3D Mark kills the power. OCCT doesn't....

[KDrago]

System:
Asus Maximus V Z77
CPU: i7-3770K(not overclocked yet)
PSU: OCZ Fatality 3 1000W
GPU: MSI Nvidia GTX 780Ti TwinFrozer OC Edition
RAM: G.Skill Ares 16GB 1866
Samsung SSD
Main Display is a 1080p HDTV so all my gaming resolutions are as such.

I have been dealing with a rather obnoxious problem in my rig. It has worked fine for months until I upgraded to a 780Ti about a month ago. My system will crash(full power loss) when under heavy load. I played around with the audio drivers like others on AnandTech had recommended, and it seemed to reduce the crashing, but it still happens regularly. The only game I am usually playing when it crashes is Planetside 2. I am thinking its a pretty decent load on the CPU and GPU together, and that is why its crashing.

The system always reboots itself after about 5 seconds.

I installed several pieces of software today to try and see if I could get some more insight.

When I run 3D Mark I don't even get 30 seconds into the lowest test (around the first rendered explosion) before full crash. I don't think any results are recorded.

I decided to try OCCT to see if I had any better results. I have run OCCT cpu and gpu tests for 20 minutes without any issues. No over heating or abnormalities. I have a power meeter attached to the outlet and it will read a consistent 362 - 368W and 223 - 229 fps.

When I try the Power Supply test using all Logical Cores I get around a 415W power draw that will drop down to around 270W after a few seconds. It will periodically spike back up to 415W and just keep going back and forth. No crashing

If I unselect "Use all Logical Cores" the power draw will stay at between 395W and 405W. And again no Crashing.

3DMark will pull around 380W before everything goes dark. Same with Planetside 2.

And thats where I am at... Not sure what to try next. Anyone got ideas?

I don't want to buy another PSU just to find out its something with the MO board or the GPU

 

[dragantoe]

my only guess could be faulty ram, I've had a lot of experiences with bad ram and this sounds like one of them...

 

[zebrax2]

Test without the GPU and with only 1 stick of ram at a time.

 

[RampantAndroid]

Test without the GPU and with only 1 stick of ram at a time.

Don't use 3D Mark to test RAM, there are simpler ways.

http://www.memtest.org/

Test with all sticks; if errors, then test with 1 stick in at a time until you find the bad stick(s).

 

[Subyman]

Usually RAM problems give a blue screen, not a complete power loss. A complete shut off sounds more like a PSU problem. Run memtest through a USB stick for 2-4 runs to test the RAM. Run Furmark without any other program to see if you can get it to crash with only a large GPU draw.

Do you have any spare parts to swap out to eliminate some things?

What video card did you have before the 780TI? If it was AMD, did you do a complete driver wipe?

 

[Burpo]

Safety on Northbridge temps may be shutting it down..

 

[bononos]

Safety on Northbridge temps may be shutting it down..

But what could be stressing the northbridge out? Isn't the memory controller on the cpu?

 

[Burpo]

Ripple from the power supply, bad caps or other issues.. Why not try a fan in that area to see if it helps?

 

[BonzaiDuck]

When I run 3D Mark I don't even get 30 seconds into the lowest test (around the first rendered explosion) before full crash. I don't think any results are recorded.

I had a problem with the same symptom but likely different causes. Then again, maybe not.

First of all, 3DMark or 3DMark Vantage won't work with something like Lucid Virtu. Right into the running of it, the whole system will seem to freeze while it seems to be gathering information . . . . perpetually. I got the general observation from FutureMark tech-support, and the experience of it was firsthand.

Lucid or no Lucid, what did you do about your Intel iGPU? Try disabling it -- turn off "Render Standby" and everything else.

Also, just as a preliminary measure, try uninstalling the NVidia drivers and software for the 780ti card, or download the latest revision and then do a "clean install" from the custom menu option.

And I'd also suggest seeing how Cinebench v.15 works. It should put the graphics processor through similar hoops. I can only say that right now -- FutureMark is on my S*** list, but then it may not be "their fault." Even so, I uninstalled 3DMark Vantage and plan to reinstall it after I've done some other things -- with a system, by the way, which now seems stand-up rock-stable.

I had gone through all the categorized suspicions including RAM. Because I couldn't wait for a week or so to see whether I'd solved the problem, and because I really wished I had a 2x8GB kit instead of a 4x4GB, I needlessly gave myself a RAM upgrade -- which, of course, passed all tests.

Anyway, something like this -- sometimes it's hardware, sometimes it's software, and sometimes it can be both hardware and software in interaction. If you still have the problem after a focus on graphics, BIOS and other related complications, write down your list of hardware drivers for everything, go out and download the latest versions, and reinstall them one at a time.

 

[Burpo]

Full loss of power & reboot is usually a temperature problem, (shrug)..

 

[KDrago]

Sorry for the delayed response. Had a busy end of the week.

I ran my system with my RAM one stick at a time to be sure there wasn't anything drastically wrong with any single stick. No difference in performance, everything still went dark at the exact same moment on each stick.

My system's Motherboard has a lot of heat monitoring software provided with it. I have been tracking the heat very readily and nothing on the board seems to go above 70 C at any time. If something is heating up to the point that its causing system shutdown, it would have to be happening so fast none of the monitors are picking it up. Also since I can play most games for hours on end without issue I am skeptical of overheating being the issue. In addition the crashing is very consistent timing with Mark 3D. If it was heat related, then I would expect there to be some slight variation in the point that the system crashes during Mark 3D testing. It is practically crashing on the same frame of animation during the tests.

Someone also mentioned the Northbridge, but the i7 boards no long use a separate northbridge setup and instead its integrated into the cpu (maybe I am wrong on this?)

Nvidia released a new drive last week that I have not tried but other than that, I have done several driver wipes and clean installations that I am pretty sure the drives are fine. (unless Nvidia just has defective drivers for the 780ti).

@BonzaiDuck I tried Lucid Virtu when I first built my system. I kept getting freezing all the time and frequent game CTD(crash to desktop) I did a fresh install shortly after and never reinstalled Lucid and things were a lot more stable. So short answer is, No Lucid.

I'll checkout Cinebench next time I get a chance.

I will also try to get some time to test the old graphics card under 3D Mark. Its a GTX 660 MSI OC Twin Frozer Edition.

 

[KDrago]

I swapped out my video card and put in the 660. I ran Mark 3D and its exactly what I expected. The 660 pulls about half the power the 780ti does at load. Thus obviously no crash.

 

[KDrago]

Interesting Development

After I reinstalled the new card, I decided to play around with ASUS Suit II that came with the Motherboard. I noticed it has a tab that says EPU (Energy Processing Utility) When I load that section they have a "Power Saving" and "High Performance" setting. I went to High performance configuration and see an option for "Dynamic Power Management" and it is set to Enabled. Well just out of curiosity I switch it to Disabled and apply it and run Mark3D.

Mark3D ran for almost a minute before I got the crash. Just to make sure it wasn't a fluke I ran it again. System runs to the same spot(almost 1 minute) and crashes. I turned "Dynamic power Management" back to Enabled and ran M3D again... System is back to crashing only 30 seconds in.

Anyone have any insight into this?

 

[Burpo]

I know you think you have a good power supply, but it's looking like the PSU more & more..

 

[guskline]

Do you have a friend that can lend you a PSU? I have PSU calculator pro and even at 100% load a 550W PSU of decent quality should be fine.

Have you tried other connections to the 2- 6/8 pin connectors to the video card?

Also, what heatsink do you have on your 3770k ?

 

[KDrago]

I did one last test last night. I turned off the EPU tool. I set my ASUS bios back to "optimized defaults" settings, and ran Mark 3D again. Back to 30 second crash.

So now I am wondering, if this is something with ASUS's motherboard power management features.

I know you think you have a good power supply, but it's looking like the PSU more & more..

Its not that I think I have a good PSU its that unfortunately the PSU is also the one component that I can't RMA... OCZ went bankrupt and is no longer honoring warranties on their PSUs. So if thats the part that I need to replace, its all coming out of pocket. So, I just want to be pretty confident that its the PSU and not anything else before I go dropping the cash.

Do you have a friend that can lend you a PSU? I have PSU calculator pro and even at 100% load a 550W PSU of decent quality should be fine.

Have you tried other connections to the 2- 6/8 pin connectors to the video card?

Also, what heatsink do you have on your 3770k ?

I have asked around, no one has a PSU that they can lend me.

Yes, I have tried different Pin connectors to the GPU and even tried moving the card down to the second PCI-E slot, 8x instead of the 16x. No difference there.

I am using one of the closed system liquid coolers from Corsair. Temp on the CPU never gets above 50C it seems. I have reseated the cpu twice while working on this issue. The thermal connection seems solid with the coolers heat pad and I use Arctic MX-2 thermal compound.

 

[theAnimal]

I swapped out my video card and put in the 660. I ran Mark 3D and its exactly what I expected. The 660 pulls about half the power the 780ti does at load. Thus obviously no crash.

Definitely seems like PSU may be bad since that PSU can supply almost all of the 1000W to 12V, so 780 should be no problem.