[SOLVED] Computer dies after new graphics card

Hi,

I’m on the ‘promo’ license so understand that there is no technical support, however this seems like a fairly serious issue.

I’ve run the software fine previously using a Quadro K5200, and have just bought a GTX 1080 - mainly for VR, but also with the hope that it will speed up software that relies on the GPU (like RealityCapture). I’ve had no problems running both cards when rendering (V-Ray RT and Iray for eg) but a few seconds after I start ‘reconstruction’ in RealityCapture the computer restarts. I’m running windows 10 and have never seen a program cause this kind of behavior before.

I don’t see any way in settings to manage the GPU usage, is this possible?

Thanks,
Kim

Hi Kim Laughton
RC can’t crash the PC just like that, from our experience it’s highly probably related to too high power draw from the PSU.
Please use this tool: http://outervision.com/power-supply-calculator#Expert and report your setup + let me know what PRECISE PSU are you using and what RAILS on the PSU are used/connected to the devices.
The priority is the GPU + CPU cables = rails…

Wishgranter wrote:

Hi Kim Laughton

RC cant just so crash the PC, form our experience its highly related to too high power draw form the PSU.
Please use this tool http://outervision.com/power-supply-calculator#Expert

and report you setup + let me know what PRECISE PSU are you using and what RAILS on the PSU are used-connected to devices.
Priority is the GPU + CPU cables = rails…

Thanks for the quick reply! - I assumed that it would be power draw at first and was looking at getting a new supply, but wouldn’t this issue also occur with other software that maxes out the GPUs? Iray (in theory) actually uses 100% CPU as well as 100% of both GPUs which ought to represent the maximum load.

The calculator gives me:
Load Wattage: 713 W
Recommended PSU Wattage: 763 W

My PSU is a 825W supply in a T7810 Dell workstation (a proprietary supply that doesn’t seem to be set up like conventional supplies, and only has one way in which to connect it)

I can get a more powerful PSU for the system but I’m not sure if it will give any more to the GPUs and want to be sure that this is the issue first…

Hi Kim Laughton
OK, thanx for the info. it seems the PSU isn’t able to deliver enough power to the system… As I see it, you need my expertise to create a good article about the PSUs…
So in short, is it possible to use only ONE ( gtx 1080 ) of the GPUs and do the testing ?

Wishgranter wrote:

Hi Kim Laughton

OK, thanx for info. its look the PSU is not able to deliver enough power to the system… As im see it need my expertise to create a good article about the PSUs…

So in short its possible to use only ONE (gtx 1080 ) of the GPU and do the testing ?

Thanks - just did some more testing last night and managed to replicate the problem with a game, using only the 1080, which suggests that the power supply can’t even handle that one card! Strange because the 1080 doesn’t draw so much power really and don’t really understand why the rendering software isn’t causing the same problem.

Because I can’t upgrade the power supply inside the computer with anything significantly more powerful I guess my option now is to get a second power supply to leave outside the case and power the GPU -_-

(I edited the post title for people looking through the forum in future)

Hi Kim Laughton
it’s because even though they write on a card some numbers, in reality the card behaves quite differently. GPUs have “peak” power draw and your PSU is on the limit of watt delivery. and if the PSU is older (in 1 year 5-10 % capacity is lost because of PSU internals’ aging), it loses its real capacity, so it crashes because the PSU can’t handle the power requirement…

Wishgranter wrote:

Hi Kim Laughton

its becasue evne they write on card some numbers in reality the card is behave quite different. GPUs have “peak” power dra. and your PSU is on the limit of watt delivery. and if the PSU is older (1 year 5-10 % capacity is lost because of aging PSU internals ) it loose its real capacity, so it crash because the PSU cannot handle the power requirement…

Ah - interesting, I didn’t know about the aging - is that something that happens only when it’s used or happens over time anyway? If I bought a new PSU and it had been manufactured a few years ago and sitting on a shelf could there still be this drop?

Hi Kim Laughton

read this article https://turbofuture.com/computers/How-t … Is-Failing
It’s not very technical but it can explain a lot…

Could also be a win10/DirectX 12 or nVidia driver issue if it happens with just one card, too.

There is a chance that only using the GTX 1080 with fresh installs of nVidia drivers and CUDA and DirectX would help.

Another thing worth trying would be to check how the card is actually supplied with it´s power.

There is usually 75W coming from the PCIe 16Slot, everything else needs to be feed in via cabling from
the PSU to the card, via 6pin or 8 pin cable connections, a 6pin connection is specified for 75W throughput,
an 8pin connection is specified for 150W throughput. Totallying 225W. Cards with 8pin+6pin can draw 300W accordingly.

The powertarget for the GTX 1080/1070 is usually 225W, with 150 TDP, e.g. 75W from PCI SLot + 150W from 8pin PCIe connector.

Using a single 6pin to 8pin adapter is not adviseable but merging 2x6pins to a 1x8pin adapater is within specs.

In a nutshell, there is a chance one GTX1080 can be made to work reliably by checking the power cabling for
strain connections and splitting out to another power lane, like trying a Dual Molex LP4 4 pin to 6pin to branch
out where the power is pulled from the PSU, separating supply lanes, so to say.

Cheers,

tim

Thanks Wishgranter & Tim!

Solved this by getting a second PSU - didn’t seem that the Dell supply could be upgraded or was intended for multiple GPUs (not sure why). Running the second supply outside the case, although not especially elegant seems to work fine, haven’t had any issues since. Thanks both for all the info - good to know for future reference.

Kim