[SOLVED] PC shutting down after installing new GPU

[zenx]

Hey guys.. this is an odd one.
So I upgraded my ASUS Strix GTX 970 to a Gigabyte G1 GTX 1080 Gaming yesterday and since then.. my pc randomly powered off during decent load (playing games / streaming them.. etc) for the fourth time..

Now the catch is that usually I'm the one helping people out with similar stuff and I'm quite aware that usually this happens for 2 common reasons.. One of them being temperatures, the other being power issues.. Well, temperatures are just fine, but will obviously double-check (I built this pc like 2 months ago and I ran some tests with prime back then and it all seemed fine) and the power supply is a Cooler Master V650 Gold, which should be more than enough to run this card, especially since the 970 doesn't use that much less power than this one and it ran that for the past 2 months flawlessly... Like I've been streaming 8-10 hours daily for the past month without any issues whatsoever with either temperatures or power or any sort of crash.. (if it was a CPU temp issue, it would've presented itself under heavy load from streaming)

So based on the above, I find it really strange that I swap out the GPU.. and this is what happens... Is there any chance the new Gigabyte GPU is somehow faulty from the factory and causes this? That would also be weird since in that case, why does it run for a while before it happens?

I also logged the system stats and at the point when one of the shutdowns happened, everything seemed alright.. GPU temperature was a mere 68 degrees C.. CPU usage at 40%... (don't have the temp for that in the log, but it should be fine)..

Right now I'm wondering whether I should just return the card...
Lemme know what you think..

Cheers

 

[10tacle]

I would normally say crashes are because of driver conflicts if one didn't remove all previous drivers and load new ones from scratch (direct from Nvidia) with a new GPU. However, complete shut-downs are unusual with a GPU swap with enough PSU power which yours has. Have you tried underclocking and undervolting that 1080 and testing it?

 

[zenx]

I would normally say crashes are because of driver conflicts if one didn't remove all previous drivers and load new ones from scratch (direct from Nvidia) with a new GPU. However, complete shut-downs are unusual with a GPU swap with enough PSU power which yours has. Have you tried underclocking and undervolting that 1080 and testing it?

No, I haven't tried underclocking, but if it was instability due to high clock speeds it would simply crash / reboot imo.. we're talking about a complete power-loss here as if someone pulled the plug and without any restarts, delayed or otherwise.. So no crash dumps.. no logs.. no mesasges.. usually this only happens when your CPU is about to fry... or if something is really messed up..
I mean I'm not sure, but I think even if it wouldn't have enough power, imo it would simply throttle the card under normal circumstances... but like I said, this PSU ran my overclocked 970 at 120% power cap so that's pretty much the same or higher as this card uses.. same cable / plug.. everything..

That's why I'm curious if there's a chance a faulty card can do something like this, since I haven't really dealt with those in the past.. pretty much every card I ever touched was working properly so idk what kind of hiccups a bad card could cause.. but I'm skeptical, since I would expect it to do a power cycle / reboot in those cases too.. not a complete shutdown..

 

[10tacle]

Yeah that's what I meant about driver conflicts causing crashes vs. complete shutdowns, which usually means a hardware failure (CPU, mobo, PSU, or memory). Not sure what your CPU is, but can you take out the GPU and try doing the exact same streaming operation running only through your CPU/APUs onboard graphics?

 

[10tacle]

By the way: I assume you put the 970 back in there and it works fine.

 

[zenx]

By the way: I assume you put the 970 back in there and it works fine.

That's the catch I gave away the 970 to my dad and already installed it for him in his PC so I don't have access to that anymore, which pretty much means if I end up having to send this back, I'll be stuck with no PC for the time being.
It's a Ryzen 1700x btw so no onboard graphics

I just did some stress testing though with prime.. and noticed that I do reach close to 80 degrees on its own during a full torture test in 10 minutes or so..
BUT to add a bit of twist.. I added furmark to the equation and ran that for a bit on its own.. gpu went up to like 73 degrees.. cooling kicked in properly.. but through heat exchange my CPU was also rising in temp.. without any load.. then I kicked in prime for just a minute on top of that and I was already above 80 degrees in the first few seconds.. then slowly rising to 82-83 at which point I stopped them all..
Now granted.. gaming AND streaming it at the same time only uses my CPU to like 30-40%, while prime throws everything at it instantly.. but we're talking longer periods when gaming so I'm starting to think that the new card might be running hotter than the old one or due to a slightly different design of the back-plate it might be giving off more heat at the top than the 970 used to.. and actually helping my CPU to overheat.. that's the only logical explanation I have for the overheating that explains why it's only happening with this card..

Still, even if that's the case, throttling should kick in, instead of a shutdown, but if it keeps getting passive heat from the GPU who knows.. it might be enough to trigger the failsafe.. (my pc seemed to be much hotter to the touch during this short stress test than it was during the shutdowns though, so I'm still skeptical, but I need to explore this possibility too)

So what I'm gonna do is.. do the normal streaming / gaming thing.. and have ryzen master open on my second screen to watch the CPU temp and see how it does in that scenario..

Will post if I get any results..

Cheers!

 

[zenx]

Yup.. the 'wild' theory was correct...
Somehow this particular card gives off way more heat at the top plate than the previous one and due to the proximity, the CPU cooler was drawing in that hot air, effectively transferring some of that heat to the CPU.

Now, I couldn't trigger the shutdown again afterwards (most likely cuz of the rainy past few days, which dropped the room temp by 1-2 degrees), but I was monitoring the temperature constantly (effectively included it on the stream so in case it shuts down I can go back and watch the video) and it was hovering around thermtrip values when under decent load so it's quite possible that was the cause.

Anyway, I ordered another noctua fan and set up my CPU cooler in push-pull config with the Y splitter (I have that NH-U12S SE-AM4 which comes with a single fan by default) and added an additional exhaust fan to the case. This dropped the temperatures all across the board by almost 20 degrees, idle temp is down to 38 with a room temp of almost 28 atm (was around 55-56 before) and the highest it went with 10 minutes of extreme testing was 88 (by extreme I mean 1080p furmark stress + prime95, which is probably something people shouldn't normaly try and definitely more load than you can ever get under normal conditions).

Sooo I guess that's it.. lesson learned, changing one component can throw off your cooling completely I should probably mark this as the solution.. (or someone else, since I prolly can't)

Cheers!

 

[10tacle]

Actually there really is no solution best here. You got it done the hard way as many of us have experienced over years of dealing with home-builds LOL! I'll pick your solution as the best. Greetings from across the pond in the US!