The majority of upgrades that people go through with their computers usually go without a hitch, but now and then problems do arise. I was lucky enough to get my hands on the ATI AIW Radeon 9700 Pro that Hubert reviewed here as an upgrade for my old GeForce 4 Ti4600. I was pretty happy to get this card since even when it becomes "old" it'll make a great basis for a future HTPC with its multimedia capabilities and the Remote Wonder to go with it. When it arrived though I discovered a rather annoying issue. Upon hardware installation, my computer wouldn't power up for more than a second before shutting down.
As soon as the card arrived, I went about clearing the nVidia drivers and my TV Card drivers from my system ready to install the 9700 Pro AIW. I powered down the machine and swapped in the Radeon. This is where all the trouble began.
As soon as I powered up the PC, it shut itself off. Ok, stuff like this happens. I checked all of the connections, made sure the card was in place properly, made sure I hadn't dislodged anything else by accident, made sure no loose screws were shorting something out, but everything looked fine.
Tried it again and the same thing happened. My first thought was that this was a common nForce2 problem with the BIOS, although usually if this is the case, not powering up isn't a symptom (not reaching POST is). My next thought was that the card was dead, despite getting no warning beeps from the system, however putting the card into a different system I had no issue at all. After a few more troubleshooting procedures it boiled down to this. The problem seemed to be with the Radeon 9700 Pro card as disconnecting the floppy power connector to the card allowed the computer to power up fine, albeit to only get a display warning me that the power connector wasn't connected. All the signs lead to a power issue, but the fact that I had just finished testing and reviewing a PSU that is personally the best PSU I have seen brought me some confusion.
So with the Ti4600 back in place I went surfing for information (or at least individuals with the same problem) and my first stop was the Rage3D forums. Lo and behold I come across with individuals with the exact same problems. They had exactly the same symptoms, with differing systems and specifications, but the main commonality appeared to be the Radeon 9700 Pro AIW card (and specifically a certain PCB revision).
The actual facts of what is happening appear to be a bit hazy but using the information I found at Rage3D, as well as from a few other sites ( is your friend) it would seem the most logical assumption is this. As your machine powers up, the capacitors on your motherboard and graphics card draw a lot of power initially. Now the Radeons draw a lot more than most, and the All-In-Wonder range more than others, which makes the PSU work overtime for that first second of start up.
The PSU I have here is the Aerocool Aeropower II Titanium 520w PSU and amongst its features are Over Volt Protection and Over Current Protection. It would seem that the combined draw from the motherboard and the Radeon is enough to trip the PSU into shutting down the power. You might think this means that the PSU is to blame, but since a lot of high end PSUs have these features for a reason I don't feel its right to place the blame with them.
I had the same issue using a Vantec Stealth 480w, and an Enermax 350w. Some users reported luck with Q-Tec or Kingwin PSU's. We could perhaps blame the motherboard but with such a wide range of differing systems having this issue it isn't likely, although some folks report swapping the motherboard for a different one solved it for them (yet others have the same problem with the same motherboards). Also the fact that some folks report returning the card for a new one solved the problem for them (and others report that it doesn't), leads me to believe that the issue is likely with the Radeon Card.
The main PCB revision recorded that is affected in this way is the ones marked "PN-xxx-xxxxx-10". This isn't to say that this affects all Radeons, or to say that it is exclusively a 9700 Pro AIW problem, but it does happen and the majority of reported cases I could find were AiW cards. It's basically a combination of the motherboard, the Radeon and how your PSU reacts to the initial power draw by your system.
Well the problem here is that the PSU doesn't like the initial draw of power by both the Graphics card and the Motherboard at the same time so this left me with only few options.
1) Send the card back to ATI and pray I get lucky
2) Swap the motherboard and pray I get lucky
3) Get a PSU that doesn't have the various protections
4) Create some form of delaying circuit to attach to the 5v rail of the floppy power connector for the Radeon
5) Install a switch in the 5v rail floppy connector and delay the initial power manually
The last 2 showed the most likelihood of success in my opinion, as the others are very much all down to the luck of the draw (pun intended). As I didn't have any diodes or resistors laying around, and being the impatient type, I chose to install a switch.
Usual Disclaimer: Neither myself or Viperlair will be held responsible if you damage your equipment or electrocute yourself by using the information below; you do this at your own risk.
For this, you will need to find a disused switch capable of handling 5v and being of the gate type. By gate type I mean that once moved to the ON position it will remain in that position until moved to OFF. A simple push button switch that returns to its original state when you let go is no good to you. Hunting around the loft I found an old lamp with a sliding switch attached to the power lead, hopefully the wife won't want to use it again. ATI supplies a Molex-to-Floppy adaptor with a pass thru so I chose to hack this up as well.
On the floppy connector of the adaptor lead, we find 4 wires, a red, 2 blacks, and a yellow. The 2 blacks are ground wires, the yellow is +12 volts and the red is +5 volts. It's the red wire I installed the switch to. It's a simple case of cutting the red 5v wire, and wiring the switch in line with it. If I need to explain how, I don't think you should be reading this.
So how this works is that when I first power up the computer, I leave the switch in the OFF position. About a second after I turn on the power to the system, I put the new switch to the Radeon into the ON position, just before I get the warning bar about having no power connector. This allows the motherboard to draw power first, and then afterwards the Radeon can get its share without tripping the OCP/OVP features of the PSUs.
Subsequent reboots are not an issue. It is only after I power down that I need to follow the same procedure again. Since I run my PC 24/7 this isn't a major deal to me (although the fact I have to do this at all is). Not an elegant solution, but a quick fix that works none the less.
Having an issue like this is a real pain in the ass, and worse is that a permanent solution can be pot luck, with you having to return the card, swap your motherboard or the PSU, or even all three in the hopes of getting a working combination. This problem doesn't affect everyone (probably less than 1%), and isn't limited to the AIW line up exclusively or even limited to specific systems, but if you are in this situation and are comfortable with a bit of minor wiring then the switch workaround will do the trick.
I would have preferred to use a delaying circuit negating the need for the switch but I'm impatient and wanted to play with my new toy, so the switch fix was a good second choice I was able to implement immediately. The Radeon 9700 Pro All-In-Wonder, now that it's running, is a great card and I'm very happy with it's performance and features, but I am pretty disappointed that I had to spend 4 hours trawling the net to find information on such an irritating problem, not to mention the bit of DIY just to get it running at the hardware level.
The folks have had a lot of discussion and testing going on in regards this issue, which lead to the above solution and coupled with my own experience of this I felt it deserved to be put into a summary. Hopefully their research will help others, and hopefully will come up with a permanent solution to this problem. If you have any questions or comments, be sure to hit us up in the forums.