Weird Issue

[morfinx]

I posted this under OC section, but thought I might have better luck here:

Just upgraded an Athlon XP 1700+ to a Winnine 3000+, and I'm having a weird problem. Whenever I do anything graphical, i.e. games, video, or even web sites with alot of pictures, the monitor loses signal from the vid card and goes blank. I would then have to restart. A couple of times, VPU Recovery (part of Catalyst driver set) recovers, and says video card stopped responding to driver commands. At first I thought it was the OC, so I lowered everything back down to stock, and still have the same problem. Here is my setup:

A64 3000+ Winchester
Zalman CNPS-7000B AlCu
Soltek SL-K8T800-Pro
Mushkin Black L II V2 (TCCD) 512MB x 2
Enlight 420W PSU
Gigabyte 9600XT (stock)
Dell 2005FPW (DVI)

The strange thing is, I can run Prime95 all day long without error, but as soon as I start doing anything graphical, the screen will blank out within minutes. The card never gave me any issues in my Athlon XP system for over a yr. Voltages look stable running Prime95, I get 3.3V, 5.0V, and 11.71V fairly stable. I've tried the latest Catalyst drivers and an older one, still no go. Could it be the PSU? Any help or suggestion would be appreciated!

 

[Ignad]

1. Check ALL connections. Something stupid might be slightly loose and causing odd behavior.

2. Check to make sure that the cooling fan on the GPU is running ok... you might have wiped it out while pulling it out to upgrade the cpu, (e.g. put pressure on the fan when removing or inserting the card) or simply unplugged the GPU fan, etc. Make sure you didnt somehow disturb the heatsink on the gpu and its slightly askew. (longshot, but worth the two seconds to check)

3. Check to make sure that the vid card is 100% straight in the slot in all directions(side to side and end to end, etc). AGP has a LOT of connections and the slightest mis-seating in the slot can cause all kinds of oddness.

4. To see if its the power supply, you could try underclocking the cpu, and even try unplugging a few non-essential items (extra case fan, optical drive(s), etc) to lighten the load and see if that helps.

5. Do you have another vid card to swap in to test temporarily? Maybe your vid card decided to go insane.

6. Try your vid card in another system to see if its stable.

7. There is a chance your new cpu is damaged... or that you didnt apply thermal paste to it correctly, or its heatsink is not seated 100% correctly... or you slightly zapped one of your components with static charge while working on the system...


let us know how it goes... i respond to way too many of these help threads to which the OP never notes the solution

 

[morfinx]

1. Check ALL connections. Something stupid might be slightly loose and causing odd behavior.

2. Check to make sure that the cooling fan on the GPU is running ok... you might have wiped it out while pulling it out to upgrade the cpu, (e.g. put pressure on the fan when removing or inserting the card) or simply unplugged the GPU fan, etc. Make sure you didnt somehow disturb the heatsink on the gpu and its slightly askew. (longshot, but worth the two seconds to check)

3. Check to make sure that the vid card is 100% straight in the slot in all directions(side to side and end to end, etc). AGP has a LOT of connections and the slightest mis-seating in the slot can cause all kinds of oddness.

4. To see if its the power supply, you could try underclocking the cpu, and even try unplugging a few non-essential items (extra case fan, optical drive(s), etc) to lighten the load and see if that helps.

5. Do you have another vid card to swap in to test temporarily? Maybe your vid card decided to go insane.

6. Try your vid card in another system to see if its stable.

7. There is a chance your new cpu is damaged... or that you didnt apply thermal paste to it correctly, or its heatsink is not seated 100% correctly... or you slightly zapped one of your components with static charge while working on the system...


let us know how it goes... i respond to way too many of these help threads to which the OP never notes the solution

1. Everything seems to be in order
2. I have two 80mm fans blowing across the GPU, so even if the fan's not working, there should still be sufficient cooling. ATI's driver panel (Overdrive) reports 35C GPU temp.
3. I took the card out and reseated it. The mobo has a locking tab on the AGP slot, similar to the locking tabs on DIMM slots. So the card would have to be properly seated for the tab to be completely locked in.
4. I might give that a try
5. I don't have another card, but I will see if I can find one.
6. I will try doing that
7. The computer runs Prime95 for hrs without error. I would think that means the CPU is alright. In fact, forum browsing is fine, but if I run anything with graphics, monitor will show no signal within a few mins.

I'll post the findings when I find the solution. In the mean time, any more suggestions would be most appreciated.

 

[UncleMacro]

This isn't exactly a smoking gun or anything, but the last couple of people I've run into with your symptoms have turned out to have power supply problems. You should definitely try Ignad's #4 to try to reduce the power consumption and see if the problem stops or the average time to "lights out" lengthens. If you underclock both your CPU and video board enough and unplug any optical drives, you should be able to reduce the load on the 12 volt rail quite a bit.

 

[morfinx]

This isn't exactly a smoking gun or anything, but the last couple of people I've run into with your symptoms have turned out to have power supply problems. You should definitely try Ignad's #4 to try to reduce the power consumption and see if the problem stops or the average time to "lights out" lengthens. If you underclock both your CPU and video board enough and unplug any optical drives, you should be able to reduce the load on the 12 volt rail quite a bit.

I will try that. And as an interesting side, I just ran my TV Tuner program (DScaler) for 45 mins with no problem. As I understand, the program utilitzes DirectDraw overlay and it puts the CPU at 100% utilization. I'm puzzled.

 

[morfinx]

Ok I unplugged two 80mm fans, disconnected my optical drives, and then drop all the voltages down to stock levels, and clocked the CPU down to 1Ghz, started Windows up, and still have the same problem. Any ideas?

 

[UncleMacro]

Well, you've probably already tried these but: the only things I can see which are not on the lists above would be underclocking the video board a lot with something like ATITool. You can underclock the GPU down to next to nothing and underclock your video RAM quite a bit before it gets into trouble. The other thing is using SmartGART to fiddle with your AGP slot. Disable fast writes (I'm not sure that's allowed on a 9600XT) and slow down the AGP speed to 4X and disabled to see what happens. AGP problems do not match your symptoms but it's easy enough to try it.

 

[morfinx]

Well, you've probably already tried these but: the only things I can see which are not on the lists above would be underclocking the video board a lot with something like ATITool. You can underclock the GPU down to next to nothing and underclock your video RAM quite a bit before it gets into trouble. The other thing is using SmartGART to fiddle with your AGP slot. Disable fast writes (I'm not sure that's allowed on a 9600XT) and slow down the AGP speed to 4X and disabled to see what happens. AGP problems do not match your symptoms but it's easy enough to try it.

By the beard of Zeus! I should call you uncle! I set AGP speed to 4x and for the first time ever the entire 3DMark2005 demo played without blanking out! Normally it blanks out within 2 mins. Wow, I'm dumbfounded. I will do more testing to make sure this fixes it. What could have caused this?

 

[UncleMacro]

Well borrowing from another one of my posts:

The reason that full speed AGP and fast writes tend to be flakey sometimes is basically because they are transfering data like a bat-outta-hell. They are transfering data on a 66MHz clock, octuple pumped, 32 bits wide, using a non-differential 0.8V swing, usually over a four layer board (if you don't know hardware then just think "Be Afraid. BE VERY AFRAID!"). Basically, that's pushing the envelope, electronically speaking, if you're talking about low-budget consumer hardware where they're trying to shave every nickel off of their bill of materials. If you've ever seen high-speed scope traces of that data, it ain't exactly perfect ones and zeroes. If you're board is laid out and decoupled properly, and the CPU, northbridge, and GPU are all nice and cool, and the power supply is keeping the rails relatively smooth, then it probably works just fine. But if anything one of those links in the chain is a little weak, then you've got problems.

Which basically just means AGP 8X tends to be flakey. My machine, for example, will crash about once an hour if I run it at AGP 8X but runs perfectly at AGP 4X. I've never heard of AGP corruption shutting down the monitor output before but it commonly causes crashes and hangs. Personally, I don't see any reason to run AGP at 8X anyway. It doesn't gain you any significant performance over AGP 4X but it definitely requires everything to be working perfectly to be completely reliable. If it really fixes your problem, then consider yourself fortunate.

 

[morfinx]

I think that fixed it. Running 2.4Ghz@1.5V, Prime95 stable for 14 hrs, 3DMark stable for 4.5 hrs. Thanks for the help UncleMacro.