GTX 1080 TDP throttling WHY?? [SOLVED]

[andydabeast]

Hi guys, Gigabyte windforce card here under water with a 1600x at 4.0 and corsair AX860 PSU.
I have a TDP throttling issue that only happens in some games and started happening more once I put the GPU block on. Here is what happens when a load is applied-

-Temp doesn't go over 33C
-clock speed is ok, ramping up to max
-memory speed is ok, ramping up to max
-GPU load is ok
-TDP won't go beyond 38%
-GPU-Z does not give a perf. cap reason, it thinks the card is loaded maximum

I tried stock OC and some other combinations with no change. Tried messing with all the overclocking settings using afterburner or EVGA Precision. I can turn the TDP limit down and if below 68% it will obey, but there is no difference between setting it to 70% or 108% (108 is maximum). Stock settings does the same thing.

I am using HDMI to my main monitor and DP to DVI to a secondary. I have tried only using one.

I googled around a lot and found lots of people whose clock speed was stuck but none with TDP.

MSI Afterburner is up to date, Drivers are 391.35 (fresh remove and install), windows 10 is up to date, I am on an SSD. I tried messing around with Nvidia settings for power management, nothing changes.

I am using a riser but have been for months with no issues. GPU-Z reports pci-e 3.0 @ 16x

If I restart my computer I have a couple minutes to launch a game and it works fine but if I wait too long it throttles.Sometimes while playing a game just fine it will decide to throttle. Attached is an image of while I am in game on the other monitor while throttling. Ignore the maximum values because it wasn't throttling before. This is how I know it is software not hardware, it happens randomly.

Anyone know what it could be? Thanks!

New Picture.bmp

[Tabs]

What block are you using? You could be having throttling if your waterblock isn't cooling the VRM's. The card itself monitors the temperature of the VRM, but as far as I know it's impossible to view VRM temperatures within windows. Everything could look absolutely fantastic, but still have your VRM's toasting if there isn't any airflow (or direct block contact).

 

 

[andydabeast]

3 minutes ago, Tabs said:

What block are you using? You could be having throttling if your waterblock isn't cooling the VRM's. The card itself monitors the temperature of the VRM, but as far as I know it's impossible to view VRM temperatures within windows. Everything could look absolutely fantastic, but still have your VRM's toasting if there isn't any airflow (or direct block contact).

 

 

It is a EK block. I no not have a backplate or much airflow. It was my first block install so I went very slowly and extremely carefully with the installation. When everything works the card gets into the low 40's and I can play for HOURS. How can I play for 4+ hours with the core at 45C but after being idle it throttles while the core is at 30C? I will point a fan at it but thermals aren't an issue.

[Tabs]

1 minute ago, andydabeast said:

It is a EK block. I no not have a backplate or much airflow. It was my first block install so I went very slowly and extremely carefully with the installation. When everything works the card gets into the low 40's and I can play for HOURS. How can I play for 4+ hours with the core at 45C but after being idle it throttles while the core is at 30C? I will point a fan at it but thermals aren't an issue.

Sorry, I must have misunderstood the problem then - so at load, you have no problem with your card?

 

What problem do you have at idle that's got you concerned? I promise I read your post multiple times but based on what you just said I can't figure out what issue you have mate.

[Arika]

17 minutes ago, andydabeast said:

-Temp doesn't go over 33C
-clock speed is ok, ramping up to max
-memory speed is ok, ramping up to max
-GPU load is ok

so what's the issue? what makes you think it's throttling?

[andydabeast]

2 minutes ago, Tabs said:

Sorry, I must have misunderstood the problem then - so at load, you have no problem with your card?

 

What problem do you have at idle that's got you concerned? I promise I read your post multiple times but based on what you just said I can't figure out what issue you have mate.

sometimes it throttles, sometimes not. I need to put load onto it to find out if it is throttling because TDP is only at 7% just sitting on chrome. If I launch a game and I have 5 frames and 28% TDP I know it is throttling.

Just now, Sierra Fox said:

so what's the issue? what makes you think it's throttling?

see above. Any game is unplayable with 500+ ms lag on the mouse, 5-10 frames at most, and the card doesn't get hot and only at 30% TDP

[andydabeast]

22 minutes ago, Tabs said:

What block are you using? You could be having throttling if your waterblock isn't cooling the VRM's. The card itself monitors the temperature of the VRM, but as far as I know it's impossible to view VRM temperatures within windows. Everything could look absolutely fantastic, but still have your VRM's toasting if there isn't any airflow (or direct block contact).

 

 

 

HOLY BALLS YOU ARE RIGHT.

See the picture. I left a fan on pointing at the back of my card while idle. Then I ran heaven. See the TDP and also Memory controller load decrease in GPUZ. at the low point I turned the fan on and that is when they went back up!

So is a metal EK backplate my solution? Still doesn't explain why I can play sometimes.

New Picture (1).bmp

[Tabs]

The VRM's aren't actually on the back of the card though, so that's likely only a temporary/hotfix solution. If your block is a full-cover block which (SHOULD) cover the vrm's, it's possible that the heat pads to cover the vrm's have either slipped (so is not actually providing heat transfer for the right components), or isn't contacting between the VRM and the waterblock.

 

If the waterblock is designed for a founders edition card but you're using it on a Windforce card, it's possible that some hot VRM components are in the "wrong" place and are therefore not being covered by the block.

 

The reason this could happen sometimes is pretty easy to explain - every game or program will stress your card in different ways. If you play one game and it pushes your vrm to 110C (which is shockingly a "standard operating temperature" for decent VRM's), your card will have no issues. If you start playing another game that has more complex geometry or something, or perhaps utilises some GPGPU functionality, or has a lot more gpu-vram communication, suddenly one VRM might be pushing 120+ and causes throttling of the whole card to prevent damage.

 

Edit, sorry, to answer your question directly: No, a backplate is not your solution as backplates do not meaningfully contribute to the cooling of a GPU. You need to reseat your block with appropriate thermal padding between it and all of the VRM's to ensure they get some cooling from the block as well, or to permanently mount a fan if the block is semi-coverage to ensure the VRM's get some airflow.

[andydabeast]

14 minutes ago, Tabs said:

The VRM's aren't actually on the back of the card though, so that's likely only a temporary/hotfix solution. If your block is a full-cover block which (SHOULD) cover the vrm's, it's possible that the heat pads to cover the vrm's have either slipped (so is not actually providing heat transfer for the right components), or isn't contacting between the VRM and the waterblock.

 

If the waterblock is designed for a founders edition card but you're using it on a Windforce card, it's possible that some hot VRM components are in the "wrong" place and are therefore not being covered by the block.

 

The reason this could happen sometimes is pretty easy to explain - every game or program will stress your card in different ways. If you play one game and it pushes your vrm to 110C (which is shockingly a "standard operating temperature" for decent VRM's), your card will have no issues. If you start playing another game that has more complex geometry or something, or perhaps utilises some GPGPU functionality, or has a lot more gpu-vram communication, suddenly one VRM might be pushing 120+ and causes throttling of the whole card to prevent damage.

 

Edit, sorry, to answer your question directly: No, a backplate is not your solution as backplates do not meaningfully contribute to the cooling of a GPU. You need to reseat your block with appropriate thermal padding between it and all of the VRM's to ensure they get some cooling from the block as well, or to permanently mount a fan if the block is semi-coverage to ensure the VRM's get some airflow.

The card is Windforce which EK says visually fits the G1 gaming block. The stock cooler had a plastic backplate with thermal pads on the back of the VRM. I also really thought that a full cover block would cool it properly. I put the thermal pads onto the VRM myself.

Because the fan pointed at the back of the card being on or off produces a direct, noticeable, repeatable difference in performance I think that a backplate with thermal pads would solve it.

How would I even know after taking the block off if the VRM's were properly being touched?

 

I have flexible tubing. I could take the block off without draining the loop? Cryofuel in there...

[Tabs]

2 minutes ago, andydabeast said:

The card is Windforce which EK says visually fits the G1 gaming block. The stock cooler had a plastic backplate with thermal pads on the back of the VRM. I also really thought that a full cover block would cool it properly. I put the thermal pads onto the VRM myself.

Because the fan pointed at the back of the card being on or off produces a direct, noticeable, repeatable difference in performance I think that a backplate with thermal pads would solve it.

How would I even know after taking the block off if the VRM's were properly being touched?

 

I have flexible tubing. I could take the block off without draining the loop? Cryofuel in there...

Take a pic of the card without the block and I (or anyone else) can circle the VRM's. You'd then be able to visually see if the block was making contact (since the pads would have indents on it to match the VRM's). If you do decide to de-block your card, remember to take a pic of the block (underside) as well, so that we can easily compare which parts are properly making contact and which aren't. You shouldn't need to drain your loop if you're careful when you remove the block.

 

Alternatively, if you have a picture of the bare pcb, or know the EXACT model of card and block you're using (so we can try to find a pic of it) we can point out where the VRM's are to try to indicate which VRM's are making contact and which aren't.

 

EK may state that the card visually fits, but the likelihood is that they are simply checking for blocking issues, like capacitors in the wrong place that will physically prevent mounting. If the card has VRM's in the wrong place (they are low profile surface mount components), they wouldn't necessarily prevent the mounting of a block even if they were in a different place.

[m0k]

sound more like a driver issue

ive had an issue in the past with my gtx 1060 where it would not use more than 40%ish in certain games.

i just reinstalled the nvidia drivers and it fixed the issue (with ddu, i might have had to reinstall twice or something, it didnt fix it in the first attempt)

i dont think it has to do with tdp or power or cooling since everything else seems fine in your system

[Tabs]

3 minutes ago, mok said:

sound more like a driver issue

ive had an issue in the past with my gtx 1060 where it would not use more than 40%ish in certain games.

i just reinstalled the nvidia drivers and it fixed the issue (with ddu, i might have had to reinstall twice or something, it didnt fix it in the first attempt)

i dont think it has to do with tdp or power or cooling since everything else seems fine in your system

 

He already tested putting extra airflow over his card, and the problem receded. It's 100% NOT a driver issue. 

[andydabeast]

9 hours ago, Tabs said:

Take a pic of the card without the block and I (or anyone else) can circle the VRM's. You'd then be able to visually see if the block was making contact (since the pads would have indents on it to match the VRM's). If you do decide to de-block your card, remember to take a pic of the block (underside) as well, so that we can easily compare which parts are properly making contact and which aren't. You shouldn't need to drain your loop if you're careful when you remove the block.

 

Alternatively, if you have a picture of the bare pcb, or know the EXACT model of card and block you're using (so we can try to find a pic of it) we can point out where the VRM's are to try to indicate which VRM's are making contact and which aren't.

 

EK may state that the card visually fits, but the likelihood is that they are simply checking for blocking issues, like capacitors in the wrong place that will physically prevent mounting. If the card has VRM's in the wrong place (they are low profile surface mount components), they wouldn't necessarily prevent the mounting of a block even if they were in a different place.

I do know the exact model of the card and block, so I will get pictures ready and post back here. Basically, I will exhaust all my options before removing the block. The GPU is vertically mounted in a scratch build and I need to remove the CPU block in order to get the GPU out. Not an easy task. At least it's not hardline. 

I have been dealing with this issue for over a month, and yall found [what we think is] the issue in minutes. Fantastic community. 

[andydabeast]

I verified with my amazon order history that my GPU is the GV-N1080WF3OC-8GD model. The block I bought is EK-FC1080 GTX G1 - Nickel.
The below page is the EK page where I saw that this block is compatible with the card. My card is the bottom one.

https://www.ekwb.com/configurator/waterblock/3831109831373#DB_

I took a picture of my PCB right after removing the stock heatsink. I grabbed it from Facebook, cropped it, and put it under the image of the G1 PCB from EK site. They look the same to me, which points to user error in block installation. 

I do know that the plastic backplate my card shipped with had thermal pads on the back of the VRM... To transfer heat to the plastic? 

 

Given this information do yall think I should just try to reinstall the block or put a heatsink or backplate on the back of the VRMs?

 

I happen to have some heatsinks that are VRM size so maybe I just spend $10 on amazon for some thermal double sided adhesive tape, slap those heatsinks on the back and see if it works with no fan. 

[Toysrme]

If DDU and a full driver re-install doesnt fix it. The riser is spotty. Happens all the time mining. Replace the riser & from now on make sure it's molex supply only. Don't use SATA power connectors on your mining riser. They are half the spec current that a card will actually use. Try to use gen6+ molex only risers. They're much better.

[andydabeast]

12 minutes ago, Toysrme said:

If DDU and a full driver re-install doesnt fix it. The riser is spotty. Happens all the time mining. Replace the riser & from now on make sure it's molex supply only. Don't use SATA power connectors on your mining riser. They are half the spec current that a card will actually use. Try to use gen6+ molex only risers. They're much better.

My riser is non-powered and off brand, purchased before the big companies started making them. How could it be the riser if turning on a fan fixes it?

Also, this happened less before I put on the block. I also had this riser with another card and that one was fine.

[andydabeast]

In this video here he has the G1 gaming model but he just puts his backplate right back on... It covers two parts, the three memory modules farthest away from the IO, and one row of VRM that are ON the back of the board. I think this is the key. I don't need to cool the back of the board, I have a whole row of power delivery chips back there!!

 

 

[Damascus]

Looks like the card has vrm controllers on the back, they might be what's throttling here.

[andydabeast]

I guess I should head over the Gamer's Nexus and learn more about PCB components. Memory controller load was throttling in sync with the TDP throttling. What does the memory controller look like?

[andydabeast]

I'll have pics tonight but my tubing kept hitting the heatsinks and now they won't stick on fully, so I am unsure if it was a success or not. The card is close to the motherboard chipset so with a 1/2" thick piece of cardboard I will let the card and heatsinks rest on the cardboard to keep the heatsink in place. I NEED to know if this fixes it. 

 

Then I will see how sticky the thermal pads are from the original backplate and do one of two things-

A. put the original backplate back on and attach the heatsinks to the bacplate. It is plastic, but it does something.

B. get a sheet of aluminum and cut a square to cover the same area as the thermal pads from the stock backplate. Then attach stock pads and new metal and my heatsinks to the back of the card. This would also allow me to not have to take out all the waterblock screws. I am not sure if I have any metal laying around so I would have to buy some. 

 

Thoughts?

[andydabeast]

The problem is solved by using the stock backplate with added heatsinks. I will never know if it was the line of VRM or the three memory modules that needed the cooling, but it works now.

[Mick Naughty]

Should add a waterblock on the back of the card too. Super cool temps and all.

[Jaqen]

might be worth emailing EK so they can advise future customers of this....

[andydabeast]

8 hours ago, Mick Naughty said:

Should add a waterblock on the back of the card too. Super cool temps and all.

That would be nice but I don't have the space and the cost wouldn't be worth it when I already had those $5 heatsinks laying around. 

 

19 minutes ago, Jaqen said:

might be worth emailing EK so they can advise future customers of this....

EK recommends you use their backplate so I imagine that would have solved it. Theirs is metal and this one is plastic. 

 

All the dozens of threads on the internet about how backplates are aesthetic and don't help much are not true in all cases though. I have proved that. 

[andydabeast]

Also, See the thermal imaging here. This is for the G1 card but nearly the same as mine. The backplate gets to 60C, but that is heat that came through a 1/8" thermal pad and 1/16" of plastic. 

 

http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_1080_g1_gaming_review,10.html