Hello! New member, and I hope I have something to contribute on day one...lol
I recently purchased a Dell Inspiron 9100 (XPS style) laptop that has the 128Mb. ATI Radeon Mobility 9700 graphics card in it. The first time I booted it up, I got the ATI infinite loop BSOD. It would boot up to the point where the login screen was supposed to load, and it would pause there for about 20 seconds and cough up the BSOD. So I held the power button down to shut down and then restarted the laptop. It went right in the second time.
I changed drivers, no good. I tried Omega drivers, no good. I reinstalled XP Pro, no good. One day I tried to start it up off of the battery instead of the power supply, and I got one infinite loop BSOD after another. I could not load XP while running on battery. Later that day, the laptop coughed up the infinite loop BSOD while just sitting there idling in XP.
I researched this on the net, and I am surprised that I have never encountered this problem before. I build custom computers for a living (own business), and I use ATI exclusively unless a customer specifies otherwise. Out of hundreds of systems I have built, none have had this problem. So that caught my eye right away. Why have I been the exception? Also, I know Dell never would have sold this laptop if it ran like this from day one. These clues led me to find and fix the problem with my laptop, and now it runs perfectly. On battery and on AC power.
In all of my reading, one thing that kept coming to mind was heat. The Dell 9100 laptop I have has a 3.2 GHz Prescott processor and Intel chipset package, on top of the Radeon. It is a small furnace when gaming. When I build custom gaming PC's, I use aftermarket GPU coolers and thermal paste (AS Ceramique on the GPU and AS Thermal Adhesive on RAM sinks) and large, well ventilated cases that are equipped with way more power supply than is necessary. So my systems run cooler because they are designed to do so.
I installed a temp monitoring system on my 9100 laptop, and I recorded the temps at idle and full gaming loads. The cpu was hitting in the mid 150's (F) under load and idling at 129. The chipset was hitting the mid 150's under load, and idling at 136. The GPU was hitting low 160's under load and idling at 134. All of these temps are high, at idle and load. I pulled the three fans out of the laptop, cleaned them (they were pretty clean to begin with) and tested again. The temps barely dropped (1-2 degrees).
So I pulled the laptop apart, pulled the thermal pads that Dell used and checked the heatsinks for contact. The cpu and northbridge heatsinks sat right on the chips, so I reassembled them with AS Ceramique thermal paste. I pulled the RAM sink and heatsink from the ATI 9700, and removed the thermal pads from them. I noticed that one of the two RAM thermal pads were discolored. Interesting. After pulling the GPU thermal pad, I placed the heatsink back on the GPU to check it for clearance.
I found the problem! The heatsink on the 9100 is almost .050 of an inch off of the GPU chip. The thermal pad showed that the GPU and the heatsink were not in close contact. I laughed my rear off when I saw the gap. How in the heck Dell thought this was ok is beyond me.
The way the heatsink mounts, the only way to bring it closer to the GPU is to shave the four mounting legs down (shorter), so the heatsink is closer to the GPU. My heatsink legs were .247 inch long, and I cut them down to just under .200 inch. The back brace the heatsink legs screw to also has to be cut down a like amount so it clamps the heatsink in place properly.
I carefully dry assembled the RAM and GPU heatsinks, and bent the tabs on the RAM sinks so they would line up with square with the RAM chips and to take up the space that the thinner adhesive strips left. The RAM sink is very thin and you can bend the legs (carefully!) on it with a pair of needlenose pliers. Once everything lined up, I replaced the thermal pads that were on the RAM sinks with thin thermal adhesive strips for computer memory heatsinks. I put a dab of AS Ceramique on the GPU and then reassembled the video card.
I put it back together, held my breath and fired the laptop up. It went into XP without hesitating. I shut it down and pulled the power and then started it up on the battery. It started and went in to XP fine. I have been using it for a week now, and no more problems. The cpu temp is 134 under load and 112 at idle, the GPU is 138 under load and 114 at idle, and the Northbridge chip is 145 under load and 115 at idle. This is a huge improvement over the original readings.
High end graphics cards are heat and power sensitive. Heat is resistance, and resistance means less power. That is why overclockers commonly mess with voltages on their system to stabilize them. Well, the inverse is true too. If you have just enough power to get the job done with little headroom left, overheating will cause problems where none existed before.
I hope this helps some people out. I do not think that this is the sole answer to the problem, but I bet it is a major factor. Another point is that switching to the PCI to PCI bridge does not push the video card as hard and that explains why it helps some people out. Temperatures are reduced with less drive, and it only masks the real problem.
Heatsink pads and compounds will eventually dry out. Service your computers, and do not assume that Dell (or any other manufacturer) did everything right the first time. That GPU to heatsink gap on my laptop was a clear mistake, and Dell expected the crappy foam thermal pad to do the job. It failed.
