Gigabyte cooling performance *REP POST*

[Gigabyte-Gaming]

Because there’s been a lot of talk about cooling performance of 1080/1070 lately we would like to open a discussion about cooling capacity.
We don’t want to use our own shots for credibility’s sake so in this case we will use some from techpowerup to demonstrate our point.
This is the underside of the cooler on our GTX 1080 G1 Gaming:

Source: https://www.techpowerup.com/reviews/Gigabyte/GTX_1080_G1_Gaming/4.html

What is the thermal load on the cooler?
-Cooling GPU through direct touch with the heatpipes
-Cooling the memories using thermal pads that connect it to the same cooler
-Cooling the VRM and chokes with thermal pads that connect it to the radiator/heatpipes

The positive effect of all this is that memory, VRM and chokes are being actively cooled and kept well under 90º which is the safety guideline. In other’s designs you will see that these are being “incidentally” cooled by the excess air that goes through the fins of the cooler, this airflow is already been warmed up by the cooler. This last is a valid approach, but we think this is designed to pass media tests and achieve easy to measure numbers, not to achieve real every day use cooling performance.
The negative effect of our approach is that the VRM and chokes input more heat into the system. We believe this is a fair price to pay for the sake of longevity of the components even though a part of the price is paid in fan speed to achieve the desired GPU temp of <70ºC.


And this is the backplate on our GTX 1080 G1 Gaming (same source)

Another issue that we have also seen on media being misreported is temperature measurements of the back of the card with the presence of a backplate.
If you were to compare one backplate that has no thermal pads connecting it with the back of the components, this backplate will have little to no impact on the cooling performance of the card. Its use would be mostly for aesthetics, protection and to provide additional rigidity to the PCB.
On the other hand if this backplate is connected with thermal pads it becomes hot and it becomes a passive cooler that uses case airflow to dissipate heat. If you were to compare the temperatures of these two cases, the second one would be hotter. Some media report this as the card being hotter which makes no sense.
In summary: a hot backplate is doing its job, a cool one is not doing much in terms of cooling.

We would like to remind our customers in the UK that we have 3 years warranty and that we have a local RMA center located in Milton Keyes.

 

[Aretak]

You guys should send out some review samples of your GTX 1060 Mini ITX OC model. EVGA's Superclocked has been cleaning up in terms of word of mouth as the 1060 to buy, but having owned both, the ITX OC is the better card and runs a good bit cooler than the SC does at similar fan speeds, with the same small form factor. It's also cheaper than the EVGA card. Yet there's absolutely no reviews of it out there (and barely any information period), so it's hard to convince people of that fact.

 

[r7slayer]

This is what i like to see. A nice backplate which acts as a passive cooler. Ive noticed a good backplate with good contact and a slow RPM case fan blowing over it can often help keep temps down and help with keeping noise down too. So keep that up i like that!

Also i don't know Gigabytes stance on cooler removal but often ive always pondered at if AIB partners starting making it more easy to start attaching AIO coolers. However currently most AIB coolers have memory and VRM cooling built into the heatsink. Is there any possibility that that the memory and VRMs could have sort of a separate heatsink? So when you remove the cooler, the VRMs and memory chips are left with a heatsink on it of some sort? I don't like the sticky little lego brick thingys you can buy to put on them.

I like the idea of the NZXT Kraken g10 where you can add a AIO cooler but your often left with no heatsink on the VRMs and memory chips so the fan is just blowing on them which is a no no. Or you buy those stick on lego bricks which don't add much more thermal dissipation. Anyways food for thought i guess?

edit -
Also sorry for offtopic!

 

[pastymuncher]

The negative effect of our approach is that the VRM and chokes input more heat into the system. We believe this is a fair price to pay for the sake of longevity of the components even though a part of the price is paid in fan speed to achieve the desired GPU temp of <70ºC.

Myself and I am sure the vast majority on here, prefer the direct contact approach for cooling the vrm's. The indirect approach that relies just on airflow can lead to severe throttling of the card as seen in a lot of XFX's cards in the past. Keep designing the coolers to properly cool the card.

 

[r7slayer]

Myself and I am sure the vast majority on here, prefer the direct contact approach for cooling the vrm's. The indirect approach that relies just on airflow can lead to severe throttling of the card as seen in a lot of XFX's cards in the past. Keep designing the coolers to properly cool the card.

Something like this? For the VRMs is not good then? My 980Ti never seems to throttle and temps seem good for 1542 clocks. I was hoping for a more complete heatsink though for memory chips too. I know there is a small square patch of VRMS to the top left of that heatsink which the cooler attaches too but when it comes to third party cooling such as AIO coolers you have to cool them with those leggo bricks.

Spoiler

 

[orbitalwalsh]

Ideally the VRM plate should be copper as well to aid in maximum heat transfer . looks like it would only be a small pieve but guessing over thousands of units it would end up costing . maybe on flagship model?

 

[Gigabyte-Gaming]

You guys should send out some review samples of your GTX 1060 Mini ITX OC model. EVGA's Superclocked has been cleaning up in terms of word of mouth as the 1060 to buy, but having owned both, the ITX OC is the better card and runs a good bit cooler than the SC does at similar fan speeds, with the same small form factor. It's also cheaper than the EVGA card. Yet there's absolutely no reviews of it out there (and barely any information period), so it's hard to convince people of that fact.

Thanks for the suggestion. We've send it over for review in a couple of non English speaking markets and we've had good feedback as well.

 

[Kaapstad]

Interesting OP and thanks for posting Gigabyte-Gaming.

I am amazed at the number of times the so called tech review sites get things wrong and this is yet another example.

It is good to see things explained from the vendors point of view and the reasoning for doing them.

 

[Gigabyte-Gaming]

This is what i like to see. A nice backplate which acts as a passive cooler. Ive noticed a good backplate with good contact and a slow RPM case fan blowing over it can often help keep temps down and help with keeping noise down too. So keep that up i like that!

Also i don't know Gigabytes stance on cooler removal but often ive always pondered at if AIB partners starting making it more easy to start attaching AIO coolers. However currently most AIB coolers have memory and VRM cooling built into the heatsink. Is there any possibility that that the memory and VRMs could have sort of a separate heatsink? So when you remove the cooler, the VRMs and memory chips are left with a heatsink on it of some sort? I don't like the sticky little lego brick thingys you can buy to put on them.

I like the idea of the NZXT Kraken g10 where you can add a AIO cooler but your often left with no heatsink on the VRMs and memory chips so the fan is just blowing on them which is a no no. Or you buy those stick on lego bricks which don't add much more thermal dissipation. Anyways food for thought i guess?

edit -
Also sorry for offtopic!

About cooler removal, we don't endorse it. But it doesn't void your warranty unless you damage something.
If we separated the VRM cooling we would run into the problem that we are trying to avoid, that only incidental air is hitting it.
In the end I think it depends more on how good the AIO kit you put on. We understand your suggestion but we can't really build something for massive consumption and add small modifications so it can fit the needs of just a few users that plan to get rid of our cooler.

 

[sgavan]

This might be a bit stupid but would a full size thermal pad covering the whole board somehow connected to the fans/heatsink work, that way the heat from everything would be transferred away.

 

[Gigabyte-Gaming]

Myself and I am sure the vast majority on here, prefer the direct contact approach for cooling the vrm's. The indirect approach that relies just on airflow can lead to severe throttling of the card as seen in a lot of XFX's cards in the past. Keep designing the coolers to properly cool the card.

That's the difference between forum users who do their research and users who get convinced by flashy marketing campaigns. But it also media's fault to some extent by standardizing their tests so much that important details go unnoticed.

 

[Gigabyte-Gaming]

Ideally the VRM plate should be copper as well to aid in maximum heat transfer . looks like it would only be a small pieve but guessing over thousands of units it would end up costing . maybe on flagship model?

I'm not sure how much difference it would make. I think the quality of the thermal pads has probably a fair impact as well.

 

[Gigabyte-Gaming]

Interesting OP and thanks for posting Gigabyte-Gaming.

I am amazed at the number of times the so called tech review sites get things wrong and this is yet another example.

It is good to see things explained from the vendors point of view and the reasoning for doing them.

It's not that they get it wrong, but sometimes they focus too much on the numbers. Important things like cooling the VRM which can't be backup with numbers for technical reasons go a bit unnoticed.
For example Hexus mentions this on their review here


What's of more interest is how Gigabyte chooses to cool the remainder of the card. Other manufacturers, such as MSI and Asus, have their respective coolers make contact with the core alone; other heatsinks provide passive cooling for the memory and VRM circuitry. The idea is that the downward air pressure from the cooler's fans is enough to keep those components at reasonable temperatures.

Gigabyte, on the other hand, uses its tri-heatpipe cooler to provide direct contact and cooling by way of thick thermal pads between heatsink and other hot-running components. Its premise is that the RAM and choke/MOSFETs needs more than a small passive heatsink; the main cooler's TDP-wicking ability should be used instead. We'll see how this translates in the power and temperature section of the review.

 

[Gigabyte-Gaming]

This might be a bit stupid but would a full size thermal pad covering the whole board somehow connected to the fans/heatsink work, that way the heat from everything would be transferred away.

I think thermal pads are good at transferring heat on the "vertical" plane but not necessarily on the "horizontal", I don't know if you are getting my idea.

 

[r7slayer]

About cooler removal, we don't endorse it. But it doesn't void your warranty unless you damage something.
If we separated the VRM cooling we would run into the problem that we are trying to avoid, that only incidental air is hitting it.
In the end I think it depends more on how good the AIO kit you put on. We understand your suggestion but we can't really build something for massive consumption and add small modifications so it can fit the needs of just a few users that plan to get rid of our cooler.

Hmm yea i guess that makes sense. It would only be a small user base that would want to put some after market cooler on it. I just like the idea of AIO cooling as it brings temps down for the core greatly. I just realised you guys do the Waterforce. That actually has VRM cooling on it as well all though its only a coldplate. I am certainly going to get a AIO cooled card next i think and the waterforce cooler did actually impress me. Can't believe i forgot about that. All though im not sure how good the coldplate with copper heatpipe on it is at keeping VRM temps down?

 

[Gigabyte-Gaming]

Hmm yea i guess that makes sense. It would only be a small user base that would want to put some after market cooler on it. I just like the idea of AIO cooling as it brings temps down for the core greatly. I just realised you guys do the Waterforce. That actually has VRM cooling on it as well all though its only a coldplate. I am certainly going to get a AIO cooled card next i think and the waterforce cooler did actually impress me. Can't believe i forgot about that. All though im not sure how good the coldplate with copper heatpipe on it is at keeping VRM temps down?

Temperatures are a little bit higher than hybrid versions, but again the reason is that VRMs are contributing. On hybrid versions the blower is taking care of that. There's a pretty good video about it done by GamersNexus, he mentions this particular issue.

 

[RanxZy]

Temperatures are a little bit higher than hybrid versions, but again the reason is that VRMs are contributing. On hybrid versions the blower is taking care of that. There's a pretty good video about it done by GamersNexus, he mentions this particular issue.

Yeah, Steve at Gamers Nexus on YouTube done a complete teardown and showed how complete the cooling solution is. I think it's worth sacrificing a bit of core temps to have other vital components at lower temps.

 

[pastymuncher]

Something like this? For the VRMs is not good then? My 980Ti never seems to throttle and temps seem good for 1542 clocks. I was hoping for a more complete heatsink though for memory chips too. I know there is a small square patch of VRMS to the top left of that heatsink which the cooler attaches too but when it comes to third party cooling such as AIO coolers you have to cool them with those leggo bricks.

Spoiler

That's direct cooling as it has a heatsink on it. Indirect is when there is nothing on it and it relies on airflow across the naked vrm's to cool them. This is the worst method and very often results in throttling, all because the manufacturer can't be bothered or is too tight to cool them with a heatsink. I would never buy a card that uses indirect cooling.

 

[Gregster]

Well I have had my Gigabyte 1080 G1 Gaming since launch on a brand new 6850K build and whilst it isn't the best overclocker in the world, it runs at its boost clocks comfortably and quietly. Gigabyte did a good job on the card and acoustics and I am very happy with the temp

 

[Gigabyte-Gaming]

Yeah, Steve at Gamers Nexus on YouTube done a complete teardown and showed how complete the cooling solution is. I think it's worth sacrificing a bit of core temps to have other vital components at lower temps.

Exactly, but that's you noticing it, not every user will go over a teardown video like that. When you only look at numbers you would think a hybrid solution is better than what we did.