Tazz <dmswaine@gmail.com> wrote:

<snip>

I use the onboard ATI graphics adapter for VMware server
and High Def video capture. CUDA co-exists well with these, 
running as it does on the Nvidia card.

Right now, I only use the Nvidia card for (CUDA and) High Def video
playback with the 28" monitor. 

<snip>

Let me see if I understand this right. I *might* be able to use the 
onboard geforce 8200 for my monitor AND have SETI running on the 9500GT 
PCI-e videocard? I ask because the last time I put a PC together (in 
'03) if there was onboard video, you had to disable it to use an AGP 
videocard. It was one or the other. I thought that was the way it was 
and forever will be.

I have been using both consistantly. About 6 months ago, I built this machine
specifically to record High Definition video off the HD DirecTivo and play it
back. This has been challenging. Playing back is more demanding than 
recording because the H264 compression is done in the $200 Hauppague USB
box, while decompression is done in the (mandatory) dual-core CPU. I was
going cheap, so I decided to use the Gigabyte MA78G-DS3H board with the 
"ATI Radeon HD 3200 Graphics Core", "Full HD 1080P Blu-ray playback". 

This setup was not up to the task of playing back the H264 compressed video. 
This was even more the case when I added a fiber optic cable so I could
record 5.1 surround sound. My principal test was the 2008 Bejing Olympics
Opening Ceremony. The motion kept jerking annoyingly. Further proof
was the fact that the little "Process Explorer" square by the clock kept
maxing out on CPU during playback.

A further complication occurred when I tried to run the whole thing using
VNC from my wireless laptop from across the room. Media Center Edition
comes with special fonts to allow you to work it from across the room, but 
MCE doesn't support the Hauppague (or vice-versa), and, besides, I don't
have it. VNC on the wireless laptop wasn't happening for the purpose of 
running the media software, the connection was too slow, and VNC screwed
up the recording process.

Therefore, I connected a second monitor, actually a 15" widescreen TV,
via a long cable to the integrated ATI Graphics adapter. That's right, I'm 
using a 28" widescreen monitor as a TV, and a 15" widescreen TV as a 
monitor. This arrangement worked fine for recording, the CPU's run at
about 45%, but was unacceptable for 1080i playback.

Therefore I bought, for a mere additional $39, a Gigabyte GeForce 8900 GS
PCI Express graphics card. I have always had less hassle with Nvidia than
with ATI. This card was a bit of a mistake, however, because I bought a 
PCI-E 1.0 card to use with a PCI-E 2.0 interface. Never mind. The card worked
great. All my playback problems went away, even with 1080i and 5.1 surround
sound. The CPU's hovered at around 75%. The Nvidia card also supports HDCP, 
which the ATI doesn't.

The problems came when I tried to use two monitors. It was a catastrophe. I 
had disabled the internal ATI graphice and connected both monitors to the
Nvidia card. It insisted on booting to the large monitor. Then when Windows
started, it would switch to the small monitor. Then the taskbar would appear
on the large monitor, with the icons on the small monitor. Then when I would get
everything "right", the next time I rebooted, it would be all screwed up
again. Then there was no telling which screen would be used when I ran a 
program. It would open on one screen or the other, with very little predictability.
Furthermore, the playback was more jerky than it was when I used the integrated 
ATI.

I was going to send the card back and get an ATI PCI-E 2.0 card that I hoped
would be "more compatible". Then I got interested in CUDA (which only runs on
Nvidia). While trying CUDA, I wanted to do some other stuff with the computer
without stopping CUDA, since I had so much trouble getting it to run in the first 
place. This was (understandably) not happening as long as both monitors were
connedted to the Nvidia card. It was too busy to let me do anything.

"Just for the heck of it", I enabled the integrated ATI adapter and connected
the small monitor to it. I too had heard that you weren't "s'posed" to run two
different graphics adapters, especially with two different chipset manufacturers,
at the same time on the same computer. Instantly, to my surprise, all the 
problems went away. Windows boots to the small screen, the taskbar and 
icons are all in the right place, and windows open where they are supposed to.
Playback is flawless on either screen, even with 1080i and 5.1 surround
sound, and the CPUs hover at around 80%. 

Furthermore, unless I am doing something "demanding" on the big screen,
I can just leave CUDA running, and I don't even know it's there. A slight
problem is that since I started running GPU and CPU units at the same time,
I have to shut down the whole project if I want to use the computer, because
I can't suspend only the CPU units without new ones starting.

I have given this rather long-winded explanation to answer your question
and to explain why I am running it as I am. I can not say that if you have an 
integrated and an external Nvidia adapter, whether you will be able to run
two CUDA units. I guess that would be a total of six at once with a quad-core
CPU? I can only say that while running them in the manner I have described, for the
purposes I have described, they are working extremely well, and doing exactly
what I want, without any problems.

Thanks for giving me a chance to explain all this.
DaveT