Ordered a new capture card

[landrover]

I''ll be watching your steps closely...

[Flying dutchman]

Update #2 (still under WinXP, "standard PC" mode):

Captured full-frame (720x576) Picvideo. Program stopped recording spontaneously after 46 minutes (10.2 gigabytes) and "hung". I had to terminate it by force. The file was OK though. Since Huffy seems to run stable, maybe Picvideo is to blame (though it never let me down before). Trying Mainconcept now..

[landrover]

Dutch,

Don't know if it has anything to do with it but I feel the urge to tell you that the system crash that ruined my former hard drive, occurred during playback of a 7054x576 file captured with MainConcept MJPEG. This happened a few minutes after my succesful capture with MC's DV-codec (see earlier on in this thread).
I must say that since rebuilding the set I have used the MC MJPEG codec again, without any problems.
If you look at Cinergy's codec list, there's "MJPEG Kompressor" listed. Do you know where that comes from? Or does it refer to my (disabled) Zoran chip?

[Flying dutchman]

The "MJPG codec" you mention is unusable, it's way too slow to capture full-resolution. I am not certain to which software package it belongs, but I assume it was installed by the Cinergy.

I'm currently trying out the Morgan codec; I had an evaluation version of Mainconcept but I found the bitrate too low (1 mb per second).

[Indiana]

Yes, this MJPEG Compressor thing is definitely installed by the Cinergy software. I have it as well - and as said I did a clean install of WinXP, so there's no other way it could've got on the system.

The PicVideo MJPEG is afaik both faster and better quality-wise than the Mainconcept one (what I read from previous threads here). But then I only have the PicVideo, so I can't do a comparision for myself.

I just captured a larger video (about 1 hour) with the PicVideo and didn't have any major problems - what I did notice, however, was that the fps counter was somewhere around 23.6 fps compared to the 24.8 fps when doing a realtime DivX grab. But then I didn't lose more frames than with realtime DivX'ing, and the results was playing back smooth - so it might be that the fps counter is just broken?

Another thing I noticed: I still dropped about 1 frame every 10 minutes (same as with realtime DivX) so this doesn't seem to be a CPU-load issue. Maybe ACPI is to blame here (I have a "standard" ACPI install of WinXP).

[landrover]

Indiana,

this fps thing is getting really interesting. I´ve always based my opinion on "frame drops or not" on the fps counter that Cinergy displays. When capturing 24.8 or 24.9 fps you should drop more than 1 frame every 10 minutes. Maybe my situation is not so bad after all: I too get 24.8 or 24.9 fps when capturing 640x480 or 480x576. I concluded that I dropped frames just because of Cinergy´s fps display. I think I didn´t even always watch the captured AVI to see if it actually suffered from frame drops.
One strange thing about your story. How can you capture full size DivX easier than PicVideo? As you know, DivX is by far more CPU-intensive. Maybe Cinergy´s capture software is more or less optimized for DivX: it´s the only third party codec they mention by name when promoting the Cinergy cards.

So it looks like PICVideo does it better than MainConcept´s. I´ll give it a try. I would like to add though that MainConcept states that their commercial MJPEG codec slightly outperforms the demo version, because of the demo´s superimposed Mainconcept logo.

[Indiana]

The Cinergy software has a framedrop counter: The number after the "F:" shows the amount of frames captured, the "D:" shows the drops. How the fps counter can show a different thing is beyond my understanding, maybe the software just doesn't measure the time correctly.

There is no real difference between framedrops here when capturing DivX and MJPEG, so I guess these few framedrops are NOT due to CPU utilization.
MJPEG still of course is definitely easier on the computer here than DivX (and not vice versa), since when capturing MJPEG I can browse the net without many additional framedrops, but when capturing DivX, I better leave the computer alone for the time of the capture.

[Flying dutchman]

My impression is the Cinergy doesn't cope with interrupt sharing too well. My ACPI system shares IRQ 9 on all PCI slots which worked remarkably well in the past (Marvel G200) but doesn't work well with the Cinergy. Even under Windows XP it didn't run stable until I disabled ACPI and gave every device a separate IRQ. Of course it may also be that one of the other PCI cards has a lousy driver that doesn't cope with IRQ sharing well.

Under these conditions HuffYuv appears to run stable even at full-frame size, but Picvideo does not. I still have to give the other MJPG candidates a real good beating to find out whether they run stable. The image quality, using Huffyuv, is truly amazing and makes my mouth water. Brightness and contrast are right on the spot and sharpness is perfect. Yet regardless of the codec, it drops a frame every 3 minutes or so (using the Cinergy's sound hardware, I haven't tried it on the sound card yet).

Newer motherboards support "APIC" which gives the system more than 16 IRQ's so no sharing is necessary anymore. Definitely something I'll consider when buying a new board.

[Indiana]

Yes, my Board does support APIC, maybe that's why the card seems to be working better here (and also stable with PicVideo)

[Ghydda]

SMP and Cinergy

Just got my first BSOD naming cap7134.sys as the culprit.
I've got version 1,4,0,0 of the file installed. I presume it's the one shipping with the card.
Seems like my first conclusion on SMP compability was wrong.

I'll go pound someone's helpdesk....


Ghydda

[Flying dutchman]

The Cinergy appears to run stable on a Windows XP system without ACPI provided it has its own IRQ. I'll try the same setup under Win2000 soon. I used the standard Terratec drivers supplied on CD.

The only codec it really doesn't like is Picvideo; that one crashes consistently after a certain period of time, somewhere between 40 an 60 minutes. Too bad! I scavenged the internet for demo and trial versions of MJPG codecs and all ran stable except Picvideo. Yet all lag somewhat behind in quality when compared to Picvideo. The Morgan codec comes close, though. HuffYuv runs perfectly. I stumbled upon a modified version of Huffyuv 2.1.1 somewhere in the Internet that alledgedly fixes some instabilities (not that I ever noticed any).

A peculiarity: When using the internal audio chip, the Cinergy consistently drops one frame per 200 seconds no matter what codec I use.

So far the device looks promising. I hope the other MURCERs don't mind my starting this thread, it's getting rather long and stays near the top of the list...

[landrover]

Have you figured out where WinDVR stores it´s channel information?
Does running without ACPI involve reinstalling the OS? On my system the Cinergy shares IRQ 5 with 2 USB Host controllers. But hell, even the good old Marvel has to share: IRQ 11 together with the Promise controller. The rig´s running very, very stable, so I would only permit myself fiddling around if it returned a leap in performance of the Cinergy.

[Indiana]

Did my first try with a scheduled Capture and this did stop without apparent reason. No crash, no HD-full or other problems, but it didn't finish the scheduled time.
My manually started captures have worked until now, though.

[Flying dutchman]

@landrover:

There's an easier option. In device manager, create a new hardware profile. Call it "video".

Then reboot - you'll have the choice between profile #1 and "video". Choose "video".

Now go into device manager again, and "disable in this profile" all devices that you don't need for video capturing, such as network, USB etc. Thus you can effectively avoid IRQ sharing without modifying your original settings.

[landrover]

Yes, but I´m dealing with a setup in which the graphics card and the RAID controller share an IRQ. Maybe the first thing I should do is move the controller to another PCI slot. Maybe then it shares it´s IRQ with something that can be effectively disabled. Good thinking? Or will XP re-distribute its IRQ´s when I disable unnecessary devices without moving the controller?