Best Way to Cool my GPU

[Alences]

I have the Fractal Design Mechify-C case with 3 120MM fans intaking on the front, 2 120MM fans on the top intaking and 1 120MM fan on the back for exhaust. My CPU temps are fine during heavy loads but my GPU (MSI RX 5700 XT Gaming X) runs fairly hot most of the times while gaming. Average temps range from 85-90 depending on settings and the junction temperature will run at 109 and spike to 112 at times but never stays at that temp. MSI says those temps are fine and I should not worry about it as it is under warranty. I am contemplating taking the two fans from the top and getting some brackets off Newegg to make them fit in the PCI slot under the graphics card. What I don't know is if it will even make a difference from a cooling perspective. Crudely drawn fan placements below. The yellow is what I was thinking with moving the two tops fans below the GPU to force more air through it. Any thoughts on this setup?

 

[Clarkius]

If you're really worried about it (as MSI said, temps should be fine especially if under warranty), you could try a water cooled alternative if you think you're capable of installing it, or if you know anyone that would be able to install it. Otherwise if you already haven't, see if your GPU fan speeds can be increased at all. 

[Alences]

1 minute ago, Clarkius said:

If you're really worried about it (as MSI said, temps should be fine especially if under warranty), you could try a water cooled alternative if you think you're capable of installing it, or if you know anyone that would be able to install it. Otherwise if you already haven't, see if your GPU fan speeds can be increased at all. 

I should have mentioned this before and thanks for reminding me of that. I have already undervolted the GPU to 1150 and cranked up the fan speed. It usually sits around 2600 RPM under load. I do not feel comfortable with water cooling. I have built a few PC's, but never really gotten to deep into repairing hardware or doing anything close to water cooling.

[8tg]

How concerned are you with aesthetics? You can slap an Arctic accelero IV on it and drop temps to the floor for about 100$

 

edit: A user here details this a bit with a 5700xt

 

[Clarkius]

Just now, 8tg said:

How concerned are you with aesthetics? You can slap an Arctic accelero IV on it and drop temps to the floor for about 100$

Completely forgot this was a thing

 

[Ralfi]

With your front fans, are they controlled by the CPU temp or GPU temp?

 

If they’re all governed by CPU, then try splitting them if possible - have the upper front remain controlled by the CPU, but have the two lower fans be controlled by the GPU temps & see if that helps before buying new brackets etc.

[Alences]

51 minutes ago, Ralfi said:

With your front fans, are they controlled by the CPU temp or GPU temp?

 

If they’re all governed by CPU, then try splitting them if possible - have the upper front remain controlled by the CPU, but have the two lower fans be controlled by the GPU temps & see if that helps before buying new brackets etc.

I have only been able to change the setting and speeds on them through Dragon Center. I don't recall an option in there for having them controlled by CPU or GPU. How did you accomplish your setup?

[Jeppes]

Are your intake fans spinning and pushing air in in a way that it reaches the gpu? Because you are about 15C above the temps in reviews and there the card was also quiet.

 

Picture always helps as then you can spot rookie mistakes like not knowing which way fan pushes air etc.

[Alences]

I added 2 pictures of the fan set up. I also decided to load up Horizon Zero Dawn and captures the specs of it running. The fans in the front are definetly pushing air over the GPU. The top fans are not pushing as much air down but they are working well. Another reason I'm thinking about moving them under the GPU. Even before I put those on the top the CPU temps have always been good.

[Ralfi]

3 hours ago, Alences said:

I have only been able to change the setting and speeds on them through Dragon Center. I don't recall an option in there for having them controlled by CPU or GPU. How did you accomplish your setup?

I was able to use Gigabyte's software to control my fans, although I have all 3 daisy-chained & connected to the one header, so they're all controlled by the same temp sensor (CPU). If I wanted to, I could split the upper two from the lower one & do as I suggested to you, but I don't have a need to.

[WoodenMarker]

Try flipping the top fans to exhaust. The current configuration relies on a single rear case fan for removing most of the heat. 

[Alences]

3 minutes ago, WoodenMarker said:

Try flipping the top fans to exhaust. The current configuration relies on a single rear case fan for removing most of the heat. 

I can try that tonight and see how it works.

[Kinda Bottlenecked]

11 minutes ago, Alences said:

I can try that tonight and see how it works.

 

Set top fans to exhuast and remove that front psu shroud plate. 

[thrasher_565]

you can remove the pci slot covers too and posably add a fan to it.

[Alences]

2 hours ago, Alences said:

I can try that tonight and see how it works.

Got those fans switched to exhausting and no change. Horizon Zero Dawn was running at the same temps as when they were intaking. 

[Alences]

I even discovered that I made a mistake when building the computer and I put in the PSU upside down and it was venting into the case rather than out the bottom. I thought flipping that would make a difference but still the same temps even if the PSU is venting out the bottom.

[WoodenMarker]

25 minutes ago, Alences said:

I even discovered that I made a mistake when building the computer and I put in the PSU upside down and it was venting into the case rather than out the bottom. I thought flipping that would make a difference but still the same temps even if the PSU is venting out the bottom.

Most fans on mainstream psus are intake and not exhaust. 

[Kinda Bottlenecked]

if you remove the side panel and the cards still has high temps, then it would be an issue with the heatsink.

[Alences]

7 hours ago, Kinda Bottlenecked said:

if you remove the side panel and the cards still has high temps, then it would be an issue with the heatsink.

Do you mean the side panel on the case? I have never ran furmark before but can it detect thermal throttling due to high temps?

[Kinda Bottlenecked]

Just now, Alences said:

Do you mean the side panel on the case? I have never ran furmark before but can it detect thermal throttling due to high temps?

 

Yes, though I would use some thing like heaven or valley or superposition instead. 

If in those benchmarks you see high temps and and the gpu clock speeds starts going down then you'll know its throttling. AMD says 110*C is the max for rx 5700 xt.  

[freeagent]

Take the mesh off the top, I run mine meshless all around. I have the top of the case completely closed off and am running 2x 140s in the front, and a 120 in the rear. 3 out of my 5 fans rev to 2500 and higher, and the other two rev to about 1900. I would consider better fans, the stock ones are ok, but you can do much better.

 

I did have all holes stuffed with 120x38s for a couple of years.

[Jeppes]

Just remove that psu shroud piece from the front and maybe swap one of those thick 25mm fans to intake on front bottom slot as that is the most crucial one for gpu. And max out all of your case fans just for testing purposes although its most likely futile. I would say 99% chance for bad paste for the cooler and another incompetent MSI pr/customer relations person doing more damage to the company.

 

You could take a tiny bit more aggressive stance and ask them if they on purpose try to mislead consumers and send golden samples to reviewers as the difference is +16C and twice as loud. And please forward that email to all reviewers if they double up as MSI is likely to be very happy when every future review will have a mentioning of MSI lottery qualitywise.

[Alences]

1 hour ago, Jeppes said:

Just remove that psu shroud piece from the front and maybe swap one of those thick 25mm fans to intake on front bottom slot as that is the most crucial one for gpu. And max out all of your case fans just for testing purposes although its most likely futile. I would say 99% chance for bad paste for the cooler and another incompetent MSI pr/customer relations person doing more damage to the company.

 

You could take a tiny bit more aggressive stance and ask them if they on purpose try to mislead consumers and send golden samples to reviewers as the difference is +16C and twice as loud. And please forward that email to all reviewers if they double up as MSI is likely to be very happy when every future review will have a mentioning of MSI lottery qualitywise.

I'm not sure I can remove that. I did remove a portion to get the slim fan in the bottom slot. I am at work right now but can check when I am done.

 

I like the idea of doing that to MSI since their customer support has been crap on this issue. I did send the card in for RMA already once and they sent it back with just "Card OK" as the description. When I called to find out if they did anything they did not and the person on the phone basically said "So what it runs hot but it still works so use it until it breaks, you have 3 more years on the warranty." I was pissed and decided right then I would never buy anything from MSI again.

[Alences]

I thought I would share the results of the Furmark test I ran. Max temp 90C. I watched the GPU core clock speed closely and it was running mostly in the 1600 to 1700 MHz range but would drop down to the 800 to 900 MHz range once for a second and then come back up and then go back down a couple seconds later. I wish it would keep track of the clock speed better and be able to output that rather than just the highest MHz.

[Jeppes]

You should just enable the overlay from amd adrenaline to make it easier to see what clocks your gpu has while under load. Also should run a few benchmarks to see if it performs correctly. Like Unigine Superposition 1080p extreme where it should score near 5000 even when ran 5 times in a row. That is the easiest way to see if it performs correctly.