Very high CPU usage with PVR-150 and GBPVR

[ubu]

So the wifi PCI card I was suspicious of did, in fact, turn out to be one of the major culprits. I removed it and its drivers today and replaced it with an external Zonet USB wifi adapter. CPU usage has dropped to 30-50% for live TV. Not quite as good as I'd hoped, but certainly acceptable.

So the moral seems to be that PCI slots should be used for additional capture cards and not for anything else.

Also, that motherboard characteristics vary more than you would expect and that a mobo that makes a perfectly good general purpose computer can cause unexpected problems when used as an HTPC.

I think I'll avoid sending the PVR-150 back to Hauppauge, even though I'm still not clear that it is working 100% "as advertised". Instead, ricklous has now inspired me to see if I can get my piece-of-junk Prolink PixelView PlayTVPro BT878 card working with GB-PVR - "just for the hell of it". It would be pretty funny if it ends up using less CPU cycles than the PVR-150.:eek:

[wtg]

Check the interrupts being used by your hardware. It used to be that PC's couldn't have devices that share IRQs. Today they can, but some do it much better than others. If you have your video card, sound card and tuner card all sharing the same IRQ, you'll notice higher CPU utilization than if they all had their own. CPU utilization goes up even though the CPU isn't really doing much other than waiting for one device to give up the IRQ so it can let another device use it.

Some PCs BIOS' let you control what PCI slots get which IRQ, giving you lots of control. Others assign 2 or more IRQs to be shared by various hardcoded PCI slots, so swapping PCI slots can make a difference, while others give you no control whatsoever. You may find you can solve the problem this way, or not .


Good luck!

[ubu]

Check the interrupts being used by your hardware. It used to be that PC's couldn't have devices that share IRQs. Today they can, but some do it much better than others. If you have your video card, sound card and tuner card all sharing the same IRQ, you'll notice higher CPU utilization than if they all had their own. CPU utilization goes up even though the CPU isn't really doing much other than waiting for one device to give up the IRQ so it can let another device use it.

Some PCs BIOS' let you control what PCI slots get which IRQ, giving you lots of control. Others assign 2 or more IRQs to be shared by various hardcoded PCI slots, so swapping PCI slots can make a difference, while others give you no control whatsoever. You may find you can solve the problem this way, or not .

Absolutely! That was my first thought too. But (on the original mother board) I had checked IRQ settings and tried every possible slot permutation for the cards which is why I figured something else was going on. That partiular ASUS board has a reputation for being finicky (in fact I had to send back an ATI graphics card when I found out on Google that the card, for no clear reason, was well known to simply not work in that particular Asus mobo).

Thanks, anyway, for your clear explanation of the IRQ issues. Since I suspect this thread may be a useful reference thread for others with CPU problems with GB-PVR, it's good to get all the issues in one place.

[gkreis]

I too have seen a leap in constant interrupts and DPCs that are making my machine very sluggish (according to procexp.exe). I can't say for sure, but I am pretty certain it coincided with the installation of the 97.13. (I skipped the one right before that). Unfortunately, I can't find a way to track down the source of this flood of interrrupts.

I did convert to using SQLlite and I did have to remove some dlls that were not compatible with this gbpvr version. Could there be some sort of registry issue that erupted because I removed the plugin dlls and didn't unregister them?

RootkitRevealer showed that gbpvr's metadata.xml is hidden from the Windows APIs. Hmm.... can anyone else download the rootkitsrevealer from sysinternals.com and see what they get for results?

[pastro]

Also normally when you see interrupt processing go high the DPCs go up too, since the DPC is usually set up in the interrupt handler to finish handling interrupt processing that doesn't need to be run at highest priority or in real time.

[sub]

RootkitRevealer showed that gbpvr's metadata.xml is hidden from the Windows APIs. Hmm.... can anyone else download the rootkitsrevealer from sysinternals.com and see what they get for results?

Thats odd it would report that. Its not hidden from Windows APIs, but it is an alternate data stream. Its used to store metadata inside the MPEG files without interfering with the actual content.

Windows APIs have no problem opening. Infact, you can open it with notepad, for example:

notepad c:\path\to\recording.mpg:metadata.xml

[gkreis]

Here is a screen capture to show what RootkitRevealer shows.

Are the interrupts that are shown by procexp only possible from hardware? In other words, am I supposed to look for a bad card? Any suggestions on how to diagnose this?

[gkreis]

I pulled out the AverMedia 150 card and the interrupts settled down. I put it back in and sure enough, they started up again. So it is the capture card, but it sure seems a strange coincidence that this started when I installed the latest GBPVR.

Could a plugin cause this kind of a problem?

I wonder how many others have this going on and they don't know it if their processor is fast enough to mask it? What is really strange, is that Task Manager will show 98% in idle, but if you look at the chart tab, it shows 30-50% of the CPU is busy.

[gkreis]

Well, all attempts to uninstall the card, uninstall the drivers and reinstall have failed (other forums indicated this helped with other hardware interrupts and DPC excesses on sound cards, etc.).

I ordered a "VisionTek XTASY Theater 550 Pro TV Tuner Card PVR MCE with Remote Control" ($30 rebate from Buy.com) to move on. I think I fried the other one with heat. I was trying to scrimp along on an old PC case, so I ordered a new one from Amazon (Antec NK4400) which seems to be a good compromise in noise, ventilation and price.

[gkreis]

Here with an update. While waiting for my new tuner card, I got a hunch that the problem might be that the overclocking might have pushed the PCI bus too far, so I slowed the machine down and the wild interrupts and DPCs went away! Strange, since it has been overclocked for a long time. Maybe the heat of summer pushed the electronics too much. It went from 37.5 Mhz down to the standard 33Mhz.

I transferred to the new Antec NK4400 case and the temperature in the case dropped from the mid 40's C to 27 C... wow. The processor had a rewarding drop from 51C to 41C, as you would expect based on the drop in the case, and the rig is much quieter. The huge 120mm exhaust fan, set at the slowest speed, is very quiet and moves a decent amount of air. I have a suspended Zalman fan (seems like it is 80 or 90mm) set to 1600 rpm blowing on the copper flower heatsink. I suspect the case would be noticeably noisier with a traditional CPU cooler, especially since the NK4400 case has a hole in the side of the case with a shroud to funnel cool outside air to the CPU.

P.S.
I still have the Visiontek card on the way (great excuse with the misses to get another tuner ;-) since it was too late to cancel.