GTX 1080 Founders Edition Died - 12V to Ground Short

[rhkcommander959]

Success! Booted into windows, realized I was using a crappy 2.5in 5400rpm laptop to boot so I couldn't get the nvidia drivers downloaded to try a quick 3D.

When testing resistance through 12V and ground I think it was ~3.5kΩ but I honestly don't remember for sure.

I'll get the heatsink off for a final inspection, button it all back together, and give it a proper benching. It'll be a few days at least though before I can get that far.

 

[Nobu]

Success! Booted into windows, realized I was using a crappy 2.5in 5400rpm laptop to boot so I couldn't get the nvidia drivers downloaded to try a quick 3D.

When testing resistance through 12V and ground I think it was ~3.5kΩ but I honestly don't remember for sure.

I'll get the heatsink off for a final inspection, button it all back together, and give it a proper benching. It'll be a few days at least though before I can get that far.

pics while it's naked?

 

[rhkcommander959]

pics while it's naked?

First post has a bunch of naked pics if there's something specific you want,

I did take a couple of the new mosfets, here's one with them removed and then installed. The brown is old Flux I cleaned off.

 

[Nobu]

First post has a bunch of naked pics if there's something specific you want,

I did take a couple of the new mosfets, here's one with them removed and then installed. The brown is old Flux I cleaned off.

Yeah, just wanted to see some of your handiwork.

 

[rhkcommander959]

Yeah, just wanted to see some of your handiwork.

Gotcha, I'll try to get better pics, the mosfet closest in the picture is angled about a degree or two off-square but the others came out perfect! I think it drifted after install but I scoped it out and the pins were all lined up to the pads, so good enough I guess!

I've seen some from the factory that were more crooked, I can tend to be a perfectionist but if it works and is solid I'm not going to bother it. I believe RazorWind said these mosfets might not like being reheated too.

 

[RazorWind]

I believe RazorWind said these mosfets might not like being reheated too.

They don't. They're real delicate. In fact, if you didn't use a preheater, there's a pretty good chance you cooked them just soldering them on the first time. It's a good sign that it works well enough to produce a picture, but I wouldn't be surprised if it dies after a minute or two of running any sort of high load application.

 

[Iratus]

this is such a cool thread.

My electrical knowledge can be summed up as “don’t lick that live wire” but I appreciate people experimenting. It’s like the early days of the [H] before all the skill went out of overclocking

 

[rhkcommander959]

They don't. They're real delicate. In fact, if you didn't use a preheater, there's a pretty good chance you cooked them just soldering them on the first time. It's a good sign that it works well enough to produce a picture, but I wouldn't be surprised if it dies after a minute or two of running any sort of high load application.

I preheated, and used a sacrificial board to dial all of my Temps in on the hot air and preheater to be as gentle as I could while getting good joints.

We'll see tonight.

 

[rhkcommander959]

So 3D test is running now, all default settings except fans 100%, probing Temps and watching them through software. Fingers crossed, so far so good.

I'm using kombuster, which says it has the gpu core at 99% itself, as well as gpuz. Core is ~55C in a room about 67F ambient. Power consumption is estimated to be only 85%, which I'm fine with honestly for a first go.

There is a "hot spot" sensor that reads 65C, where is it pulling that temp, VRM area?

Just passed the 5 minute mark, going to run it till dinner time. I don't want it running unobserved until I'm more confident in it.

 

[pendragon1]

where is it pulling that temp, VRM area?

its the core hot spot. on amd its usually ~20c difference.

 

[rhkcommander959]

Well got to 6 minutes before pizza showed lol, I'll try again later!

 

[rhkcommander959]

Went 20 minutes this time. The back of the pcb was uncomfortably hot, but behind the gpu core was nearly as hot so I think it's fine. Just being careful while I ramp up testing it. I forgot to grab my infrared thermometer....

See pic for results, core temp stayed in low 50s C

I didn't want to leave it unattended until I had more faith in it, I'm going to try loading it up longer and push for higher wattage when I can be nearby to monitor the situation. If anyone has ideas I'm all ears!

 

[rhkcommander959]

Is there anything able to get us closer to 100% power draw, msi kombuster won't go past 85% according to gpu-z, although kombuster says it's at 100%, perhaps a built in safety? Any better stress tests?

Ran it for an hour straight on kombuster without issue. I'll probably replace the thermal pads and call it a success.

The system I'm testing with is watercooled, so there is very little airflow on the back of the PCB, I would estimate it is about/up to 60C opposite the mosfets and core going off of touch, stable over the hour. I think the metal backplate was probably inhibitory or insulative since there were no pads, just an air gap.