Flakey Rampage 2 Extreme

Hey guys, I finally got my new system together and running (see "monolith" in sig), so far I've been able to stabilize the cpu/ram/uncore at 4.2ghz/1600mhz/3.2ghz, several hours of LinX stable, decent volts/temps.

I've been having strange issues with graphics-related crashes and "display driver failures" I found that when I scrubbed video around in VLC or played minecraft (not a graphic-intense game if you know it) for a while, I would get graphic artifacts around windows, followed by a graphics driver failure and sometimes a hard lock or instant reset. Needless to say I've been perplexed.

I ran furmark for an hour with no problems, so I figured it wasn't the gfx card itself, most likely something on the mobo.
At first I thought maybe it was a clock stability issue, so i started downclocking and upping volts across the board. No change, though some settings seemed worse than others... as my Rampage 2 Extreme has been kinda funny with voltages and sensors (half the time the windows voltage sensors don't work or display maximum values). I tried giving the PCIE busses more volts and such, no luck. I went to stock settings, found the same issues. Tried re-installing graphics driver, going back to an older driver, forcing 3d clocks in 2d applications, and other software tweaks.

Finally today I threw in my older HD3850, set CPU/RAM to stock and tried some video-scrubbing. Hard reset.

It must be the motherboard or power supply. I'm leaning towards motherboard, as a PSU issue would likely have come out during some of the hours of LinX or Furmark testing.
I can RMA the board and wait around for it to come back, or just buy a brand new board and re-sell the resulting RMA-ed board.
Unless someone here knows some secret setting on the R2E for "normal graphics function", I'm leaning towards getting a Rampage 3 Formula, or telling ASUS boards to f*** off completely (I've had mixed luck with them).
Gigabyte? EVGA? MSI?
What does I do?

1st Point...every time you run FURMARK you damage your card more and more, FURMARK is the worst stability test...so dont use it, rather use Crysis benchmark, run 20 loops and if it passes it will be fine.

What GPU drivers are you using, did you remove the drivers and re-install when you swapped video cards?

What are your voltages across the bios?

I have had several RIIE and they are amongst the most stable and reliable boards.

Why would furmark damage the card? because it puts a heavy load on the hardware and makes it warm? I never broke 80c. Does it use some kind of code that stresses the hardware beyond spec?

I started with 10.12 drivers, cleaned them out and installed 10.11. Then when I swapped the 3850 in I re-installed 10.11.

for 4.2ghz linx stable, voltages were
1.375 vcore (+load calibration, goes up to 1.41) (have tried up to 1.385)
1.36 VTT (have tried up to 1.42)
1.13 IOH (which translated to 1.11 actual) (have tried up to 1.17, higher voltages seemed to cause the graphic corruption to occur faster)
DRAM at 1.51 (have tried up to 1.55)
everything else on Auto, though I tried raising all the other voltages (PCIE, IOH, etc) up to 2 steps above stock one at a time, with no different results.
speedstep, c1e, spread spectrum disabled

For stock speeds I put everything on Auto except DRAM at 1.5

With all of these voltage settings, the Asus PCProbe2 sensor program would sometimes display normal voltages, and other times just show maximum values or really low ones, at first it seemed like certain voltage combos would cause it to act normal vs freak out, but multiple boots at the same settings would sometimes be trouble and other times not.

Brolloks said:

1st Point...every time you run FURMARK you damage your card more and more, FURMARK is the worst stability test...so dont use it, rather use Crysis benchmark, run 20 loops and if it passes it will be fine.

Click to expand...

I'm a little curious about this as well. I trust you and believe what you're saying, just personally curious.

Okay so today when I got home from work and tried firing up the system the problem had gotten even worse, display driver failures just after reaching windows desktop...
I booted into safe mode successfully and tried un-installing the drivers from there but of course installers don't work in safe mode.
So with little other choice I ran driver sweeper in safe mode and essentially gutted the driver. Booted back into regular windows mode and besides the display looking low-res, no crashes! Ran the normal driver un-install routines and played around for a while with no problems.
When I tried re-installing the 10.12 driver I had some issues, like it would appear to install, but wouldn't actually become active after reboot.
I cleaned it all out again and installed 10.11, booted to desktop with no problems, played VLC, minecraft, did some 3d rendering...no problems for 2 hours!

I'm just amazed at the trouble these ATI drivers seem to be causing...I was convinced my mobo was f*ked! (further testing is needed, but so far it looks okay)
Just now I set the system back to 4.2ghz, let's see if it behaves now. Next step after that if everything seems okay is to put the 5870 back in...and re-install the drivers again

Ran for an hour or so at 4.2ghz last night, threw the same load that was causing problems on it, no issues...when I get home the 5870's going back in.
Cautiously optimistic about not having to RMA/get a new board now...
I also installed an alternate voltage monitoring program and I'm not seeing the weird fluctuation I got with the ASUS software. Maybe their sensor program just sucks?

this is turning into a "hey i fixed my own problem" thread...
i'm a little disappointed with the lack of response, I guess everyone's too busy with "omg which P67 board does I get" threads

Cleaned out the drivers, put in the 5870, re-installed the drivers...got graphic artifacting when I loaded up minecraft, but no crash. Quit minecraft, artifacts went away. Used ATI TrayTools to put a load on the card (artifact scanner), and had no issues, though my PSU was squealing a little, but it's done that before, and I checked all my volrages with a multimeter and they're all well within spec under load
Forced 3D clocks with Tray Tools... no artifacts. I had tried this before, but it still had problems, perhaps the driver was already corrupted at that point. Anyway, it would appear that my 5870 doesn't play well in 2D mode, or at least with minecraft? I've been playing for an hour while doing a 3d render in the background and no artifacts or crash...
I guess the culprit is the 5870, or the drivers...but I seem to have a solution for now (force 3d clocks).
Conclusion thus far: Rampage 2 Extreme is not the problem.

Check your ram. I have seen bad ram put little red dots all over screens. Like very little the size of pixels all over screens.

Bad ram can make other things look bad.

I've been running fine with forced 3D clocks for hours now, doing resource-heavy stuff. It must be the 5870. I thought maybe the PCIE link was bad or needed more voltage or something, but it appears to be the 2d clock mode's fault.
The ram is fine, it would have been causing problems with the linx testing, which is stable for hours.

not out of the woods yet...just got a hard reset... not sure what triggered it
edit: now it's crashing every time i try to play a flash video in a browser, with 3d clocks applied...I think the driver is corrupted again somehow?

was having problems at 2d and 3d clocks today, so i cleaned out the drivers again, tried installing 10.12 again, but it wouldn't actually install the display driver part (again), so I'm back on 10.11 and things seem okay at 2d and 3d clocks. This is so weird. First I thought it was the board, then the gfx card, now it looks like it was the drivers all along, but how did they get corrupted in the first place???
edit: and i just got a driver error again. what the hell is going on???

It really sounds the memory subsystem to me. Either you have a bad stick of ram, or your IMC could be unstable in some way. One of my friends recently did a build with an HD6850 with issues very similar to yours; BSODs whenever he'd open youtube, driver errors popping up frequently, the system seeming to run fine with the drivers not installed, etc. I thought it had to be a bad video card, but instead it was a single bad stick of memory. LinX can often overlook them.

I had similar problem with an ASUS M4A78T-E, my 965BE, and a XFX 4830. I at first thought it was driver problems and tried similar things to you, but then I read a little article about switching between 2D and 3D mode can cause problems if your GDDR speed goes from one speed to another. I used RBE (Radeon BIOS Edit) to look at the different settings for UMA, 3D, Boot, 2D, and idle mode. The memory was changing when going to between 2D, 3D, and idle modes. So, I applied my best stable core O/C to 3D, then a reduced core for 2D, and a severely reduced core for idle (boot should be the same as 3D according to this RBE article) and correct voltages for each. Then I just put the same memory freq for all of them, changed the fan to go a little faster and flashed it. I stressed it a few time (I got paranoid it wouldn't work and ran furmark for a good 3 hours, temps stayed around 50c). The card worked great after that! I think it is because Windows 7 causes non-focused windows to lose driver control priority and the desktop usually switches the card into 2D or idle mode, while minecraft (I play it too ) and other games use 3D mode. So when you are switching focus back and forth you are causing the card to switch its memory speed a lot, causing instability. Trying to make memory change freq while it is responding to requests can cause it to corrupt transactions or memory blocks causing problems of different sorts. Is your card factory oc'd? Mine was, and it seems like factory oc'd cards will have the different mem settings.

I think RBE works for 58X0's but not sure... maybe you should give it a try. Make sure you save a backup of your original first though!

thanks for the responses guys,

gautam, what's the best way to determine if the memory/imc is the issue? I tried another round of linx testing and actually got a reset after an hour, so I tried raising my qpi voltage and actually got errors faster that way. Now testing with lower qpi volts...
I may need to start my overclocking routine from scratch with more rigorous testing at each interval, as I determined stability at each 100mhz inteval on 30 minute Linx runs, followed by an hour at 4.2ghz before and it may have not been enough.

dooms101, good information! no my card isn't factory oc'ed. I thought it might be a 2d/3d switching issue too, but it seems that I can get crash/resets after running at forced 3d clocks for a while too, so it's not switching anymore but still having trouble.

It looks like some combination of cpu/qpi/ram instability, causing resets and driver corruption which cause gpu issues right now...ugh what a headache.

Today I booted into safe mode (to rule out driver interference) and did 3+ hours of linx testing at full memory usage, no errors or resets. As far as I'm concerned my cpu/ram/uncore are stable. I did end up having to bump my vcore a little. But after that, no issues for hours.

Just now, while in normal windows mode, playing minecraft, and setting up a 3d render in Maya, my screen went grey and recovered with a driver crash message. Either my 5870 is flakey or somehow the driver is failing. I'm really puzzled, because I can run the ati tray tools artifact scanner with no problems.

You might want to try a reinstall of windows on a separate drive. Windows and Windows installation software can leave a lot of things behind that will interfere with new/different drivers.

Windows re-installation will be my next step.
I just downloaded gpu-z and ran its monitoring tab in the background while doing some 3d work in Maya, and noticed the clocks jumping up and down (2d/3d). This may have been the source of today's error. I forced 3d clocks again, and am testing minecraft/maya to see if i get another error. If I do, then it's fresh install time.

From what I've experienced, core clocks going up and down don't really effect it too much. What about you're voltages? They might be set too low.

Problems continue, but after doing some reading on the AMD support forums, it's pretty clear it's a problem with the drivers or the card, not my motherboard/ram/cpu.
I'm going to start a thread in the ati section.

Edit: I just found out they released new drivers, I installed them and am testing now, so far so good, much less clock-switching, almost none in fact, which seems to be helping.
Edit2: lol, foolish of me to think it would help... it just froze again while zooming in on a photoshop document *sigh*