The Power Supply Fan is still working, but does not appear to respond to increased loads, and its speed cannot be increased using SMCFanControl. Any suggestions?

Question

Hi all - My mac pro 1,1 has an enigmatic problem. It's quad-core, 3Ghz, running OS X 10.7.5. The problem began some months back, when running HD video conferences, HD streaming videos and some other graphics-moderate (I won't say intense) uses, such as slide shows in Just Looking. My machine has an ATY, RadeonX1900 card and 2 monitors - both Apple displays, one being 30 inch 2560x1600, the other being 24 inch 1920x1200.
When I start those types of graphics-moderate apps, after about 5 minutes, something would crash (the monitors making a noise as it did) and the displays would both go black. The power did not appear to be fully cut off from the monitors. I think that the underlying applications, however, did not crash. So, I could still hear and speak to people during video conferences, and I could hear the sound tracks of videos.
Attempts to restart the computer and get the monitors back did not immediately work to undo the problem, if done straight away. The only way to correct the problem (so as to have the monitors work) was to wait for a few minutes and then restart.
After a lot of puzzled looking at console logs (and I confess that I am not an expert), I ran into a blank.
But, because of the timing of the crashes (after ~5minutes use) and restarts (after some minutes wait), I reasoned that the problem might be temperature related.
Once on that path of thought, I found this article:
http://worldtv.com/blog/guides_tutorials/fixing_an_overheating_mac_pro_no_expens e_required.php
I installed both Temperature Monitor and SMCFanControl. And I experimented, and got interesting results:
Now to the puzzling bit: SMCFanControl shows that the speed of the Power Supply Fan doesn't vary much, even if I use SMCFanControl to ramp fan speeds right up. All of the other fans react to SMCFanControl, going up to and above 2500rpm when commanded. But the Power Supply Fan does not.  Having watched it for a few days now, the Power Supply Fan, according to AFCFanControl, never varies outside of a range from about 600-640 rpm, regardless of what I do with AFCFanControl.
Also (and not surprisingly), whenever the graphics-moderate apps run, when I ramp all the fan speeds up using SMCFanControl, Temperature Monitor shows that the all of the temperatures within the machine plummet, except in the power supply locations. Power Supply Location 2 is the worst affected. Because the Power Supply Fan does not react as the other fans do, the "crashes" still occur.
I could post some screen shots of the read outs of SMCFanControl and Temperature Monitor, although I think I have covered the key points in this post.
It seems that the Power Supply Fan is still working, but does not respond to increased loads.
Instinctively, I'm worried that if that's right, it might be a hard problem to fix. Any suggestions? Help! I would really appreciate them!
Cheers
P

Answer 1

Thanks for your response Grant. I had already read about, and I think excluded, dust in graphics card/fan/assembly as a cause. The entire inside of the computer, including the graphics card/fan/assembly, is as clean as a whistle. It looks clean. I've cleaned the dust out every now and then (including a few days ago). Also, my undersanding (derived from the article, linked in my first post) is that I might expect to see high temperatures in the Northbridge and memory module B if the graphics card was overheating. According to Temperature Monitor, however, locations do not increase in temperature when I run the relevant apps. Finally, the graphics card/assembly does not feel too hot to touch.
It is strange that SMCFanControl lists the Power Supply Fan as being software controllable, if that fan is not. Still, your comments are consistent with my observations, as the fan simply does not respond to software control.
There are two upsides: First, I've something of a work-around. I've worked out what temperature must be achieved in Power Supply Location 2 for a crash to occur, and I can watch that location in Temperature Monitor, terminating the process before that treshold is reached. Of course, that's not ideal. Second, ramping all the other fans up does significantly delay the onset of the problem, and makes recovery time much quicker.
Is there any donwside to peridically ramping up the other fans when running the relevant apps? Given that it causes all of the other locations to cool significantly, could this lessesn the life expecancy of the mahcine? (I don't want to run the fans up all the time - that's like sitting next to a jet ready for take off!)
Also, a correction to my first post: Yesterday, with all the other fans running on full and running an HD streaming video, the Power Supply Fan did, eventually, slowly, come up to ~740rpm, which is much higher than the ~640rpm maximum I had previously seen, and this may have aided delaying the onset of the problem. 
I'm not sure where this leaves me? The problem might now be able managed. Well, sort-of. But it's not eliminated.