GPU has problems [SLOVED] originaly: IMC is dying...

[buildzoid]

Basically no matter what I do my 3960X won't boot windows(I can get to the BIOS) and I always get a RAM related error like IRQL_NOT_LESS_OR_EQUAL or FAULT_IN_UNPAGE_AREA. The POST code display on the board doesn't work half the time the debug LEDs always show a ram or vga error. I've already tested 2 different sticks of ram in every memory channel with varying lengths time separating me from the inevitable crash. The only thing I can think off that could have done this is that I ran my ram at above 1.65V(1.685V) for about 5 months. Though recently I was forced(helped stability) into running my VTT and VCCSA/VCCIO at 1.2V and 1.25V. From all these things I've deduced that the IMC is hovering somewhere between life and death. Does anyone know a way I could keep the thing running for 4 months?
I know for sure that I don't have a VGA error.

 

[EarthDog]

Not sure if its the IMC, but try clocking the ram speed down would be the most logical thing to me.

 

[buildzoid]

Ok I just set 1333 11-11-11-33 and now I'm stuck at the windows logo pulsing.
1066 9-9-9-27 windows seems to be working. Nope that crashed once I got to the desktop.
Now it won't boot at all I give up. Goddamn and such a good chip too. Where am I gonna get a 4.8+Ghz Ivybridge-E. Well atleast I still got the Athlon x2 370K to play some games on.

 

[cadaveca]

I'd try fresh OS, and a full CMOS clear. After clear, re-flash BIOS if possible.

 

[buildzoid]

I'd try fresh OS, and a full CMOS clear. After clear, re-flash BIOS if possible.

So you think it's still alive?

 

[cadaveca]

So you think it's still alive?

Yep, it is possible. I would give it a try, anyway, not much else you can do that doesn't involve spending money directly.

Corrupt driver or OS, or BIOS/MEI can cause this. Also, having different Intel MEI driver compared to BIOS MEI version can cause similar as well.

CPU degraded is definitely possible too. Fresh OS, BIOS, drivers, etc can ensure that the problem is related to hardware.

 

[buildzoid]

Yep, it is possible. I would give it a try, anyway, not much else you can do that doesn't involve spending money directly.

Yeah I think that so far every chip that stopped working for me stopped working after a ram OC and so I'm just following the trend of probably dead IMCs since all the ram sticks are still good.
My mac is being really weird about unzipping the ASUS BIOS zip files. They all end up as .CAP instead of .ROM

 

[cadaveca]

Yeah I think that so far every chip that stopped working for me stopped working after a ram OC and so I'm just following the trend of probably dead IMCs since all the ram sticks are still good.
My mac is being really weird about unzipping the ASUS BIOS zip files. They all end up as .CAP instead of .ROM

That's normal - .cap instead of .rom. That's part of the USB BIOS flashback function and UEFI. Toss the .cap on a FAT-formatted drive, flash from within BIOS. BIOS tool checks BIOS before flashing, as far as I know, to prevent flashing improper BIOS.

 

[buildzoid]

That's normal - .cap instead of .rom. That's part of the USB BIOS flashback function and UEFI. Toss the .cap on a FAT-formatted drive, flash from within BIOS. BIOS tool checks BIOS before flashing, as far as I know, to prevent flashing improper BIOS.

Yeah except for the part where my BIOS doesn't recognize any of the files other than the hidden files on the flash stick. Files like .Trash and .fsevents.

Just to check that it really wasn't the second GPU I decided to turn of the PCI-e lane in which my R9 290 with a modded cooler is located and now everything works again so I guess the ram on my second GPU was throwing the errors.Which is odd because when I ran the R9 290 solo it was fine. So tomorrow I'll have to test the R9 290 on it's own to see if my CLU spill killed it or not. It might also be that after so many disassembles of the cooler some chip somewhere got pulled on too hard by the T pads and lost contact. Either way a potentially dead R9 290 is a major improvement over a dead 5.3Ghz(1.725V for about 3 minutes) i7 3960X. Thanks for all the help.
I'm still gonna try to flash the boards BIOS as that might fix the GPU issues I keep having when running more than one card.

 

[cadaveca]

I have issues with running multiple cards with my 3960X, too. Some slots on some boards just don't work at all, and since I have three 780 ti...man...I need all slots to work!

Now I have 4960X CPU, VGA stuff works fine, but memory OC is wide-open, and I keep wanting more, so I'm constantly tweaking trying to get more memory speed...even though it matters not for what I use the PC for, really.

Anyway, point is, try other slots on your board for that VGA if things don't work right, maybe, once you've sorted the other stuff.

 

[buildzoid]

I have issues with running multiple cards with my 3960X, too. Some slots on some boards just don't work at all, and since I have three 780 ti...man...I need all slots to work!

Now I have 4960X CPU, VGA stuff works fine, but memory OC is wide-open, and I keep wanting more, so I'm constantly tweaking trying to get more memory speed...even though it matters not for what I use the PC for, really.

Anyway, point is, try other slots on your board for that VGA if things don't work right, maybe, once you've sorted the other stuff.

That could be related to the forced PCI-e 3.0 that the Rampages use because the 3960X doesn't support that but the 4960X does which would explain why the 3960X is all weird about multi GPU but the 4960X works fine. I think there is a setting in the BIOS for disabling 3.0 so I'll try that. As for tweaking ram I managed to get that 4x4 2133 9-11-11-31 RipjawsX kit you reviewed to 2448mhz 10-12-12-34 on 1.75V with loosened secondary timings.

 

[cadaveca]

Yeah, I tried those settings, and in my case, it doesn't work.

2448 is REALLY GOOD for SB-E. Many chips won't even hit 2400, no matter what ram you use.

 

[TheHunter]

Yep, it is possible. I would give it a try, anyway, not much else you can do that doesn't involve spending money directly.

Corrupt driver or OS, or BIOS/MEI can cause this. Also, having different Intel MEI driver compared to BIOS MEI version can cause similar as well.

CPU degraded is definitely possible too. Fresh OS, BIOS, drivers, etc can ensure that the problem is related to hardware.

I have bios IME 9.0.30, while in OS 10.0.1204.. I use to have 9.5.27..


And I noticed this newer v10 is kinda weird, had to put my ram current back to 120% from 110% (digi+) and Cache from 1.135 to 1.137v or it could sometimes have issues by sleep (failed to boot and rebooted).

Also 1-2 games started to crash while before it was fine.. Now after readjusting those 2 settings I got better perf. vs before at same settings, cinebench15 extra 2-4 points and by 3dmark physics scores up to 150-200points more, but yeah its heats a little more vs 110%


So is it better to use lower then or just stick with current 10.0.0.1204?

 

[broken pixel]

RMA the CPU, I did when my first 3930k IMC got sick. I ended up with an even better 3930k that runs 4.7GHz/2133MHz @ 1.32v. Intel has fast RMAs.

 

[buildzoid]

RMA the CPU, I did when my first 3930k IMC got sick. I ended up with an even better 3930k that runs 4.7GHz/2133MHz @ 1.32v. Intel has fast RMAs.

It's 3 years old and actually it was just whining about running my secondary GPU but since the post code on the board doesn't work I assumed RAM issues but now with the PCI-e slots set to run at x16 2.0 instead of x16 3.0 stuff seems to be working.

 

[buildzoid]

Ok so I got the R9 290 to run on it's own after completely removing all the CLU and since I was working on the core I got the stuff everywhere and so I gave the core and it's surroundings a shower with distilled water. The CLU didn't all come of some of it congealed when I was removing it with alcohol and got stuck to the core. The heatsink is a mess since the stuff became one with the copper. The GPU boots when I run it alone but I think there might still be a few drops of distilled water somewhere as I was getting some artifacts but it works and should be perfectly OK by tomorrow.
BTW I'm never gonna use CLU on a GPU again without covering all the caps ICs and resistors surrounding the core with liquid electrical tape.

 

[Random Murderer]

Ok so I got the R9 290 to run on it's own after completely removing all the CLU and since I was working on the core I got the stuff everywhere and so I gave the core and it's surroundings a shower with distilled water. The CLU didn't all come of some of it congealed when I was removing it with alcohol and got stuck to the core. The heatsink is a mess since the stuff became one with the copper. The GPU boots when I run it alone but I think there might still be a few drops of distilled water somewhere as I was getting some artifacts but it works and should be perfectly OK by tomorrow.
BTW I'm never gonna use CLU on a GPU again without covering all the caps ICs and resistors surrounding the core with liquid electrical tape.

It sounds like you have it mostly figured out, but I did want to pop in since we have/had similar setups.
When I was on my 3820 (which I've decided to keep, it has a champ IMC, runs 2520) and 3x 7970s, I had similar issues: the odd no-POST with a RAM/IMC error code, or really slow loading of Windows, the odd BSODs that point to RAM... yet I could run MemTest for 24 hours straight, not a single error. Finally it threw an AMD driver related BSOD, and I talked to Dave about it. Ended up setting the PCIe lanes to Gen 2.0 in BIOS, and not only did it solve the problems, I was able to lower several voltages (VCCSA, VTT, VDIMM) and remain stable at 4.9 on the CPU, 4x4GB 2400 C10 on the RAM.

Another thing I thought I'd mention: If your BIOS isn't recognizing the new .CAP files on your USB stick, chances are you're on an old .ROM and need to use the CAP converter tool before flashing to a .CAP. What BIOS is your RIVE currently using?

 

[buildzoid]

It sounds like you have it mostly figured out, but I did want to pop in since we have/had similar setups.
When I was on my 3820 (which I've decided to keep, it has a champ IMC, runs 2520) and 3x 7970s, I had similar issues: the odd no-POST with a RAM/IMC error code, or really slow loading of Windows, the odd BSODs that point to RAM... yet I could run MemTest for 24 hours straight, not a single error. Finally it threw an AMD driver related BSOD, and I talked to Dave about it. Ended up setting the PCIe lanes to Gen 2.0 in BIOS, and not only did it solve the problems, I was able to lower several voltages (VCCSA, VTT, VDIMM) and remain stable at 4.9 on the CPU, 4x4GB 2400 C10 on the RAM.

Another thing I thought I'd mention: If your BIOS isn't recognizing the new .CAP files on your USB stick, chances are you're on an old .ROM and need to use the CAP converter tool before flashing to a .CAP. What BIOS is your RIVE currently using?

I'm on the very first RIVE BIOS. I think its 0703 or something like that. Where can I find this CAP converter?

 

[Random Murderer]

I'm on the very first RIVE BIOS. I think its 0703 or something like that. Where can I find this CAP converter?

Oh shit, you need to update that BIOS, lol. Do you know how many BIOS revisions have been released since then that do nothing but improve stability? A lot.
The CAP converter can be found in the same section on Asus' site where you download the actual BIOSes. Let me see if I can dig up a link for you.
EDIT: because the site is so Java-heavy, can't give you a link that'll bring you directly to it, but the CAP converter is in the BIOS-Utilities section. It will update both BIOS chips to 2105 and convert them to .CAP structure.

 

[buildzoid]

Oh shit, you need to update that BIOS, lol. Do you know how many BIOS revisions have been released since then that do nothing but improve stability? A lot.
The CAP converter can be found in the same section on Asus' site where you download the actual BIOSes. Let me see if I can dig up a link for you.

I know how many there have been but I approach all updates with: If it works then it doesn't need an update. And up until I got 2 GPUs everything worked.
Found the converter.

 

[Random Murderer]

I know how many there have been but I approach all updates with: If it works then it doesn't need an update. And up until I got 2 GPUs everything worked.

I generally have the same thinking when it comes to BIOSes, but with the fact that mine came with the same BIOS as yours (0703, which can't even be downloaded from Asus anymore) and I'm now on an IVB-E, I had to update. If you read through the notes for most of the earlier BIOSes, they are improvements that anybody will benefit from. A lot of "improves system stability," "improves memory compatibility," "enhances OC capability," and "enhance PCIe 3.0 support capability."

 

[buildzoid]

Yeah well I'll update this weekend today it's too late.

 

[Vario]

Ok so I got the R9 290 to run on it's own after completely removing all the CLU and since I was working on the core I got the stuff everywhere and so I gave the core and it's surroundings a shower with distilled water. The CLU didn't all come of some of it congealed when I was removing it with alcohol and got stuck to the core. The heatsink is a mess since the stuff became one with the copper. The GPU boots when I run it alone but I think there might still be a few drops of distilled water somewhere as I was getting some artifacts but it works and should be perfectly OK by tomorrow.
BTW I'm never gonna use CLU on a GPU again without covering all the caps ICs and resistors surrounding the core with liquid electrical tape.

This killed my h100i's finish. I got trolled when I mentioned that on here twice now. Gonna mention it again. CLU sucks.

Also my 3770k (used, binned) didn't run my hyper X very well at XMP 1600, constant IRQL errors, and I didn't want to raise the voltage or vtt, so I had to reduce to 1333 default profile. The prior i7 3770k which I had bought new had no issues. Might be that the IMC on Intel's are weak with prolonged ram overclock or maybe just coincidence in my case. I am assuming the prior owner ran ram overclocked and overvolted as he ran a watercooled performance setup. Hopefully running my memory stock will keep any further degradation to a minimum. I haven't had any blue screens in the two months since I set it to 1333.

Glad PCI-E mode fixed it, thats pretty weird.

 

[buildzoid]

This killed my h100i's finish. I got trolled when I mentioned that on here twice now. Gonna mention it again. CLU sucks.

Also my 3770k (used, binned) didn't run my hyper X very well at XMP 1600, constant IRQL errors, and I didn't want to raise the voltage or vtt, so I had to reduce to 1333 default profile. The prior i7 3770k which I had bought new had no issues. Might be that the IMC on Intel's are weak with prolonged ram overclock or maybe just coincidence in my case. I am assuming the prior owner ran ram overclocked and overvolted as he ran a watercooled performance setup. Hopefully running my memory stock will keep any further degradation to a minimum. I haven't had any blue screens in the two months since I set it to 1333.

I still think it's a good TIM performance wise but when you make a mistake applying it and try to clean it up it goes from a small blob in the wrong place to thin layer of CLU everywhere which is much worse than the small blob. Also it's ability to soak into the the imperfections on a heatsink's base is annoying. The biggest issue is that Coollaboratory managed to get it to congeal on contact with alcohol so when you use alcohol when getting rid of it's easy but if there is even the smallest amount of alcohol still on the thing you just cleaned you can't reapply CLU.

 

[Vario]

I still think it's a good TIM performance wise but when you make a mistake applying it and try to clean it up it goes from a small blob in the wrong place to thin layer of CLU everywhere which is much worse than the small blob. Also it's ability to soak into the the imperfections on a heatsink's base is annoying. The biggest issue is that Coollaboratory managed to get it to congeal on contact with alcohol so when you use alcohol when getting rid of it's easy but if there is even the smallest amount of alcohol still on the thing you just cleaned you can't reapply CLU.

It formed this white alloy clump that would not remove. I'll sand it off when I lap the thing eventually. Try using Shin Etsu, I've had good luck with that.

I'm paranoid my IMC or anything else will go wrong with my 3770k since its used but so far so good besides instability with the xmp. Then again I have had weird issues with this ram on other systems (Phenom II, 3570k) so maybe they set the xmp vtt too low.