Video RAM bad, or something else?

[Septfox]

Sorry, wall of text in here. Just feel that detailed information might be helpful.
First off, this machine is running a R9600.

So, I was gaming the other day (a new game, infact), and suddenly...garbled screen/ripped polygons everywhere.

In the past, this was a very rare occurrence that happened happened after long-term gaming. It actually hadn't happened in a LONG time before this incident.

Anyway, since I have Ati Tray Tools running, I also have the Reset Display Driver function hotkeyed. My response to this problem was to use that; and it fixed the display, as it always had for others.
About five minutes later, it happened again. Again, I reset the driver, and it was fine for another 20 seconds or so before happening again. After this, I decided to close the game and restart the laptop.

lo and behold, my eyes were met with a vertically-striped bios screen, followed by green squares everywhere (pre-Windows cursor blink), followed by a vertically-striped bootscreen. In Windows, oddly rectangle-checkered patterning everywhere.

At this point, I'm thinking overheat, despite nothing ever actually overheating (you know how it is; gaming-hot, but not melt-transistors-hot). So, I decide to shut it down for the night.

The next day, I switch it back on. Clear bright screen, no problems whatsoever, once in Windows it ran fine, ran a couple games perfectly well with no difficulty, etc. Then I made the mistake of running the game that had started it all again...halfway through loading, patterning is back, polygons in the load screen are ripping all over. Shut it down, and after messing with some things...clear screen again.

I started getting garbling off and on after that, usually during web browsing. I found that switching NHC's Powerplay control between Performance and Mobile sometimes "froze" the garbling and, following a video driver reset, I would have a clear screen again. This little cycle became less and less effective until it got to this point; stuck.

So, the current status is:

• Consistent display corruption. While the patterning does change depending on the display mode and resolution (BIOS screen, pre-windows, bootscreen, windows w/proper drivers installed, windows w/o proper drivers installed, etc), for each of these situations the display corruption is the same. The only thing that varies is the colors.
• No SMARTGART support. On installing Omega drivers "properly" and following a reboot/relogin, I get a black screen and after 10 seconds, a bluescreen from the video driver. On reboot, I only have very slow DirectDraw support; no D3D.
• Reinstalling the drivers "improperly" (from the device page) gets me back support for everything, but no FastWrites and unbearably ugly patterning.
• Different patterning depending on whether the Display>Troubleshoot>H/W Acceleration slider is on the left or right.
• "Smearing" of the contents of moved windows unless H/W acceleration is completely disabled.
• Generally low performance everywhere unless H/W acceleration is completely disabled.
• Flickering pixels in various areas of the screen whenever something is moving (think a slightly unstable overclock effect).

Striping of "broken" drivers (no D3D), 1024x768: http://i1192.photobucket.com/albums/aa3 ... oard02.png
Switching to 1280x768 turns the striping into a much more pleasant/manageable fullscreen speckling.
Any resolution with "working" drivers has an ugly checkerboard effect.


So, given all this, any ideas? Is it simply bad video RAM? I would think that failing memory wouldn't give such consistent results...was also thinking that, somehow amidst all the display driver resets, something in the BIOS or video BIOS might have (somehow) gotten screwed, but that doesn't account for the random clear screens I'd gotten earlier. Nor can I actually try replacing other of those, since I have the newest BIOS revision and to my knowledge, there is no upgrade for the video BIOS (along with the chip apparently being read-only).
I've also run VMT; it spits out errors like nobody's business, but it still doesn't *feel* like it's a problem with the actual memory...*shrugs* any help/advice is appreciated.

[goofyGAguy]

Your machine has fallen victim to the GPU becoming unsoldered from the motherboard. Sorry.

[poshgeordie]

Sorry, wall of text in here. Just feel that detailed information might be helpful.
First off, this machine is running a R9600.

So, I was gaming the other day (a new game, infact), and suddenly...garbled screen/ripped polygons everywhere

Hi

9600 and FireGL GPU's do suffer from from broken artifacts and I have traced that to damaged VRAM chips. Unfortunately it's next to impossible to repair, since the first challenge is to identify which of the 4 chips it is (no idead how to do that!) and to then try and remove it. Each chip is stuck to the GPU planar with black gung that is not easy to remove without damage to nearby tiny surface mount components.
Only solution is to replace the GPU. Note that 9600 (with 64MB on board RAM) and FireGL (128MB) are interchangeable BTW, but it needs a specialist reflowing company to do it because they aren't so straight forward to do.

[poshgeordie]

Update to my previous post - just today I had a 7500 chip with the speckled picture and again it was a VRAM.

[Septfox]

Thanks for the replies, guys.

Your machine has fallen victim to the GPU becoming unsoldered from the motherboard. Sorry.

I'd already considered that. All my physical testing (applying pressure to the GPU and various points on the planar board during running and startup, making sure there was adequate pressure on the heatsink, slightly flexing the chassis) turned up nothing. My GPU has never had problems indicative of becoming unattached, and when it DID start having problems, it was stable sitting on my lap; no flexing stress at all.
Still though, I'm very much considering baking the board, just to make sure. Not like it can really hurt anything at this point.

Hi

9600 and FireGL GPU's do suffer from from broken artifacts and I have traced that to damaged VRAM chips. Unfortunately it's next to impossible to repair, since the first challenge is to identify which of the 4 chips it is (no idead how to do that!) and to then try and remove it.

Yea, but...
If you're referring to physical/shock damage, I've ALWAYS been careful with my laptop. I don't see how it could have possibly happened, and as I said above, when this started happening, it was sitting on my lap.
As for heat damage...well, let me put it this way: when all this started, I was playing a relatively light-graphics MMORPG (Lunia). The laptop was cooling properly, to my knowledge; actually technically overcooling, since I have NHC installed and its ACPI config set to kick fans on sooner, to keep the CPU/GPU, and by proxy my lap, cooler. On the flip side, I've played TES:Oblivion on this thing, without the ACPI control installed, with no overheating (though it did run quite warm, it wasn't runaway, leveled around 180F max IIRC). Dunno if you've heard of or played Oblivion, but it's VERY hard on everything in the system, and especially on the technically-underpowered 9600. If it was going to fail due to heat damage, I'd think it would have done it long before now.

I'd like to note one thing, though: during vmt's testing, it found errors throughout memory. One would assume that if it were one or two chips gone bad, at least SOME of it would test ok, right...?
At this point, I'm thinking memory controller, or perhaps still something, somehow, skewed in the video BIOS (though, to my knowledge the Mobility's BIOS is read-only *shrugs*).

Which still means either a planar board replacement or new GPU+reflow....*sigh*

How's the FireGL, anyway? Worth looking for one over another 9600? They seem to be pretty rare.
Edit: from my research, it looks like the FireGL is basically a 9600 with higher fillrates, 1.7gb/s more memory bandwidth (compared to the 64mb 9600), double actual memory, and 20mhz higher CPU clock/100mhz lower memory clocks standard. Definitely think I'll look into finding one, hmmm.