[-EaS-]

After quite some puzzling I found out that my Digital Devices CineS2 card (a DVB-S2 satellite reception card) does not like my passively cooled (Sparkle) GT430:



The above combination has been tested on three systems (2x Intel and 1x AMD based), and it is clear:

• If the GT430 is active alongside the CineS2, tuner 1 has a *lot* of problems receiving quite some channels
  
• Tuner 2 is not affected
  
• If the GT430 is physically installed but *not active*, disabled or enabled in windows device manager - doesn't matter, and the iGPU is active - no problems at all.
  
• If the iGPU (HD2500 from the Celeron G1610) is active, but Nvida CUVID is selected for hardware deinterlacing (LAV filters) - problems are back again
  
• Linux or Windows - doesn't matter
  

Conclusion: The CineS2 card is nearly unusable alongside this GT430 card.

Solution:
1 - Use the CineS2 in combination with the Integrated Graphics from the Intel Celeron G1610 (HD2500) and remove the GT430
2 - Use the GT430 card for all content, and remove the CineS2 to a backend system

Actually I don't prefer to use a seperate system as backend, so hence my question:
Will I see a loss in picture quality when I switch from the GT430 to the integrated HD2500? Since loads of satellite channels come in 1080i, or whatever other exotic interlaced resolution, I want the best deinterlacing possible. I know that VDPAU is perfectly intergrated in XBMC, interlacing works flawless. But how is VAAPI hardware deinterlacing at the moment? Is hardware deinterlacing through VAAPI on the HD2500 even comparable to VDPAU on the GT430? Does the HD2500 actually even *do* hardware deinterlacing under linux?

[nickr]

Do you think there might be an irq sharing problem.

[-EaS-]

Do you think there might be an irq sharing problem.

No shared IRQ's (or I/O or Memory Ranges for that matter):

[nickr]

I meant between the Cine card and the nVidia card but no doubt you can check that out too.

To answer your original question, on linux the nVidia card will beat anything at deinterlacing, no competition.

[wsnipex]

on linux nvidia is superior, specially when it comes to deinterlacing. Vaapi deinterlacing is kinda broken too.
If you can, move the tuner to a seperate system.

[-EaS-]

I meant between the Cine card and the nVidia card but no doubt you can check that out too.

Oops, that is what I meant to show

Code:

xbmc@HTPC:~$ cat /proc/interrupts
           CPU0       CPU1
  0:         43          0   IO-APIC-edge      timer
  8:          1          0   IO-APIC-edge      rtc0
  9:          0          0   IO-APIC-fasteoi   acpi
16:         65    3969153   IO-APIC-fasteoi   ehci_hcd:usb1, nvidia
17:        136          5   IO-APIC-fasteoi   snd_hda_intel
23:        111       9556   IO-APIC-fasteoi   ehci_hcd:usb2
41:       9579     137535   PCI-MSI-edge      ahci
42:          0          0   PCI-MSI-edge      xhci_hcd
43:     605803          3   PCI-MSI-edge      em1
44:         15          0   PCI-MSI-edge      mei
45:    1781600     300284   PCI-MSI-edge      DDBridge
46:    1281458     232210   PCI-MSI-edge      snd_hda_intel

As you see: No shared IRQ between 'nvidia' and 'DDBridge'
I already emailed Digital Devices and explained the problem. Let's see what they answer.

To answer your original question, on linux the nVidia card will beat anything at deinterlacing, no competition.

on linux nvidia is superior, specially when it comes to deinterlacing. Vaapi deinterlacing is kinda broken too.
If you can, move the tuner to a seperate system.

Very clear answers. Thanks! Hadn't followed developments on the Intel VAAPI side of things, seems not much has changed.
My solution for now will be using Tuner 2 and keep the GT430 in my HTPC.