970 Mod Failure! Where did I go wrong? (Kraken X31 via G10 Bracket) (Pics)

[WhiteGoblin]

Hello everybody! I have ended the life of a EVGA 04G-P4-3975-KR 970 SSC ACX 2.0+ card & I'm coming to ask you guys for help. Obviously I've done something wrong & I'd like your input so the mistake can be avoided in the future. Hopefully this thread will not only help me out but save some other people from these troubles in the future. To start out I want to say I've been running these amazing bios from over on the Overclock.net forums for three months with constant everyday use. I ran this card at 1532 core & 7906 memory with no problems! I never adjusted the power via sliders or anything else once the modified bios were put in place, I just used EVGA's PrecisionX 16 software to modify the speeds & run a more aggressive fan profile. I had tested this card at higher speeds via AIDA64 Extreme, 3DMark Pro, Cinebench 64, & Valley Benchmark where it did fine. I bumped it down to those speeds for a little bit of safety instead of putting it on the very edge all the time. At absolute worse it would get to 80C with the fans at 100% sounding like a jet was taking off in my office. (though it never downclocked)

 Skip forward a couple of months & I was really tired of the noise. Specifically when I was doing video editing, the scrubbing would make those fans rev up and down enough to make you pull your hair out. Or even while streaming on Twitch, no matter how good of a audio filter you setup the noise could become a problem at times. Wanting to solve two problems at once I figured I could help protect the long term life of the card & bring it's noise down by installing an AIO water setup on it. I have alot of NZXT products in my build so I figured I would stay with what's worked so far & purchased a Kraken X31, the G10 mounting kit, some Noctua NT-H1 paste, & 16 additional copper heatsinks. I was worried there was a lot of this card that needed cooling that the fan included in the G10 kit wasn't going to adequately take care of. I have a ton of air flow in my case so I thought with some well placed heat sinks I wouldn't have any issues keeping this card nice and quiet while it screamed on those modified bios.

 So I put everything together & it worked great for a while. I wanted to see how it would handle maximum stress so I fired up every benchmark I could. This card would never get over 55C, it was rock solid & finally silent. I was in heaven for about an hour. I tested some additional overclocks just for grins. The card would bench at 1600mhz core & 8,000mhz memory but after a while would start to stutter. I kinda laughed that even on water that was probably pushing it to far so I backed it down to 1550 on the core & everything was fine. I mean that's only 18mhz faster then I had it running on air for months. I really wasn't concerned one bit since this wasn't about getting it faster. I had a bunch of work to get done so I just let it run endless benchmarks in the background to see how it was going to handle the new setup. About 45 minutes into this I got a business call that took me away from my desk for about 20 mins & being naive or just really dumb I didn't for a second think to turn off my stress tests. I came back to a black screen, the system was still on but there was absolutely no video response.

 From that moment on this card was officially dead. I tried separate PCI Express slots, I ran completely different cables from different ports off my PSU, I took the card out & put it in another system all together. Then I took an old Quadro FX 3450/4000 SDI card I had laying around & tested everything. My system was fine, I couldn't find any damage to anything but the 970. No matter where I tried to run this card, the systems would never power on for more then a split second. I mean you could push the power button & watch the system fans just twitch once not even really move. That was the fate as I tested it over and over. I removed all my mods off the card completely, reinstalled the default stock cooling system, & well you guessed it, obviously still dead. I got a magnifying glass out just to see if there was any sign of physical abuse on the card, like I nicked something during the mod or I don't know, just if there was any sign of what had happen.

Nada, card looks great. I'm truly baffled as to what went wrong. I have speculated that a few of the tiniest chips I couldn't fit additional heatsinks on failed when they overheated. I'm also rolling the idea around in my head that I should have tried to install the X31 on top of the secondary plate that covered all the chipsets / memory instead of putting on all the copper heatsinks myself. I'm really not sure, I've highlighted the spots that didn't get new heatsinks in the pictures below. So I've got another of the exact same card coming. It'll be here Monday. Hopefully with all of your help I can install this X31/G10 on it & not ruin another $350 card. The sticky pads that came on the copper heatsinks have got to be dry by now & I didn't order any permanent bonding compounds yet as I wanted to hear what you all had to say before going down that route again. I watched a bunch of videos of people doing this same mod & clearly I've screwed up somewhere. ALL ADVICE IS WELCOME. Please use my mistakes as an educational grounds to improve your own future builds.

 

Pictures in the spoiler box below:

 

 

 

 

 

Thanks in advance,

'Gobby

[Liam-McShane]

Probably down to the fact the fan was blowing on absolutely nothing at all. You probably fried the VRM's.

[Kherm]

oh that's awful. 

[winningsince1337]

Probably down to the fact the fan was blowing on absolutely nothing at all. You probably fried the VRM's.

This. RIP 970. IMHO balls to the wall overclocking isn't very smart for long term use

[79wjd]

This. RIP 970. IMHO balls to the wall overclocking isn't very smart for long term use

It's fine if the card, and VRMs are cooled. 

[winningsince1337]

It's fine if the card, and VRMs are cooled. 

The VRMs had no fan blowing on them so I doubt they were. The fan on the G10 is towards the back of the car (away from the io bracket) but on this 970, the VRMs are on the other side with no airflow. His VRMs likely overheated and fried given they didn't have a fan blowing on them. 

[79wjd]

The VRMs had no fan blowing on them so I doubt they were. The fan on the G10 is towards the back of the car (away from the io bracket) but on this 970, the VRMs are on the other side with no airflow. His VRMs likely overheated and fried given they didn't have a fan blowing on them. 

Oh I know. I was speaking in general.

[WhiteGoblin]

Alright guys so today my second card came in! First off I want to say EVGA sent me a card in a very nice labelled box that was shrink wrapped. It's not a retail box but it's the nicest packaging I've ever seen an RMA come in. They actually shipped that box within another much bigger box that was filled with tons of packing peanuts. Good on them for amazing shipping. So I opened it up and was a bit dissapointed in what I found. The video card had been forced through the plastic holder that fills the whole box to hold the card snug during shipping. The bottom of the card was literally cutting through the plastic & the rubber protective sleeve was forced up in the corner of the packaging. I have no idea what happen to this card but that was a worrying start. When I pulled it out and inspected the physical quality of the card it looked like some techs had this thing kicking around in the office for months just getting tossed around. Some of the heatsink rows where bent & the card itself had a lot of scratches on it, as in, I don't think they'd ever take this thing back if you needed to RMA a second time.  :unsure: 

A bit shocked at the quality I plugged it in and everything fired up fine. I tested the ASIC quality & this card is 75.5% while my old one was 64%. If you're one to look into those numbers then that's a nice improvement. I took the card out & once again removed it's stock cooler but left the secondary plate on. To my surprise the X31/G10 fit on their perfectly. There was zero problems attaching it with the secondary plate still on the card. I really should've tested that last time but I was foolishly sure my copper heatsinks would do a fantastic job keeping everything in check. To address the fact that there's no direct air flow on the left side of the card where most the components are located I mounted an extra NZXT FN V2 120mm fan across it. I didn't have another 92mm to match the one that comes with the G10 kit perfectly but whatever I'm fine with this. It actually fits nicely from the hoses off the Kraken X31 to the side of the card & fits in the metal ridges the old stock cooler used. (oddly perfect fit)

I booted this thing up & immediately opened AIDA64 Extreme to watch the sensors tab. My VRM's were reporting 56C & my GPU was reporting 42C. I did a fresh driver install & fired up the benchmarks. Over the course of the next hour the reported temps of my VRMs never went over 60C while maxed & they idled at 56-57C. I also grabbed an IR thermometer & couldn't find any higher temps across the card. It appeared to be running extremely cool, which was making me feel more comfortable this second time around. On this setup the default bios pushed the card all the way up to 1532mhz itself which was a bit shocking to me. At those speeds without a modified bios I've seen tons of voltage crashes when gaming. (overclock.net has alot of articles about this across the maxwell architecture if you're interested) I went ahead and modded the bios again to make sure I wasn't going to come across any of those issues.

Back to the cranked bios I fired up benchmarks for another two hours while obsessively watching the sensors & checking with the digital thermometer. The GPU when maxed in benchmarks never exceeds 60C if I tell my Kraken X31 to run at max speeds across the pump & fans. (still silent btw) It idles at 45C in the middle of the day when my office is in the mid 80s. That's actually 5C hotter then my CPU runs but we're talking a huge difference in radiator size & fan quality. The VRM's still never go over 60C & still idle at like 57C. Given I can google plenty of 970 owners with VRM temps going all the way up to 100C I'd say I've done this correctly the second time around. Three hours of testing & my office has been silent. I'm surprised how noisy the foot traffic outside is now that I can hear. If anything changes I'll come back & inform you guys but for now I'm considering this issue closed. Please if anyone is doing this mod yourself then take notes from my journey.

Thank you to everybody who gave helpful advice! You guys rock! <3

'Gobby