Very high CPU usage with PVR-150 and GBPVR

[csy]

This is my results:

Menu(idle): 0-3% CPU, 0-1.5 DPC, 0-1.5 Int
LiveTV(preview): 16-36% CPU, 0-9 DPC, 0-1.5 Int
LiveTV(timeshift): 7-20% CPU, 0-9 DPC, 0-1.5 Int
LiveTV(paused): 3-14% CPU, 0-9 DPC, 0-1.5 Int
Playback: 7-17% CPU, 0-9 DPC, 0-1.5 Int
Record(rec-service doing a recording, GBPVR shutdown): 2-5% CPU, 0-1.5 DPC, 0-1.5 Int

Notes:
1) Using ATI550, Purevideo decoder, VMR7 (having issues with VMR9), ATI9550
2) LiveTV, recordings, and playback-video are all MPEG2 8Mb/s CBR

[ubu]

Well, that gives me something to shoot for, I guess! I've got a service call in to Hauppauge and also asked for advice on the WinTV PVR Board so, hopefully, I'll get some feedback from one of them. I'll try the PVR-150 in another box this weekend, just to make sure, but you've got me pretty much convinced that I'll end up asking Hauppauge for a replacement. (No wonder I've had problems with some of the "greedier" decoders).

I notice you're running everything at 8000 Kbs and constant bit rate. I'm using the defaults of 6000 Kbs CBR for live tv and 3500 Kbs VBR for recording (I mostly record at "medium"). I assume you've raised the bitrate to produce a higher quality result but I'm confused by the constant/variable choice. Why does GB-PVR default to CBR for live tv but to VBR for recording? And why have you chosen to go with CBR for both?

[csy]

Could another PVR150 user out there please run these tests for comparison
(my results are not a true one-for-one comparison because I'm running an ATI550 capture card).

Yes, I'm using 8000Kb/s for quality reasons. Anything I want to keep, I'm transcoding to H.264 to reduce filesize (reduces filesize to the equivalent of 5000Kb/s without any quality reduction - ideal for burning onto DVD).

I deliberately use CBR because the ATI550 does a very poor job of VBR (VBR 8-12Mb/s looses definition in fast moving scenes).

I suspect the GBPVR VBR/CBR defaults are simply Sub's arbitary settings.

[ricklous]

I use 8000kb/s CBR for everything and my figures are broadly similar to those of csy using Overlay on my specs below.

Main reason for posting though, ubu, how do you install GBPVR? I have two user accounts on my machine, one purely for running GBPVR and the other for remoting in and doing config stuff. GBPVR was originally installed when logged into the GBPVR account but when 97.7 was released, i remoted in and used the config user account to upgrade, and found that i was having very similar problems with CPU useage. tried everything to try and fix it (even uninstalling and installing a fresh 97.7 via the config user account but found that the problem remained until i restored a pre-97.7 image and then did the upgrade from the same account as the original account it was installed from.

I really dont understand why this happened, and havnt found any other threads about it so assumed that i wasnt seeing the real reason for the CPU chewing, but if you're using different accounts to upgrade GBPVR, maybe you're seeing the same thing i did?

Probably not though lol flipping PCs are so infuriating sometimes...

[ubu]

Probably not though lol flipping PCs are so infuriating sometimes...

Tell me about it!

GBPVR was originally installed when logged into the GBPVR account but when 97.7 was released, i remoted in and used the config user account to upgrade

lol - I did exactly the same thing! To try to debug this problem I did a "clean" install of everything on a freshly formatted drive. But, being a lazy lad, once I had XP installed I installed everything else, including 97.9, remotely using my non-gbpvr account. Discovered that this is a VERY BAD IDEA. The remote was talking to both machines at the same time, timeshift wouldn't work (I guess 'cos it was recording on one account and trying to watch it on another - two copies of the recording service were running at the same time). Etc., etc. So I did an uninstall from the non-gbpvr account and then re-installed from the gbpvr one and everything "seemed" to be OK. Probably should have scanned the registry after the uninstall though.

So that might explain my marginally higher CPU with 97.9, but I was having the problem before with 96.x. Also, I get similar CPU usage with WinTV2000 which does seem to suggest a bad card, I guess. The 85-100% cpu use I'm experiencing compared with the 16-36% that you and csy are getting seems like a really big difference. Since a lot of GB-PVR users have PVR-150 cards, I was hoping somebody who had experienced similar problems might read this thread.

Yes, I'm using 8000Kb/s for quality reasons. Anything I want to keep, I'm transcoding to H.264 to reduce filesize (reduces filesize to the equivalent of 5000Kb/s without any quality reduction - ideal for burning onto DVD).

So both you and ricklous are using 8000Kb/s. Once I get this problem fixed, looks like I should try fiddling with these settings.
What do you use to do the H.264 transcoding? I've had very good results with H.264 streams using Quicktime (and now VLC also supports H.264, so I can stop using that evil app).

[csy]

What do you use to do the H.264 transcoding?

I manually transcode using Nero Recode 2.2 (note h.264 was only introduced on version 2.2)
I had Nero OEM 6.0 supplied free with my DVD drive, and I download the free NVE-3.1.0.25 upgrade from Nero FTP site, which has Nero recode 2.2
Nero Recode is not perfect, it only encodes progressive format, so it has to first deinterlace your recording, which it does ok but not fantastic.
I am waiting for CoreAVC to release their encoder that will encode interlace format.

For H.264 playback, I purchased the CoreAVC H.264 decoder, which works very well.

[ubu]

The plot thickens..... Tried removing my wifi card and CPU use dropped by 10-15%. Was always a bit suspicious of that card.

Then moved the PVR-150 to my dev PC - similar spec (2.23Ghz P4) but different mother board - and found that I was clocking a mere 25-35% for live TV with minimal DPC usage (which is more in line with you guys) suggesting there's nothing much wrong with the PVR-150. So, filled with foreboding, I then dropped the wifi card into the dev PC as well and.... no change to the CPU usage at all! Still around 30%. :confused:

This all confirms my suspicion that matching up hardware is pretty much a random crap-shoot requiring luck, wild hunches and, ideally, witchcraft to get the happy combination that actually works together. Software does at least follow (usually) the rules of logic. Sheesh!

At least swapping the motherboards will be cheaper than buying a new capture card. A few hours of wielding the screwdriver and cursing, hoping all the while that you grounded yourself properly and haven't fried too many components, seems a small price to pay if I've finally nailed this problem.

[csy]

Interesting! Let us know how you get on.

[ubu]

So I swapped motherboards and, after a great deal of swearing at Bill Gates and his sorry excuse for an operating system (XP seems to behave differently every time you install it), I found that it helped the CPU problem some but didn't really solve it.

Bottom line - with the newly installed MSI mobo, I'm getting 65-85% CPU while watching live TV. If I remove my wifi PCI card, I'm getting more like 45-65% utilization. So I've ordered a cheap external USB wifi adapter from newegg on the theory that the PCI card disagrees with the PVR-150. If so, that's probably as good as I can expect with this card.

Meanwhile, back at the ranch, I've been carrying on a dispiriting dialogue with Hauppauge tech support. ("It's probably caused by using a tweaked version of the Forceware driver. Did you try the standard one?" Yeah. Right!). They finally trumped me by offering to "test" my card if I send it back to them. So now I have to decide if I can live without GB-PVR for several weeks (months? years?) only to have them tell me it tested "just fine". Or do I accept defeat and live with not getting the 20-30% CPU usage that csy and ricklous (and tipstir, I think) lead me to believe I should be getting?

[ricklous]

at least its process eh?

is there any way you could get hold of another tuner? maybe from a different manufacturer or something just for the hell of it, see if CPU usage is drastically different. maybe that would give you something more to hammer hauppauge with

i dont know about over in the US, but here in the UK we have a law covering internet purchases from uk websites, where you can return it with very little reason within 14 days. Supposed to account for the fact that you cant try clothes on/feel the thing purchased/taste wine over the net...but comes in useful for things like this too. I had to send a CPU cooler back once due to it not fitting on my motherboard. if id bought it from a bricks and mortar shop i wouldnt have been able to return it!

You could take the view that you may end up buying a new tuner anyways and just take the pinch before your card goes off to hauppauge....