Overclock.net banner

[guru3d] EVGA FTW 1080 and 1070 Have Overheating Issues

26K views 356 replies 87 participants last post by  Bal3Wolf 
#1 ·
According to Guru, they noticed other models overheating as well.
Quote:
There is chatter on the web that EVGA FTW 1080 and 1070 graphics card are overheating resulting into a black screen. Initially people figured this to be the memory issue exposed last week. The company made a statement and issues a fix.

However reports from end users and some testing at Toms Hardware indicate overheating. Some users even reported the card dying completely. From the looks of it the issue has been uncovered due to the absence of any thermal pads over the VRM area of the FTW family of graphics cards. Some reports show temperatures running 107 ºC on components on the PCB.
Quote:
While we never reviewed the FTW editions, we did notice and made remarks on high load temperatures in the VRM area in our 1070 SC review. So we think that other models like the SuperClocked editions are effected as well. Our above thermal image was not measured with Furmark but with a 3DMark FireStrike loop.
http://www.guru3d.com/news-story/evga-ftw-1080-and-1070-have-overheating-issues.html
 
#5 ·
This is a very misleading article.

Quote:
The results in both tests show the temperature of PWM and memory is within the spec tolerance under the same stress test, and is working as originally designed with no issues.
Quote:
We agree that Furmark should not be used to stress a card (way too viral on the board and not emulating game load).
 
#6 ·
I just bought an EVGA GTX 1060 SSC GPU. Wonder if the 1060s are having this issue? I have yet to install the GPU and try it out, been too busy cleaning up the mess left behind after the hurricane.

I would send it back if I knew this was the case.
rolleyes.gif
 
#8 ·
how long have the 1070 and 1080 FTW card been out now? I feel like if this was a real more serious issue, it would be much more widespread. But articles gotta find something to post i guess..
 
  • Rep+
Reactions: LAKEINTEL
#9 ·
EVGA will probably give people with issue Titans or the new 1080TI
biggrin.gif
 
  • Rep+
Reactions: BiG StroOnZ
#13 ·
#14 ·
While FurMark certainly isn't representative of game loads, many, if not most, GPUs have under-engineered VRMs and/or VRM cooling issues.

Even if they don't like unrealistic loads, GPU manufacturers should be engineering their cooling and power delivery to handle absolutely the highest loads the parts can draw, indefinitely, plus a fair margin. It would cost all of five dollars per board to do so and would dramatically improve reliability and public perception of the products.
Quote:
Originally Posted by Smanci View Post

Not really
typer.gif


The funny thing is, these are subject to more indirect heat under normal CPU+GPU load. Gaming power consumption for the GPU alone often isn't too far off from furmark, and total system heat produced during gaming flies past furmark numbers.
The only thing non-GPU loads do to the GPU is increase the ambient temperature around the GPU. This does result in a warmer GPU, but no where near as much as the GPU itself needing to dissipate even a much smaller total amount of heat.

Increasing the load on a GPU by 20w will make it hotter than having a high-end OCed CPU dump 200w extra into a moderately well ventilated case.
 
  • Rep+
Reactions: CrazyElf
#16 ·
This is a great reason to get the FE version of the cards. They are heavily tested to pass strict standards, where as with some 3rd party cooling solutions, it's all about shaving a buck here or there, but providing a "visibly" better cooling solution.

When corners are cut, no amount of over engineering in other areas will compensate for that.
 
#17 ·
Quote:
I shouldn't need to cite a source for the color of the sky.

In a well ventilated case, virtually no heat from the CPU should be making it to the GPU. A small amount of heat is conducted through the motherboard to the card, and and even smaller amount of heat is radiated as infrared by the CPU heatsink on to the back of the video card, but the overwhelming majority of the heat produced by a CPU is going to be conducted into the air, which should be exhausted from the case before it can significantly increase ambient temperature in the vicinity of the card.

Test it yourself. Run FurMark at a fixed GPU fan speed while checking peak GPU power consumption and temperatures. Then, knock the power limit on the GPU down to about 20% below what you saw for the peak and test it again. Now, repeat the tests while running seven threads of Prime95 small FFTs on your CPU. If the power limited FurMark test isn't making the card warmer with Prime95 in the background than the uncapped FurMark test without Prime95 in the background, you've demonstrated my point. If running Prime95 during the power limited test does make your GPU hotter than more GPU load only, you have very poor case cooling.
 
#19 ·
Quote:
Originally Posted by Mad Pistol View Post

This is a great reason to get the FE version of the cards. They are heavily tested to pass strict standards, where as with some 3rd party cooling solutions, it's all about shaving a buck here or there, but providing a "visibly" better cooling solution.

When corners are cut, no amount of over engineering in other areas will compensate for that.
How much would omitting thermal pads save? I can't imagine their cost being very high.
 
#20 ·
Quote:
Originally Posted by lombardsoup View Post

How much would omitting thermal pads save? I can't imagine their cost being very high.
Cents per card. Apparently worth it when all sorts of limiters can be put in place and people told that anything exceeding common levels of load is a "power virus" and not supported.

All of which is tantamount to putting 100mph tires on a card advertised to do 150mph because the speed limit is only 65.
 
#21 ·
Quote:
Originally Posted by Blameless View Post

Cents per card. Apparently worth it when all sorts of limiters can be put in place and people told that anything exceeding common levels of load is a "power virus" and not supported.

All of which is tantamount to putting 100mph tires on a card advertised to do 150mph because the speed limit is only 65.
And EVGA will undoubtedly have to replace cards with burnt out VRMs...costing more than omitting the thermal pads.

Someone at EVGA didn't think this cost cutting through...
 
#22 ·
Not surprised.

My EVGA GTX 970 SC exhibits poor VRM cooling which affects the rather lackluster voltage section.

In my model, it seems EVGA ran a heat pipe off the GPU to like a quarter to half inch away from the voltage regulation transistors. So instead of the heat sink cooling both the GPU and voltage section well, the GPU dumps heat into or significantly hinders the cooling of the voltage transistors. So, even though it's a "high performance" ACX 2 cooler, the fans need to be ran over 55% to maintain stability. My GPU can be pushed further, but it's voltage locked, and the excessive heat causes the VRM section to thermal-throttle the available current. It's a bad heatsink design on their part. The GPU is never over 60 C, it's only the VRM section that overheats and throttles down to save itself (causing the card to crash).

Asus and MSI aftermarket GPU coolers have a separate heatsinks for the VRM section and the GPUs heat pipes aren't nearly as close to them in any case.

With the EVGA 1080 and the 1070, I hope it's something as simple as missing thermal pads and not another sub-optimal design on their part.
 
#23 ·
Quote:
Originally Posted by Derek1 View Post

The word on the EVGA forums from moderators is that this only effects cards with ACX 3.0.
All the others are supposedly unaffected with abnormal heat issues in the reported areas of the VRM, VRAM.

http://forums.evga.com/badHOT-PCB-Layout-on-GTX-1070-FTW-m2565921.aspx
This is a misleading statement. ONLY FTW cards are affected, because they use custom PCB. SC and regular cards use reference PCB.

https://www.reddit.com/r/TEAMEVGA/comments/58m8gi/evga_1070_sc_should_i_be_freaking_out_xpost_from/?st=iuogldgn&sh=fd0c98b3
 
#24 ·
Quote:
Originally Posted by keikei View Post

This is a very misleading article.

Quote:
The results in both tests show the temperature of PWM and memory is within the spec tolerance under the same stress test, and is working as originally designed with no issues.
Quote:
We agree that Furmark should not be used to stress a card (way too viral on the board and not emulating game load).
I do understand that this issue could be way out of context given that furmark stresses the GPU too hard. But usually GPU's overheat first, then the VRM. The VRM's dont have high temp shutoffs this is very convenient and safe. which is why this is a real issue and not just solved by not running furmark. Other legitimate software could damage the GPU unintentionally because of this.

Another problem is that high temps causes the VRM's to age drastically faster. If they dont have good thermal padding then I doubt they are of good quality.
 
#25 ·
Quote:
Originally Posted by lombardsoup View Post

Odd. Normally haven't had any problems with EVGA. Their quality control is usually very sound, but not this time apparently.
I think EVGA rushed the 10series release was late to the party then rushed when it was too late.

shame really
 
#26 ·
Quote:
Originally Posted by lombardsoup View Post

Quote:
Originally Posted by Blameless View Post

Cents per card. Apparently worth it when all sorts of limiters can be put in place and people told that anything exceeding common levels of load is a "power virus" and not supported.

All of which is tantamount to putting 100mph tires on a card advertised to do 150mph because the speed limit is only 65.
And EVGA will undoubtedly have to replace cards with burnt out VRMs...costing more than omitting the thermal pads.

Someone at EVGA didn't think this cost cutting through...
They did, they just forgot furmark was still a thing reviewers used.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top