Via Hyperion 4in1 4.47

Koepi's XviD Build 14.05.03 (actually, my computer crashes using any Xvid Builds, going back to the stable build 04.10.02)

VirtualDub 1.5.4 (crashes with all Versions)
&
VirtualDubMod 1.5.1.1a (all versions)

AviSynth 2.52 (as well as 2.51, 2.08 .....)

--------------------------------------------------------------------

My XviD settings (2 pass):

Motion Precision: 6
Quant Type: MPEG & H263 (crashes with both)
FourCC: XVID
VHQ Mode: 0 - 4 (crashes with every mode)
Max I-Frame: 250
Min I: 1
Unchecked all except Chroma Motion

BFrames: -1 & 2 (of course, it crashes with both settings)
B-F quant Ratio: 150
B-F quant Offset: 100
Threshold: 100
Checked DX50 B-VOP comp.

all other options at default values

--------------------------------------------------------------------

I exclusively encode DVD source material.

Now my thoughts / observations:

1. The computer seems to crash EARLIER if b-frames are being used
2. disableing VHQ does not help
3. using another XviD version does not help
4. using earlier VirtualDub (-mod) Versions does not help
5. using other AviSynth versions does not help
6. AVS Script being used is the standard from GKnot, no additional filters are used
7. The computer does not seem to crash because it gets to hot (crashes at different temperatures, even tgough the temperatures _ARE_ pretty high!)

Now, what do you think?
I thought it could have been the power supply (I had a no-name before), but changing to an Enermax one did not help.
Do you think that the CPU gets too hot?

Are there known problems with recent NVidia Detonator / Via Hyperion 4in1 Drivers?

Did I choose unwise settings?

Do you think that it's a hardware problem?

I would be grateful for any help!

Cheers

MoritzT

[zulu]

Did you try the "load defaults" button?

[MoritzT]

Yep.

[Nic]

It probably is overheating im afraid :( As no one else is really having this trouble. Crashing doesn't always occur at a particular temperature...the high temperature causes CPU errors, it only takes time before one of those errors is enough to make it crash...

Sorry I cant be more help.

-Nic

[MoritzT]

Cheers Nic,

since this really seems to be the problem, I'm going to spend my next 70€ on a new CPU cooler....
Thx all

MoritzT

[Acaila]

I don't know what all you've got in your case already, but I suggest not to buy a new cpu cooler (and they cost way less than 70 euros btw), but additional ones. A cooler on the cpu can only do so much, what really helps are fans that direct airflow through the case, in and out.

I suggest you check out a few forums dedicated to hardware out there for advise/recommendations and general info if you're not familiar with cooling.

2cpu.com (http://forums.2cpu.com/forumdisplay.php?forumid=16) is my personal favorite, just to mention one :).

[MoritzT]

@Acaila

I already have two additional casefans (front + back), creating a flow through the case and my power supply sucks air out of the case as well.

The cooler will cost about 70€ because I want it to be as silent as possible!

Cheers

MoritzT

[geoffwa]

Have you tried the latest ffvfw (http://cultact-server.novi.dk/kpo/avisynth/avs_cvs.html) from sh0dan's page? (the 03/06 one and not the 25/04 one)

Changing your YV12/YUV2 codec seems to effect AviSynth 2.52 rather dramatically.

If not, give prime95's (http://www.mersenne.org/freesoft.htm) torture test a whirl, and see if that reports any errors.

[ciper]

@Moritz ...

This might be offtopic .. but may i ask what cooler are you going to buy for 70€ .. i'm having noise and heat problems aswell with my duron 1.3

[MoritzT]

@ciper

It's a Thermalright SLK800A with a Papst fan

[DoW]

I dont know if they changed the design since my roommate bought one, but the SLK800 he bought when they first came out was a litteral nightmare to get on the board. The clip was so tight and we (yes, it took two of us to get the thing on) had to put so much pressure on it to get it to clip on that I was afraid we would crack the core. One of us had to pull the clip outwards with a pair of pliars while the other one had to press down with a screwdriver to get it on the socket. On the flipside, I love my Thermalright SK-6+ w/ Sunon 4800 rpm 60x60 I got on my machine. I have it on my 2100+, never goes over 50C (and usually stays at 46-48) now that the Arctic Silver III has settled in (and the HS is only rated for a 1700+ too).

[MoritzT]

@DoW

Thx, but I wouldn't have bought this fan + heatsink anyway.
I'm gonna buy the Arctic Cooling Copper Silent 2.

Moritz

[-=Stan=-]

Go to debug page and select force optimizations instead of automatically detect optimizations .
Now check off Integer SSE if it is checked and try encoding again .
It solved my crashing problem.

Maybe it wud save your 70€ as well.;)

[MoritzT]

At first I thought that your way worked, but trying to encode a longer clip (30 min), the computer crashed after 18 seconds of encoding!

Thx for the advice anyway,

Cheers

Moritz

[kaitsuburi]

Originally posted by MoritzT
I already have two additional casefans (front + back), creating a flow through the case and my power supply sucks air out of the case as well.

The cooler will cost about 70€ because I want it to be as silent as possible!

Cheers

MoritzT

In order to have balanced airflow through your case, IMHO it's best to have an intake fan on the front and an exhaust fan (or fans) (plus the PSU fan) on the back. Putting the stronger fans on the back/exhaust side helps create better "pull" pressure and improves aiflow I believe.

Might want to invest in some Arctic Silver thermal compound to go with your new CPU cooler; it brings temps down even further.

-kaitsuburi

[Wilbert]

In the mean time, try the following:

Make sure that your computer is not overclocked and lower your bus speed of your processor in steps of (say) five MHz till the crashes are gone.

[MoritzT]

@kaitsuburi

That's exactly what I did.

I already have two additional casefans (front + back), creating a flow through the case and my power supply sucks air out of the case as well.

The one in the front sucks the air into the case, the one on the back is an exhaust fan.

@Wilbert
Make sure that your computer is not overclocked and lower your bus speed of your processor in steps of (say) five MHz till the crashes are gone.

The cpu isn't overclocked, and I don't think that I can lower the bus speed in such a way on my ECS board, but I'm gonna give it a try anyway.

Cheers

Moritz

[MoritzT]

* UPDATE *

Got the new cooler, cpu temp is down by 10° Celsius, and still the computer chrashed after about 20 mins.
Hope I'm not gonna have to change my mainboard.......

Moritz

[MoritzT]

* UPDATE 2 *

Just did an encode with DivX 5, which turned out to be fine (not quality wise ;) but in the sense that my computer did not crash).

This is really confusing, since this would mean that it is not a hardware problem. But what did I do wrong?
So my question is again, did I choose unwise settings?
Can I use the latest Nvidia Detonator?

Cheers, any help appreciated

Moritz


EDIT:
It seems that I was just lucky, my second DivX 5 enode crashed!
Damn it! I'm either going to have to buy a new Mainboard + CPU +RAM or give up my hobby of Encoding.

[Koepi]

Did you think about your power supply unit as well? Sometimes it can cause such a behaviour as well.

"Nice to see" that it's definatly not XviD's fault - sorry for your hardware being messed up.

Regards
Koepi

[MoritzT]

@Koepi

I don't think it's the power supply because I changed it to an Enermax one (which just won a comparhison in C'T magazine), and it happened with the old one (which was a no-name) as well.

Cheers

Moritz

[MoritzT]

@all

I switched my FSB from 133MHz down to 120MHz and as a result, my 1,4GHz Thunderbird is now running @ 1260 MHz - and it's working! I just did a 5h 1st pass without crashing!!!!
So I'm probably going to slow down the CPU everytime I want to encode with XviD.

Thx @ all, you're a great community :D

Moritz

[gpower2]

I had the same problem with you and thanks to this thread I solved it, I have a P4@1,7GHz and I used to have it overclocked @2GHz. Everything was fine, except XviD and some(fewer) DivX encodings. When I read the replies in this post, I guessed I should try to use the defaults settings, meaning to return to the 1,7GHz. Well, it did work! But you are using settings below the default, if I got it corerctly, so I am proposing you another solution:

Leave all the settings to default, except the CPU Voltage core. Increase it just a little. It may work, because by increasing the voltage, the 1 and 0 have bigger distance, so the CPU can understand more easily which is which!

I hope you see my reply and I hope it works!

[Sastraxi]

As you probably know, AMD heatsinks can only go on the chip one way- or else theye're going to be propped up on one side, diagonally. I did this once, and this caused any media or burnin programs to immediately shut off the computer. So, I turned around the heatsink. Unfortunately, my heat issues have never gone away, but the crashes have stopped. And I've never had a problem with anything like that since... =/

Though I think the *heat* is just my mainboard reporting a higher temperature than actually exists =/

[Mole]

Since turning down the FBS solved your problem, I think it might be your RAM settings which is the problem.

Find out whenever your RAM is CAS 2 or 2.5. Make sure the CAS setting is correct in BIOS as well.
You should set the RAM settings to "Use SPD" or something similar.

I've experienced similar problems with crashes like this if the RAM is 2.5, but is set to run at 2 in BIOS.

It would also be interesting to know WHAT type of crash you are getting.

Is it "blue screen of death", computer just freeze or program illegal exception.

Because if it just freeze, then it's most likely a hardware problem. BSOD is usually a hardware problem, but can also be software. While program crash is most likely software problem.

FYI, I use an Athlon XP Thoroughbred B 2000+ OC to FBS 166x12.5. (2080 MHz) and it's running rock stable. :D

[visigotik]

You should test your ram with microsoft memory diagnostic (http://oca.microsoft.com/en/windiag.asp)

[MoritzT]

@ gpower2

Leave all the settings to default, except the CPU Voltage core. Increase it just a little. It may work, because by increasing the voltage, the 1 and 0 have bigger distance, so the CPU can understand more easily which is which!

I don't think my mainboard (Elitegroup...) supports this...


@ mole

Find out whenever your RAM is CAS 2 or 2.5. Make sure the CAS setting is correct in BIOS as well.
You should set the RAM settings to "Use SPD" or something similar.

I've experienced similar problems with crashes like this if the RAM is 2.5, but is set to run at 2 in BIOS.

I have CAS 2 Apacer RAM, and the BIOS setting is actually "by SPD"


It would also be interesting to know WHAT type of crash you are getting

The whole system just freezes, I cannot move the mouse anymore etc, no BSOD - hardware reset needed

@ visigotik

You should test your ram with microsoft memory diagnostic

I've tested my RAM with this prog and many others, RAM seems to be ok


Another thing, ever since I "under-clocked" my CPU, C&C Generals doesn't crash anymore (it did before)

Thx @ all

Moritz

[Mole]

Try setting your CAS to 2.5, but keep the FBS at 133.

See if it still crash.

If it's OK, your RAM is either really CAS 2.5, or can't run reliably at CAS 2.

If you are positively that it's CAS 2, then I suggest you get your RAM replaced, because naturally, if it's CAS 2, it's supposed to run reliably as well.

[TheXung]

I would suspect the RAM as well. Use less agressive settings. Just set your memory timings to higher numbers. If you have more than 1 stick, try leaving only one in and see if it crashes.

If you really think it's not the RAM, then see if the heatsink on your chipset fell off or is loose. If not, then I would see if your voltage regulators or power capacitors are damaged in any way. Check if any of the big capacitors look over-inflated or if your voltage regulators have any burn marks at the solder joints. Finally see if you are missing any surface moint capacitors near the RAM, like if anything was rubbed off or fell off (they really can just rub off). Surface mount capacitors looked like little rectangles soldered to the surface of the board attach by their two ends. That's the order of things I would check, that is if you're totally sure it's not your power supply.

When you tested your RAM, did you have it at 120 or 133?