Sudden Reboot

Morning All,

I have seen this around but nothing too recently and this was a first for me so I thought I would pick the collective brain.

Two nights ago I upgraded from my old twin 970s to an EVGA 1080ti. Prior to this, the system has always been rock solid.

• i7-4790K
  
• 32GB Kingston HyperX RAM
  
• Samsung 850 EVO SSD
  
• Asus Z-97A
  
• Corsair H100i water on the CPU
  
• Corsair HX750 PSU
  

After installation I was only able to screw around for about 30 minutes, things looked good and everything was running very cool, I am struggling now to remember exactly but I believe the GPU maxed at 68 C.

Last night I had a little more time, I was flying the A10 for a little over an hour, getting ready to call it a night, when the machine just completely rebooted. At first I thought maybe our electricity flickered at the house, but nothing else had any issues from what I could tell. I haven't had a chance to try to look at the Windows event log, but that is my next move.

My immediate thought is power supply. The Corsair PSU in there is 3.5 years old. I know 750 watts should be plenty for this rig, so I would guess maybe age is taking its toll?

Regardless, are there any other logs I can try to access to see what the problem might have been? Are there any GPU issues that anyone is aware of that would cause a sudden reboot, as opposed to a blue screen, game crash, ect?

Thanks

Run prime95 and Realbench together, that will max out your power draw.

If the machine reboots while you try that = PSU too small/old

750w should be ok but if the PSU is not 100% integer it might trigger the reboot ( and may kill your HW down the road )

Thanks Bitmaster,

I ran a Realbench 15 minute stress first. All passed OK, 62C max on GPU, 58C max on CPU.

Restarted, waited a bit, and started Prime95 (I did the in place large FFTs since it said it would have max power consumption) and then immediately started the Realbench 15 stress test. Realbench said all OK. All 8 windows on Prime95 showed no issues/no warnings. CPU maxed at 65C, GPU at 63C.

From my Event Log from last night, 2 events seem to have precipitated the restart, both with an Event ID code of 1074.

• "The process explorer.EXE has initiated the power off to computer GAMERIG on behalf of user GAMERIG\GAMERIG for the following reason: Other (Unplanned). Reason Code 0x0. Shutdown Type :power off
  

-2 Seconds Later

• "The process C:\Windows\system32\winlogon.exe (GAMERIG) has inititiated the power off of computer GAMERIG on behalf of use GAMERIG\GAMERIG for the following reason: No title for this reason could be found. Reason Code: 0x500ff. Shutdown Type : power off
  

After that the log shows a cascade of services stopping, then the shutdown.

Try a new PSU. Last time when my PC randomly rebooted itself, I changed the PSU and now everything is fine.

Turn the PSU off. Install your anti static wrist strap. Unplug and replug all power connectors except the 120v (to the wall). Also make sure the card is fully seated into the PCIe slot. Remove, clean (pencil eraser) and reconnect.

I suspect it could either be the PSU or the Graphic card.

Try what Demon said reseating the graphic card and unplug the power cables and reinsert them and try again.

With the PSU 750 is more then enough for that setup. If your using cable extensions with the graphic card try a direct connect to the card.

If you still have your old cards you can try putting one of them and see what happens in the same slot.

Usually reboots with a new card point to the card itself or psu. Bitmaster is right too with it could be old which psu lose there rating over a certain amount of time.

Hope it helps and keep us updated.

If the PSU cuts out there will be no log entries or such, it's a very quick & sudden OFF.

Something is telling the OS to power down gracefully, this is something different han a PSU stalling and cutting out abruptly.

Still, if you can, change the PSU with a friend etc. and see if that cures it.

Thanks all. I am out of town this weekend but I think at this point I am leaning towards picking up a new PSU. I think there is merit in rechecking all the connections, and if I am going to go through that I would rather be just hooking everything up with new cables to a new supply. This way I can try to cover all the bases at once and won’t risk a bad PSU taking anything with it if it decides to really fail. I just don’t have time anymore to spend troubleshooting, if I can be by the rig I want to be flying, not cracking it open on the workbench.

Also, since I have realbench now, does anyone have an idea what a baseline score (nothing overclocked) should look like, given my specs above? Maybe that could help point to a GPU issue if I am way off...

Las try before change PSU: if you are using the two 8pin cables from the same rail, try using 8pin from 2 different rails.

With my old Alienware Aurora 875W PSU I was having freeze and system restart using cables from the same rail.

Fair enough, I think there are enough ideas here to warrant not pulling the trigger on a new PSU quite yet, particularly in light after considering @Bitmaster's point that a PSU failure probably wouldn't lead to an event log, that makes sense to me. I'll try all the given advice with regards to cabling.

@Demon, with the pencil eraser, you were talking specifically about the contact points on the GPU card where it goes into the PCI slot, correct? (Pencil eraser makes sense to me, whether comm cords or headset plugs that has always been my go to.)

Las try before change PSU: if you are using the two 8pin cables from the same rail, try using 8pin from 2 different rails.

With my old Alienware Aurora 875W PSU I was having freeze and system restart using cables from the same rail.

That depends on the PSU's layout. Most of the recent PSU's I have read reviews about are single rail, so it won't matter from the ampere but only heat.

Whereas, if you have a multi-rail layout your suggestion does make perfect sense.

Yes, the copper.

Alright everyone, thanks for all the suggestions. I complied with them this afternoon, put it back together, and ran a couple more benchmarks. I haven’t had another shutdown. I have a friend with a basically identical system (we built at the same time) and we are benchmarking essentially the same.

I guess at this point I just have to play some more and monitor. On the good news front, while going through my documentation I found out that PSU has a 7 year warranty, so if I do have to RMA anything I am covered either way.

Thanks again all. I will give an update if I make it a while without problems or if something happens again. This is a great community.

Bad news. Maybe. So the first time this reboot happened it was pretty deep into mission 2 of Georgian Hammer CA. I started that up again tonight , once again about 30-40 minutes in- I was watching a helo burn in when WHAM, sudden reboot. So same mission, roughly about as far into it when the total crash happened.

I ran a DCS repair, played again. I caught some damage and had to RTB but this time ran the mission to conclusion with about 30 min of run time and no issues. I had GPU-Z going, nothing got even close to 65 C....so now I really don’t know what to think.

tom 19d just a though :

- you run a liquid cooled cpu, you probably have it OC ? if so it's probable that your Vcore is not enough (or not anymore since you upgraded your gpu ? , or you had to set again your cpu settings after a MB bios update and messed up?)

- you say you run realbench stresstest for 15mn, that is not enough : it can pass 30mn OK but crash after 1 hour. On my side to be sure it's ok for my usage I run a 2hours stresstest 1 time. I think less is useless to be sure for gaming OC.

Just be sure your Vcore is not the issue.

Thanks but I have never overclocked any of the components and I haven’t touched any of the MB settings in a long time. I get the idea behind a longer stress, but that seems kinda weird to me since DCS isn’t pushing my cpu to near the level that the stress test does.

Does it reboot slowly or black screen completely off, then reboot? I had a PC at work doing that all of a sudden, it could sometimes run all day fine. It ended up being a bad stick of ram.

It just shuts off completely, like someone held the power button and forced a shutdown, then it immediately goes into reboot.

I ran the 2 hour BenchMark stress test. Test OK was the result, 24x Result Hash Match. GPU maxed out at 63C. CPU maxed out at 61C. Obviously the machine didn't reboot mid-run.

Get hold of a USB stick with Ubuntu or any derivate there of, those are easy to use even for a novice. You can also make one yourself, 2GB Stick is plenty.

Once you got that stick, boot from it and run your PC on Linux for a few days and see if the error is OS independent or tied to 10 only.

I myself have a similar problem, a 2-5 sec freeze and then goes on, and I could limit it to 10, Mint Linux has no issues. 10 has it even after my new install from last weekend, clean drivers etc... and dang..FREEZE on day 2 for 5 sec in YT...sucks....so my search goes on and I hope for new Bios, drivers, new 10..everything but get those freezes away. luckily..if they happen in DCS I do stay connected, if I dont crash it's all good. It came by itself, it will go by itself, just keep calm ;)

This is less dangerous:

1: DL Rufus -> https://rufus.akeo.ie/

2: DL Ubuntu -> https://www.ubuntu.com/

3: Take a 2-16GB USB20 stick and follow theguide at rufus website, basically point Rufus to the ISO file and let it do it's work, use the Presets presented.

4: Boot your PC, hit "F8" to get to the boot menue and select the USB stick

5: Do your daily stuff now, Firefox, VLC, etc... you can download prime95 if you like, or stressapptest from google and torture the rig a bit.

If it wont shut down in Linux it is likely a driver. Whenever I have such a reboot or BSOD the first door I knock on is Nvidia drivers.

tom_19d just another though….

This time focusing on your 1080Ti. And considering the option : stock values (frequencies core and Vram as it seems to be your will), and the fact (I suppose) that your one is air cooled.

Mine is also air cooled and considering this pecular gpu, the default voltage in balance with temperature is not good. A liquid cooled 1080Ti, from what I've heard from several users, can work flawlessly with stock voltage (1.043v in full load) and even with an OC profile over 2Ghz core.

But a 1080Ti even with a good custom air cooling system can't work flawlessly (stock frequency can't be sustained) with stock voltages. Run GPU-Z, activate GPU-z render test, look the sensors and you will see at default 1.043V the "problem" will be voltage reliability.

Maxout software voltage with afterburner (100% of allowed V increase, that gives 1.093 v max) and you will still see voltage reliability be the "problem".

Then, with afterburner, put the air cooling to 100%, wait some seconds, and the "problem" will disappear (no more V reliability issue, gpu takes between 1.083 and 1.093V), "problem" being perf cap issue in gpu-z sensors. At this time the 1080Ti runs flawlessly at its max stock frequency.

That is to say that, like a weak Vcore can crash your cpu, a weak gpu core voltage can also crash system, and with default values with a 1080Ti air cooled, voltage is too low (not the case with a liquid cooled one I think because temps are way lower and that means less voltage needed) and your default core frequency even can't be sustained (for it you have to max voltage and have aggressive fan curve, mine is 100% at 60°c and 90% at 50°c)

Just my experience and many will disagree and even say it's unsafe for longevity.

Tried inverting the cable as I've suggested?

Set at Single or Multiple rail?

Taken from the official corsair forums:

I have a RX Vega 64 in my system. With link set to multi rail (850hxi) the system would flat power off when I loaded Far Cry 5 levels. Battlefront 2 made it 5-10 minutes. Same thing.

 

Set to single rail and the situation cleared up immediately.

tom_19d just another though….

 

This time focusing on your 1080Ti. And considering the option : stock values (frequencies core and Vram as it seems to be your will), and the fact (I suppose) that your one is air cooled.

 

Mine is also air cooled and considering this pecular gpu, the default voltage in balance with temperature is not good. A liquid cooled 1080Ti, from what I've heard from several users, can work flawlessly with stock voltage (1.043v in full load) and even with an OC profile over 2Ghz core.

 

But a 1080Ti even with a good custom air cooling system can't work flawlessly (stock frequency can't be sustained) with stock voltages. Run GPU-Z, activate GPU-z render test, look the sensors and you will see at default 1.043V the "problem" will be voltage reliability.

Maxout software voltage with afterburner (100% of allowed V increase, that gives 1.093 v max) and you will still see voltage reliability be the "problem".

Then, with afterburner, put the air cooling to 100%, wait some seconds, and the "problem" will disappear (no more V reliability issue, gpu takes between 1.083 and 1.093V), "problem" being perf cap issue in gpu-z sensors. At this time the 1080Ti runs flawlessly at its max stock frequency.

 

That is to say that, like a weak Vcore can crash your cpu, a weak gpu core voltage can also crash system, and with default values with a 1080Ti air cooled, voltage is too low (not the case with a liquid cooled one I think because temps are way lower and that means less voltage needed) and your default core frequency even can't be sustained (for it you have to max voltage and have aggressive fan curve, mine is 100% at 60°c and 90% at 50°c)

 

Just my experience and many will disagree and even say it's unsafe for longevity.

I agree 100% with that. Under water it can be undervolted quite well too.

The key with Pascal is to keep it below 30°C for max OC beyond 2.1GHz, below 50°C for a reliable 2050-2100MHz oc and for sure stay away from anything higher than 50°C.

No air cooled card will ever be that cold, they run anywhere 60ish and up.

HardOCP tested the HKIV Pascal waterblock and it kept that thing well in the 2.1GHz range...that is 18°C cooler than my shabby Poseidon.

From my personal pov, watercooling a GPU makes more sense than CPU, if only 1, then take the GPU and do it right 1st time, no Poseidon, no Waterforce, go full cover HeatKiller-IV.

Amd yes, in order to sustain 2050MHz on my GPU I need to bump the volts, so I settle with a few MHz less and leave the volts where they are, default.

If, then i tune it down, not up.

Wow, thanks everyone.

I travel for work so it makes it a little hard to keep up right away sometimes, particularly when the issue is hardware and my machine is hundreds or thousands of miles away.

I'll start with Automan. I attached the (stock) photo of my PSU. The single/multiple switch isn't present. Initially after installing the card, I had both PCIE cables plugged into the top two blue connectors. While replugging all the power connections, I moved one of the PCIE cords to the bottom blue plug, not that I think that would matter. Also not that I think it would matter, but I am using two of the peripheral plugs.

Bitmaster, thanks for your detailed write up on booting from a Linux drive. I have zero Linux experience so I guess my question would be if I boot from one USB drive, I will still have access to all of the software I need for testing (DCS obviously, TrackIR, ect)?

Toutenglisse and BitMaster- I have never done any overclocking but I think I am mostly following you here. (And just to confirm, I am air cooled on the GPU). But if I may ask some questions to make sure I am following you.

-I have always used GPU-Z to monitor new cards/setups/ect but I never really understood the vRel which always shows up as the "capping" factor on the card. Also, I can confirm that when the card is being pushing I almost always see 1.043 as the voltage, although I have seen it on occasion go to 1.050, never higher than that. So are saying that as the card senses it is warming up, it is limiting itself to 1.043 volts, when it is capable of handling more? I guess I just don't have the background to understand what vRel really means.

-Second, not that I am questioning anyone's expertise, but if the PSU/CPU/GPU were all able to run through a 2 hour stress test without a hickup, I would think that points to hardware not being the problem. Am I missing something?

-Finally, and again not to seem disrespectful to anyone's time since you are all being so generous, but I see all sorts of people with 1080ti's in the signature lines, lots (or most, idk) of which I am sure are air cooled. If the 1080ti has such a potential for instability in its natural state, wouldn't there be more people in my shoes? Essentially, I have always been taught in airplanes if something suddenly stops working, undo the last thing you did...the last and only thing I did here was install a factory new GPU and suddenly a once robust systems is failing. Not to be a defeatist but would attempting an RMA with EVGA be out of line? (I don't think I am quite at this point yet, I also did do the very basic step of running a DCS repair after I ran the last stress test. I would like to give it another chance at actually running the game before I make another move when I get back home).