Games/ OCCT PSU test rebooting PC ( PSU Problem?)

[Eadan1]

As the title says my PC started rebooting during games, and after running many stress tests and following suggestions, I finally found a way to replicate the problem. When I run OCCT PSU stress test, my PC reboots within a second. It was unexpected since running Prime 95 together with Furmark didn't cause any reboots.

Anyway, do you think the issue is my PSU (corsair hx750w)? I already have a new graphics card arriving within 10 days. How should I go from here to find the issue? My specs here: http://pcpartpicker.com/p/Xhn9D3
Edit: Linked the correct specs.

 

[basroil]

Prime+Fur means that maximum power draw isn't an issue, but something like a transient load would slip through. Try running a medium complexity game (something you would normally get 200-300FPS on ) but locked to 60fps, that might also cause large transients. If the game regularly crashes, it might be a bad set of capacitors on the output side of the power supply causing voltages to dip below the 12V specifications.

Only way to really make sure this is the issue and not something like memory or motherboard is to test the system with either another PSU or to monitor the voltages continuously in millisecond resolution (or preferably better than ms resolution). You should also test the memory just in case.

 

[Eadan1]

Thanks for the reply.

I already tested the RAM with MemTest earlier, it took 5 hours and no errors were found. Should I test each stick seperately by removing the other?

All voltage monitoring software I have shows correct values in the lines. The one with highest deviation is +3.41 volts in +3.3V Standby line. Do I need a PSU tester for better than ms resolution?

I don't have another PSU I can try currently, but I will try running a medium complexity game locked to 60 fp/s, and post the results.

 

[Eadan1]

I underclocked my GPU and the OCCT PSU test doesn't seem to cause any reboots.

Stock --> Gpu VCore:1.012V Core clock: 880 Memory Clock: 2100

Underclocked--> Gpu VCore: 0.95V Core Clock:500 Memory Clock:1500



Btw GPU temperature with either setup doesn't go above 55C if you are worried about overheating.



What does it mean? Should I try running tests under different settings like CPU OC'ed and such? The software already saves plots of a wide variety of voltages/temperatures/clocks, as well as taking a screenshot showing all, so I can post them here if anyone is willing to take a look. Here is the screenshot for underclocked settings: http://imgur.com/PS6GmdD

 

[basroil]

It could be either graphics card issues or PSU issues still. I would venture to guess a graphics one, but really hard to tell without more data. Try seeing if you can up the clocks slowly until you have crashing again, and look at the temperature and voltage readings. If you see the voltage on 12V drop below 11.5 you have issues.

You should also take the time to confirm that the PCIe power connectors are plugged in correctly, and if possible switch them with another PCIe power connector if your PSU supports multiple cards.

 

[Eadan1]

The issues were persistent before and after I did the following: I recently reapllied thermal paste to GPU and changed the fan, removed the PSU from the PC case to clean the huge amount of dust collected on the fan, readjusted the plastic spacer under it to give the fan more breathing room. Changed the sockets that the 2 PCIe power connectors from the GPU were connected to. ( PSU has 4 PCIe sockets, I connected to the 2 that weren't used previously)

However, I didn't change the PCIe connectors as you suggested. There is PCIe connectors that came with the graphics card, but I read that the connectors that come with the PSU should be preferred. I also have 2 previosuly unused PCIe connectors that came with the PSU, and I will try them soon and post again.

I am currently running the OCCT PSU test for 3 minutes, since I was getting instant reboot originally, but in games my PC can reboot any time randomly. (Though it reboots sooner more likely)

With GPU underclocked, I overclocked the CPU and still didn't get a reboot during test: http://imgur.com/oXMxTbA

Stock CPU--> Max Vcore: around 1.2V, Core clock:3.3Ghz

Overcloked--> Max Vcore: 1.4V, Core clock:4.5Ghz

Can I assume the issue is with my PSU or GPU most likely? And if it may help, OCCT saves many plots after a test in a folder which I can compress and send.

 

[basroil]

Connectors that come with graphics cards are usually for using molex connectors (originally meant for hard disks and the like), not really sure what your setup looks like but molex to PCIe is definitely wrong for a card of that power (or anything above ~100W for that matter). Make sure it's connected directly, though I doubt it'll help.

I have had issues with 560Ti failing due to poor memory/asic, and in that case lowering either the GPU or memory clocks helped, but ultimately it got worse and I ended up replacing it with a borrowed 660 (which was faster and used less power). Hopefully it's not that issue, but it's looking more and more like it. Try seeing if 3dmarks tests show artifacts at normal clocks, if they do, it's a sign that the card is starting to go the wrong way.

 

[Eadan1]

I have had issues with 560Ti failing due to poor memory/asic, and in that case lowering either the GPU or memory clocks helped, but ultimately it got worse and I ended up replacing it with a borrowed 660 (which was faster and used less power). Hopefully it's not that issue, but it's looking more and more like it. Try seeing if 3dmarks tests show artifacts at normal clocks, if they do, it's a sign that the card is starting to go the wrong way.

I actually would prefer if its my GPU that's failing. I already ordered a MSI gaming gtx970 4g, and want to make sure the issue is related only to my current graphics card before the new one arrives within 2 weeks. If it is most likely either my PSU or GPU, I can try RMA the PSU, so both possible causes will be eliminated. I just want a little more confirmation that my PSU may be failing before I invest in the abroad shipping fees and hassle.

Connectors that come with graphics cards are usually for using molex connectors (originally meant for hard disks and the like), not really sure what your setup looks like but molex to PCIe is definitely wrong for a card of that power (or anything above ~100W for that matter). Make sure it's connected directly, though I doubt it'll help.

You are right. The connectors supplied with the graphics card are PCIe to molex, and I am not using them anyway. Excuse my lack of PC building knowledge, I only started dealing with it after the problems arised. At least I am getting knowledgeable enough to build my own PC next time

I also tried keeping GPU VCore at the stock value and only decreasing clocks, and again there were no reboots. I thought only the voltage requirement of the GPU would effect the PSU, am I wrong? Here is the screenshot with stock voltage gpu, but underclocked (VIN0 is GPU VCore): http://imgur.com/wChY1IT

 

[Eadan1]

Correction : My current GPU is MSI gtx560-Ti Twin Frozr II/OC. My specs link show the ordered card. I will update the specs so it isn't misleading. Here is the correct specs: http://pcpartpicker.com/p/Xhn9D3

 

[basroil]

I also tried keeping GPU VCore at the stock value and only decreasing clocks, and again there were no reboots. I thought only the voltage requirement of the GPU would effect the PSU, am I wrong? Here is the screenshot with stock voltage gpu, but underclocked (VIN0 is GPU VCore): http://imgur.com/wChY1IT

It's actually a bit more complicated than just that, since most GPUs nowadays use dynamic scaling of voltages. If you undervolt it but keep clocks high (as close to reference as possible) and it stops crashing, it certainly can be a wonky card.

In any way, you'll probably have to wait for the new card to know for sure.

Oh, and a heads up, make sure you have the newest firmware for your motherboard before you put in the 970. If you don't, there's a chance that it will simply not recognize the video card!

 

[Eadan1]

The newest BIOS for my mo-bo was released in 2012 November. I already have it, so let's hope it supports the new GPU. Thanks for your help, I will keep posting any relevant updates.

 

[Eadan1]

It's actually a bit more complicated than just that, since most GPUs nowadays use dynamic scaling of voltages. If you undervolt it but keep clocks high (as close to reference as possible) and it stops crashing, it certainly can be a wonky card.

When I undervolted it but kept clocks at stock values, I got a reboot after 5-10 seconds of running the test. Other configurations result in either instant reboot or no reboots during the whole test. Also using Force constant coltage setting in afterburner didn't make any difference.

Try seeing if 3dmarks tests show artifacts at normal clocks, if they do, it's a sign that the card is starting to go the wrong way.

Also got a reboot during the demo part of 3DMark 11, then tried Benchamrks only and got a reboot again.

I tried running a test for 20 minutes with stock GPU Vcore (1.012V), but underclocked core (500Mhz) and memory clocks (1500MHz) . I noticed there are some interesting things happening with the plots. Mostly the CPU usage was 100%, but GPU Vcore was 0.95V, core clock 405 Mhz and memory clock 325 Mhz, but every 5 minute for 10-20 seconds the CPU usage would drop to 30%, GPU Vcore rise to 1.012V, and core and memory clocks would reach the set values (500MHz and 1500MHz).

Then I inspected the plots and noticed many unordinary things happening in those intervals. I will provide a link to all these plots at the end of the post, but here is a summary: It seems CPU Vcore rised by 0.05V when CPU usage dropped to %30. ( Note: I later tried with an overclocked CPU, with all throttling and monitor features turned off to make sure it wasn't intentionally done by mo-bo, and seems it wasn't.) Also the AVCC (+3.3V) seems to fluctuare between 3.33 and 3.31, but when the CPU usage drops to 30%, AVCC stays perfectly stable.

In the next test I did with the overclocked CPU (from 3.3Ghz to 4.5Ghz, with throttling and monitors turned off) the plots were very much alike. This behaviour might be intended by the test, but I don't think they would want to run only CPU or GPU at its max at a time when they are testing PSU stability. Anyway, here are all plots:
4 unordinary plots : http://postimg.org/gallery/2vf5i6ufk/a9932864/
A screenshot of the behaviour: http://postimg.org/image/xvifwrjcj/
All plots for CPU at stock: http://postimg.org/gallery/2mjdqj31m/b3a80cbb/
All plots for CPU overclocked: http://postimg.org/gallery/sfxikng4/fc06245d/

 

[Eadan1]

Seems the issue is resolved. I reassambled the PC, and changed sockets and connectors with spare ones. Also changed the extension cord and the wall outlet. Don't know which was casuing the reboots, but it seems fixed now. If I ever find the cause I will update this thread so people with the same issue can see. Thanks for your help.

 

[basroil]

Hopefully it was just an install issue, if there are any issues do update it. Just make sure to put as much info as possible (though memory dumps aren't really necessary for this system I would think.

 

[Eadan1]

Apparantly the issue wasn't solved completely. I got my new gtx 970 gaming 4g, run some benchmarks (Heaven Unigine) etc. to make sure it works. But when I tried playing Star Wars battlefront today, my pc seems to reboot occasionally during game menus but never during the actual gameplay. I monitor CPU and GPU temps and they are always below 60 C. I read my graphics card may cause crashes while transitioning from load to idle due to sudden voltage drops but it shouldn't result in a reboot as far as I know.
I just started playing witcher 3 to see what fps I will get, now I will try actually playing the game to see if I get reboots in this game as well, and update this thread..

Btw Corsair accepted my RMA request for the PSU, but sending it will still take some time so I want to avoid it if possible. What do you think may cause the PC to reboot at game menus particularly?

 

[basroil]

Battlefront is unstable without the newest nvidia drivers, so make sure you have the newest drivers.

As for the voltage drop, how much are we talking? 0.5V or dropping from 12V to 11V or lower? If it's under ~200mV drop it's not that bad, but if it's well above it that might be a bad PSU.

 

[Eadan1]

Battlefront is unstable without the newest nvidia drivers, so make sure you have the newest drivers.

As for the voltage drop, how much are we talking? 0.5V or dropping from 12V to 11V or lower? If it's under ~200mV drop it's not that bad, but if it's well above it that might be a bad PSU.

First I have to say reverting the overclock seems to fix the issue. Overclocked settings were 110% power limit, +140 core clock and +500 Memory clock with no additional voltage in MSI afterburner. These settings I found at:
http://www.hardocp.com/article/2014/09/29/msi_geforce_gtx_970_gaming_4g_video_card_review/8#.VhMoSitS09o
This overclock was stable for half an hour under Heaven benchmark . No crash, artifacts or throttling that I noticed, but there was a popping sound with no errors periodically which I assumed was part of the benchmark.

Thanks for the driver suggestion. I will update to the newest and try the overclocked settings again.

I meant voltage drop for GPU. It is 1.200 V for my card in games without overclock, weirdly it jumps to 1.225V mostly in menus as in following screenshots (Also drops to ~0.98V when idle):

Edit: Still rebooting with the new drivers while OC' ed.

 

[basroil]

That's way too high an extra memory clock for some card, and if you're pushing the 970 140mhz above factory overclock, that might be a bit much too. Also, unless your card is one of the few that uses a 167W or lower power limit, don't push power limit higher. Most cards already have a 10% or more power limit boost built in.

 

[Eadan1]

That's way too high an extra memory clock for some card, and if you're pushing the 970 140mhz above factory overclock, that might be a bit much too. Also, unless your card is one of the few that uses a 167W or lower power limit, don't push power limit higher. Most cards already have a 10% or more power limit boost built in.

+500 in MSI means total memory clock goes from 7000 to 8000. I don't know why it is defined that way either. The power limit boost may be the %10 that I set in the afterburner, further requires modding BIOS which I am not interested in now.
I tested around 2 hours with stock settings but if there is an issue with PSU, I might have been lucky and it might cause reboots at stock settings in other games as well.

 

[Eadan1]

Lots of kernel power entries in the event viewer. Could they be useful somehow?

 

[Eadan1]

As I feared it started happening at stock settings too, I will try reassabmling everything one more time and then just RMA the PSU.

Edit: Weirdly I am now suspecting my CPU may be causing the reboots. I was reverting every change I made since the PC seemed to work last, and one of them was reducing CPU cooler (Themaltake FRIO) fans from 100% to %70ish. I thought this was okay as CPU core and GPU temps were always below 65 in games with the new 70% setting. But after recent reboots there were some kernel power events with "thermalzone" in their description in the event viewer. Anyway I increase the CPU fans to 100% again, and haven't got any reboots within an hour. Of cource if I try OC ing the GPU again, I get reboots within minutes. ( CPU overclock is fine though)

I forgot to put the washers between the cooler and the motherboard while installing the CPU cooler. It was without the washers since I had the PC, and the temperatures seemed fine, so I thought they were not really necessary. Next reboot I get, I am reinstalling the cooler and putting the washers.

 

[Eadan1]

It probably was the Cstates. Enabled c1e and disabled c3e and c6e. No reboots so far. Guess I shouldn't have trusted the guides, some even official, saying I should enable c1e and leave others at auto. The latest BIOS had one of c3e and c6e at disabled. It makes sense as fully loading both GPU and CPU at the same time didn't cause issues, but since games put varying load at cpu and gpu at different times it was probably causing problems.

 

[Eadan1]

I tested prime stability to rule out motherboard VRM s causing the reboots. The only left possiblity was PSU. It was all along obviously a PSU problem but I couldn't be sure. Corsair accepted my RMA but I would have to spend more than the actual PSU cost to send it abroad and would also need to wait for the return. So I just bought a FSP Aurum PT 1000W. No reboots in 4 hours of Unigine Heaven. Also, with this PSU I think I can add another msi gtx 970 gaming 4g to the system in the future.

 

[basroil]

That PSU is beyond overkill for your system. Even overclocking it to the max (without using LN2) would keep it under 600W. Putting 1kW is just going to waste money (about $50 now, and another 1-5 cents per hour of use)

 

[Eadan1]

The retail store I bought the PSU from didn't have many options, and this was the only tier1 supply I could find there. I could have bought from amazon a better and cheaper PSU but I didn't want to wait 3 weeks for their shipping.

Also, wouldn't I be saving on electric bills using a platinum rated PSU? Since the PSU has maximum efficiency at 50% load, if my system requires 500W at load, 1000W is the optimal capacity. If it requires like 400W, 800W PSU would be better but wouldn't leave me much room for added parts.