Sounds like progress

No, I run each separate and try to get the lower card up to the higher one with voltage and MHz tweaks
 
Power & Temp Targets ---> Full Right - Max

Calling her stable!  Was able to go through valley for 3+ hours straight no issue.  Changes and timeline on how I got there below.  Short summary, I can't say that it was one thing changed to make my setup stable.  To find out if it was, I would need to do more testing.  My opinion, though, is that it was a few things.  Upping the mobo voltage, changing the memory gap, loosening memory timingshelped, I can tell by how far I can run Valley before the issue starts, and it was getting better.  But I also think that running the stock GPU clocks and voltage in SLI was not stable.  Maybe because of the difference in performance of each of the cards, don't know for sure.  I had to reduce my gpu clock to get it stable, but 1392mhz clock was what it would go down to when the cards reach 60C+ (before water cooling).
 
Thanks for the help!
 
timeline and changes:

• Went back to my 4.2Ghz setting (21x201)
• increased VTT to +150mV, and CPU PLL to 1.995 (all other settings from before carried over, ie memory gap at 3, uncore at 15x/3000mHz, 1.3375 vcore), power target in precision to 115%
• 1 hour of valley okay
• re-tested valley to make sure it was consistent, and failed at ~55 minute mark
• In precision I added +37mV to each card and unlinked them
• 1 hour of valley okay
• At this point, it was feeling pretty stable, so I re-installed the 12gb of dominator GT ram I had, set the timings real loose (what auto populated it with, then set it in manually)
• Failed valley after ~40min
• Gave the GPU's more voltage, went with 50mV on each card
• Failed valley after ~15min
• downclocked GPU's by putting in a -4 offset in precision for each card (went from 1405mhz to 1392mhz, only loss <1fps in valley)
• Ran valley for 3 hours and 40 minutes okay!  (fell asleep)
• Felt stable here, so I ran prime95 blend test for 15 hours while I was away from home to ensure all the bios settings for CPU, ram, mobo okay, and it was
• Ran valley for 1 hour and still okay

bill1024
Glad it seems to be stable. One thing, I would try to lower the PLL to 1.8 and see if it is still stable.
There is a huge x58 thread on another forum and they are lowering PLL to 1.6 claiming it makes the system more stable and cooler.
Some MB will not let you go below 1.8.  1.8 seems to be a default voltage.
If it were me i would try to get it lower if I could.

thanks for the advice, i'll try it out. My mobo can go as low as 0.6.
 
Unfortunately, I think I spoke too soon.  Was gaming in battlefield 1 for a few hours this morning, and it did the same thing, restarted unexpectedly.  I'll try above suggestion and also maybe downclock my GPU's even further.

I called evga premium support and the only suggestion they gave me was to try a different pcie slot. Said it's either my mobo or gpu. Was expecting more from premium help.

sicminded

bill1024
Glad it seems to be stable. One thing, I would try to lower the PLL to 1.8 and see if it is still stable.
There is a huge x58 thread on another forum and they are lowering PLL to 1.6 claiming it makes the system more stable and cooler.
Some MB will not let you go below 1.8.  1.8 seems to be a default voltage.
If it were me i would try to get it lower if I could.

do you mean CPU, IOH, or QPI PLL?

CPU PLL. Usually going higher results less stability. 1.775 seems to be a real popular compromise.

Sajin
To me this sounds like a psu issue. Have you done extensive testing at stock clocks to make completely sure you're having no issues at stock clocks?

Either that or a GPU issue, I doubt it is a CPU overclock issue.
I would run it stock clocks for a while, and then one GPU at a time. One at a time would be a PITA beings it is a custom water loop.
If he can run prime95 and Intel burn test at 4ghz 1.35v stable it is in the PSU or one of the two new cards.

I tried the new evga 1000w p2 power supply, and my rig won't even start up. Put my old psu back, starts right up. Tried the evga psu again, and still doesn't start up. My luck with evga products is not doing too good.

Sajin

sicminded
I tried the new evga 1000w p2 power supply, and my rig won't even start up. Put my old psu back, starts right up. Tried the evga psu again, and still doesn't start up. My luck with evga products is not doing too good.

Try testing the psu using the paperclip method. 

I did, it turns on with paper clip. I told the evga Rep it might be something with the cables. The mobo cable does have a hard time getting into the x58, but i inspected both the pins on cable and mobo, looked fine. The mobo light turns on too. When I hit the power button, i hear click, like a fuse tripped or something. Tried holding down power button, I hear the click again. Only thing that confuses me is when I put my old psu back in, works normal.

Purchased from a local Fry's. Not sure if it was an earlier model, it doesn't have the red gpu cables as pictures on Amazon.

Must be my lucky day, the second evga 1000w p2 power supply I picked up also does not work.  I got on the phone with evga tech support, and they said that the power supply for the second time is the problem.  What are the odds.  I went to another location to pick up the replacement too to try and increase my chance on a different batch of power supplies.  Oh well.
 
Good news is that the evga tech support had a great idea on how to test each of my 980 ti classifieds separately.  He helped me find the jumper cables on the x58 classified motherboard to disable one pci-e lane at a time.  Ran some testing tonight and each card was able to run 90 minutes of valley on default Precision settings no problem.  Going to game on each card tomorrow to see if I can reproduce the problem (recall I was still crashing in game even though I can run Valley for hours).  Hoping to isolate the problem to one card, or maybe if the cards work without issues individually, I either have an issue with my motherboard running SLI or a power supply that cannot handle both cards at the same time.  On the note of power supply, to maybe rule it out as a cause of the issue, going to still try a different psu (maybe the corsair ax1200i).

I'm sad to say that the motherboard is under RMA process.  Really sad, I'm the original owner of the board from years ago and has been good to me.  Somewhere in the middle of testing the GPU's one at a time and testing different power supplies, something went wrong.  The motherboard now no longer posts, it goes straight to FF error code which is a boot error.  Did some trouble shooting with the evga tech support to eliminate the GPUs, CPU, and RAM as the issue and concluded that it was the mother board.  Good thing is that I have the lifetime warranty on the motherboard from when I originally bought it.  Bad thing is that I'm worried sick if any other hardware, like the CPU, RAM, GPU's or hard drives were affected.
 
To be fair, I don't think the random restart issue is related to the motherboard dieing.  After further research, I strongly believe my old power supply is the issue.  Although it is a 1200W power supply, it doesn't have a single 12V rail.  In other words, each of the pci-e connectors have a rated amperage they can deliver.  Recall that I was mentioning that 2 of the pci-e connectors on the power supply had the same "12V5" label?  If I'm reading the spec sheet right, they are probably the same rail, meaning they were not supply enough power (or running borderline) to the 980ti's in SLI.  See my math below.
 
the four pci-e connectors on power supply and labels:
#1, "12V5"
#2, also "12V5"
#3, "12V4"
#4, "12V3"
 
spec sheet:
"12V5" = 20amps
"12V4" = 20amps
"12V3" = 20amps
 
I believe the equation is 12volts multiplied by 20 ampere = 240 watts (W).  So I had "12V3" connected to one GPU, "12V4" connected to second GPU, that leaves "12V5" supplying the other 8 pin connector to each GPU.  Say the "12V5" rail is sharing 50% of it's power (120 W) to one GPU and the other 50% to the second GPU.  So each GPU is getting only 360W (240 + 120).  Now, theoretically that should be enough, given that each 980 ti is rated at 300W.  But say if the power supply is only 90% efficient, the power it's delivering to each GPU is only 324 (0.9 x 360).  At 80% efficiency, it is delivering 288W (0.8 x 360).  I have a feeling that I've been borderlining the power supply the whole time, and if it hits some current draw limit, it shuts down.  Probably explains why my CPU overclock is stable, it has it's own seprate power supply rail which is supporting enough power.  Once i get a new motherboard (and hoping the other components are still okay), I really want to try a new power supply with a single rail.  I bet that it would fix the random restart issue.

I'm back.  EVGA did not have a replacement board for my X58 classified.  So they ended up giving me a brand new X99 classified.  Love their customer service.  This was cool of them.  They also acknowledged that I had a limited lifetime warranty on the original X58 classified, and they carried it over.  Only thing bad that came out of this is I had to buy a new processor and RAM because they are not compatible with the X99.  This led to purchasing a new case, water loop, etc.  Mod bug is really hitting me right now.
 
I got the X99 up and running tonight (or should I say this morning).  Rigged up something quick just to test things out.  The video cards and ssd's are still working.  But my 1TB hard drive is no longer working.  Keeps giving me errors when booting up, posts, but does not let me get into windows.  Probably died along with the X58.  Other than that, everything is working, was really worried something happened to the video cards, but turned out okay, I think.
 
Some issues with new setup:

• Weird restart issue one time when I was booting up, expect things to be smooth with everything basically new, will continue to monitor.  Feels like it boots up slower than my X58 with X5670.  Then again, i'm not overclocked.. yet
• Valley score seemed low, at 1080p extreme setting, only averaging 100fps, does this seem right?  I'm on my old monitor, so I can't compare to my previous scores from my higher resolution monitor.  Also it was choppy, stutters here and there throughout the benchtest, much smoother on my X58.
• Device manager doesn't recognize the X99 chipset, tried removing it and reinstalling the drivers a few times with no luck, suggestions are welcome

 
Hoping there's just a few simple things I'm missing, because setup doesn't feel optimized, feel like somethings off.  Will get in touch with EVGA tech support tomorrow.
 
new parts:

• x99 classified
• 6850k (using ek supremacy evo)
• corsair dominator 2400, 4x8gb (32 gb total)
• corsair ax1200i psu

 
max temps were 36C for CPU, 44C for gpu's after a few runs of valley and cinebench, only on a temporary 2x120 radiator!  and the cpu and gpu's are in parallel (trying this out, looks awesome this way, but I know the flow is not optimal, feedback welcome)
 
 

Latest x99 chipset drivers can be downloaded from here...

https://downloadcenter.in...river-for-Windows-?v=t

happy to report that the 980ti's have been running stable on the new x99.  again, I believe it was the psu and not anything wrong with the x58.
 
new issue though:  when playing battlefield 1, the screen flickers!  this is a downer especially since the flickering didn't happen on the x58.  after googling around, seems like the flicker issue is a game software issue, not hardware related.  just really weird that my x58 running the same cards in sli didn't flicker in battlefield 1.  I asked evga tech support why this is, they said that the way the drivers work in x58 may be different.  tried running borderless, but still had flickering.  sad.