EVGA GTX FTW series becomes FTL due to VRM overheating issues

It’s not often that you get a factory fault within a range of GPUs that necessitate large scale action on behalf of the manufacturer. We get different revisions ironing out kinks and maybe more efficient/cost reducing designs, but nothing really of any note. The last hoo-ha made in recent memory was the noisy water cooler on the Radeon R9 Fury X, and that got sorted out pretty timeously.

Now, however, it’s EVGA’s turn to be in the spotlight for an apparent oversight regarding adequate cooling design on their ACX series coolers in their Pascal line-up, namely the GTX1070 and GTX 1080 FTW series. The thermal issues, initially pointed out by the German edition of Tom’s Hardware, found last week that there was something not cool with, well, keeping things cool. They were reporting a maximum VRM temperature of 107-114C, which although were still within operating range (Max operating temperature for the VRMs is 125C), it did not give much room for variation in temperatures—remember  Tom’s Hardware tested in a controlled and air-conned environment, in an open test bench.

Although EVGA have acknowledged the problem on their forum with a view to things being blown out of proportion. However, GamerNexus do make the claim that the problem is so distinct and design based, that the type of case cooling can affect VRM temperatures by at least 5-7 C. The apparent reason is a lack of thermal interface between PCB baseplate and ACX heatsink for effective heat dissipation. This lack of or poor implementation of a thermal interface, such as thermal pads, between the VRMs/VRAM on the PCB and the baseplate meant that heat would end up trapped in this area with little or no airflow.

This brings the fault tolerance of the VRMs under serious stress, since unlike modern GPUs–which throttle their core speed and voltage when encountering thermal limits– the VRM MOSTFET will approach breaking point and once reached it “will thermal runaway and burn out with no warning”. Although possibly only affecting a insignificant amount of users, there seems to be a few people to have experienced this critical MOSFET failure resulting in their EVGA GPU letting go the smoke ghost.

To address this issue, EVGA performed their own internal testing and found that their cards performed fine without the thermal pads, although once tested with additional thermal pads between the backplate and PCB and baseplate and heatsink fins, it resulted in significantly better temperatures.

However, in light of the review from Tom’s Hardware, and probably to save face, EVGA has offered a free thermal pad mod (shipping included) for all EVGA owners to install on their EVGA cards because they want their users to have a lower temperature. Of course this is the correct action to take, even though its making the buyers who’ve spent hundreds of dollars have to take apart their GPU and apply thermal pads, with some perhaps never having done such a thing before. At issue here is the frequent request on the forum for at least a video or post detailing steps to add the new thermal pads; this would go a long way to allaying EVGA customers’ fears of correctly adding thermal pads. I’ve reached out to EVGA support with questions regarding warranty and issues regarding users who might have no experience doing this sort of customisation mod, but at the time of publishing I’ve not heard back.

Although the lack of proper cooling design should not have made it past any sort of QA, the quick turn-around and generally customer-friendly solution by EVGA should be commended, with an easy to fill out form requesting the thermal mod. If anything, think of this as a way to legitimately open up your EVGA card without voiding warranty and perhaps adding better thermal paste or put it under a custom water loop.

Last Updated: October 31, 2016