adelli
Only real question I have are the VRM: did you have a chance of checking them?
-No I don't own an open test bench or a infrared thermometer...yet =]

Oh and what is your YouTube channel?
-Tech Professor (same logo) . I just started so really nothing on it yet. I am a college professor but tech stuff is my side passion project.
 
Question, was the first test with the tape on the stock or original TIM?  I ask because replacing the TIM alone can have substantial impact on GPU core temps.
- YES...forgot this part. Ugh, I am usually better at reporting so my results aren't suspect. I have 3 sets of data. Stock with Tape, TIM Replaced with Tape, TIM Replaced without Tape. I reported only Stock With Tape and TIM replaced without above. If I recall correctly, the raw temps for the results (TIM Replaced with Tape) I left out were ~69°C . I don't have access to the data in front of me right now because I am at work. 

Okay, so if I read you correctly your first test was with the stock thermal paste and tape and the second test you reported was without the tape and new paste.  Okay. That makes more sense for the thermal delta of ~5C or so. Blocking even a relatively small area can affect thermals significantly as even moderate dust can replicate the same effect. It may be they decided that redirecting some external airflow to other components was worth the trade off in marginally increased core temps.  That would take more testing to determine.

Yes - correct three tests in this order (1) Stock Card with Tape, (2) Replaced TIM with Tape, (3) Replaced TIM without Tape.
I will edit the post tonight with graphs (bar graph with mean temps & line graph with temps over time) when I get home. So the delta from just replacing the TIM was ~5 °C, and the delta from replacing the TIM + Removing the tape was ~5°C . 
 
I will also share the data to make it publically available tonight. I believe this is the most important aspect of testing, is independent verification and sharing data so we all can use it. Many tests in the PC tech community use the empirical method (really only descriptive data) whereas I am trying to make a push for open access to data and the use of the scientific method for testing.
 
Also - the VRM temps I do not seem to be that serious of an issue. There are thermal pads (1mm) on the VRMs and Memory covered by a black aluminum heat spreader (1mm) see pic below. You could replace the stock thermal pads with something fancier (i.e., Fujipoly Extreme XR @ 17W/mK), but anything around 6W/mK thermal pads would do just fine.
 

 
 
 
With a custom voltage curve, I can OC to +160Mhz on the clock at 1.00 V stable. I haven't pushed the card since removing the tape but will do more tinkering tonight. 
 
Also - the best part of removing the tape isn't the improvement GPU thermals. It is improved thermals inside your case. I noticed that all of my components were noticeably cooler. (Again, you can verify with data I'll upload tonight, so looks like the video will be delayed a bit).
 
 

adelli
Yes - correct three tests in this order (1) Stock Card with Tape, (2) Replaced TIM with Tape, (3) Replaced TIM without Tape.
I will edit the post tonight with graphs (bar graph with mean temps & line graph with temps over time) when I get home. So the delta from just replacing the TIM was ~5 °C, and the delta from replacing the TIM + Removing the tape was ~5°C . 
 
I will also share the data to make it publically available tonight. I believe this is the most important aspect of testing, is independent verification and sharing data so we all can use it. Many tests in the PC tech community use the empirical method (really only descriptive data) whereas I am trying to make a push for open access to data and the use of the scientific method for testing.
 
Also - the VRM temps I do not seem to be that serious of an issue. There are thermal pads (1mm) on the VRMs and Memory covered by a black aluminum heat spreader (1mm) see pic below. You could replace the stock thermal pads with something fancier (i.e., Fujipoly Extreme XR @ 17W/mK), but anything around 6W/mK thermal pads would do just fine.
 

 
 
 
With a custom voltage curve, I can OC to +160Mhz on the clock at 1.00 V stable. I haven't pushed the card since removing the tape but will do more tinkering tonight. 
 
Also - the best part of removing the tape isn't the improvement GPU thermals. It is improved thermals inside your case. I noticed that all of my components were noticeably cooler. (Again, you can verify with data I'll upload tonight, so looks like the video will be delayed a bit).
 
 

Okay, but you cannot have it both ways. You went from we need verified objective data to 'oh never mind I think the VRMs are good enough'... or some random thermal pad is 'just fine'.  Having done extensive testing professionally, that's not a good methodology. l hinted at that when I discussed reasons for redirecting airflow. You're not using the 'scientific method' when you substitute your opinion for data. This is why objective analysis is expensive and time consuming. You've only scratched the surface if you truly believe you're going to provide an objective analysis of the thermal design.  If your components are 'noticeably cooler with that small amount of redirected flow, I'd think the evidence would suggest you need a bit better airflow in your case...but that's just an opinion based on the evidence...

HeavyHemi
 
Okay, but you cannot have it both ways. You went from we need verified objective data to 'oh never mind I think the VRMs are good enough'... or some random thermal pad is 'just fine'.  Having done extensive testing professionally, that's not a good methodology. l hinted at that when I discussed reasons for redirecting airflow. You're not using the 'scientific method' when you substitute your opinion for data. This is why objective analysis is expensive and time consuming. You've only scratched the surface if you truly believe you're going to provide an objective analysis of the thermal design.  If your components are 'noticeably cooler with that small amount of redirected flow, I'd think the evidence would suggest you need a bit better airflow in your case...but that's just an opinion based on the evidence...

Thanks for pointing out the flaw in my logic. A healthy discussion is based on constructive criticism. In future posts, I will attempt to explicitly indicate what information is derived from data, what is an interpretation of data, and what is opinion (albeit informed or uninformed). I by no means have the equipment, facility, or time to perform a proper experiment on thermal design of a graphics card. This is why I included a measure of noise in the data (error) with a 95% confidence interval. Also, kudos for the appropriate use of methodology and methods. Many do not know the distinction and think methodology is just a fancier word when (s)he actually mean methods. 
 

adelli

HeavyHemi
 
Okay, but you cannot have it both ways. You went from we need verified objective data to 'oh never mind I think the VRMs are good enough'... or some random thermal pad is 'just fine'.  Having done extensive testing professionally, that's not a good methodology. l hinted at that when I discussed reasons for redirecting airflow. You're not using the 'scientific method' when you substitute your opinion for data. This is why objective analysis is expensive and time consuming. You've only scratched the surface if you truly believe you're going to provide an objective analysis of the thermal design.  If your components are 'noticeably cooler with that small amount of redirected flow, I'd think the evidence would suggest you need a bit better airflow in your case...but that's just an opinion based on the evidence...

Thanks for pointing out the flaw in my logic. A healthy discussion is based on constructive criticism. In future posts, I will attempt to explicitly indicate what information is derived from data, what is an interpretation of data, and what is opinion (albeit informed or uninformed). I by no means have the equipment, facility, or time to perform a proper experiment on thermal design of a graphics card. This is why I included a measure of noise in the data (error) with a 95% confidence interval. Also, kudos for the appropriate use of methodology and methods. Many do not know the distinction and think methodology is just a fancier word when (s)he actually mean methods. 
 

I apologize if I came off a bit strident there. But, I make a distinction between casual discussions over design preferences and likes and dislikes and empirical fact based discussions with objective data and defined metrics.  Personally I'm much happier with the casual conversations. It's too easy to nitpick results and the endless 'well what about...' and individual results are not a substitute for sampling of course.  Just in your example, half of your purported gain would have been just from the change in TIM. If someone had not asked my question, there would have been a misleading impression of a large temperature delta from the change versus the relatively moderate one in addition to not knowing the other effects of the change.  Anyhow, not trying to beat a dead horse but to illuminate a distinction between formal, 'scientific method' discussion, and the more typical casual discussions. 

As promised... link to data https://drive.google.com/file/d/1fKPPHQav3du0k47S5XEjnqxEXDWbRznO/view?usp=sharing
https://drive.google.com/...bRznO/view?usp=sharing

adelli
As promised... link to data...

Thank you so much.

I wonder if one could leave that front cooling plate and use an Arctic cooler...

Very nice job adelli!
Nice methodology btw.

the tape is to direct airflow to the vrm's, the temp differences you are seeing are from the TIM, since you are only seeing the gpu core temps, a 8~10c drop is not that unusual from using better TIM, and a few degree difference between applications is not unheard of either, it's the margin of error. the only real way to see the effect of that tape is to measure the vrm temps with a thermocouple in a controlled environment, which i'm sure evga did since adding something that actually makes temps worse would not happen.  

quadlatte
the tape is to direct airflow to the vrm's, the temp differences you are seeing are from the TIM, since you are only seeing the gpu core temps, a 8~10c drop is not that unusual from using better TIM, and a few degree difference between applications is not unheard of either, it's the margin of error. the only real way to see the effect of that tape is to measure the vrm temps with a thermocouple in a controlled environment, which i'm sure evga did since adding something that actually makes temps worse would not happen.  

Good point. Margin of error is important to consider to attribute results to each condition. There are two types of error: individual-level and condition-level (also known as intra-cell and inter-cell) with the latter being more useful for distinguishing and attributing effects. With one run, I was only able to obtain individual-level error.
 
What would be needed to appropriately distinguish effects of TIM replacement vs Tape removal and attribute them to thermal results, is condition-level error. This requires a larger sample size (i.e., more runs) to generate enough data to have a reliable error distribution for each condition. Around 25 per cell for a total sample size of 75 runs would be a decent start based on typical experimental designs (as well as previous experience as a scientist). 75 x 30mins = 37.5 hours of testing. This is one of the challenges with practicality vs accuracy that HeavyHemi hinted at earlier. 
 
I cannot make an informed opinion on the VRM issue. I agree that it is logical that the tape would be used to direct more airflow over the aluminum plate to assist with VRM and memory temps.