Replacing the thermal paste on a Sapphire Pulse 5700xt

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
I have a Sapphire Pulse 5700XT that I keep underclocked/under volted (1900mhz, 1080mV) because the performance is very close to stock but the fan noise and heat are significantly less. While my pc is completely silent during normal operation, it is somewhat audible during a full gaming load. To this end, a fellow forum member said he used Thermal Grizzly Kryonaut to replace the stock thermal paste. You remove the screws to the backplate, fan shroud, and heatsink, remove the old paste, put the typical rice dot in the middle, and put all the parts back together again. 15 minutes of work really. He said his core temps and hot spot (junction/edge) temps were reduced by 10C.

I decided to give this a try today using Noctua NT-H1, an excellent thermal paste but not as good as TG Kyro. Reviews show that, at most, it will be 1-3C worse. So I am expecting 9-7C improvement. The results were not great. In FurMark, I see that the min temp increased from 38 to 45, Max stayed the same at 63, but average was reduced from 57C to 55C. In COD MW, Core Max was reduced from 75C to 74C, but junction increased from 90C to 93C.

This temps are still fine, but nowhere near a 10C improvement. Has anyone else had any experience replacing the stock thermal paste of their 5700XTs?
 

pendragon1

Fully [H]
Joined
Oct 7, 2000
Messages
20,345
not on my 5700 but it helped on all other gpus ive done, 5-10c. maybe you didnt use enough paste or screws arent tightened right. did you check for coverage before reassembly?
 

THUMPer

2[H]4U
Joined
May 6, 2008
Messages
3,327
A few things to try. Maybe a little more paste to get good coverage. Make sure screws are evenly tight, and increase fan speed on the gpu. Each chip is different and may require a few tries to get a good result.

With that said, are you able to lower voltage anymore and remain stable?
 

Aegir

Limp Gawd
Joined
Apr 15, 2020
Messages
202
Don't forget to boil water and let your Kryonaut tube sit in the boiling water.
Microwave water in a cup, and then put in your tube, and let it sit until hot.

Then use a heatgun on low or a hairdryer on highest heat to heat up the heatsink. Then squish the hot paste with the hot heatsink, and really let it melt perfectly flat.

I did that on a Vega 64, and the temp decrease was incredible.
It's a Vega 64, so it still gets VERY hot, but it's still way better than it was.
And that was replacing MSI's liquid metal paste that was on by default. Yeah, Kryonaut beats stock-liquid-metal. Not bad.

Just remember that you only have ONE chance to place the heatsink on perfectly flat, and you have to do it FAST before the paste and heatsink cool down.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
not on my 5700 but it helped on all other gpus ive done, 5-10c. maybe you didnt use enough paste or screws arent tightened right. did you check for coverage before reassembly?
A few things to try. Maybe a little more paste to get good coverage. Make sure screws are evenly tight, and increase fan speed on the gpu. Each chip is different and may require a few tries to get a good result.

With that said, are you able to lower voltage anymore and remain stable?
I did not check for coverage because I assumed I did it right. The GPU die is significantly smaller (half the size?) of a CPU Integrated Head Spreader (IHS). I used the same rice size drop I use for CPUs on the GPU die, so coverage should not have been an issue. When it came time to assemble the heatsink back onto the pcb/die, there is an X bracket that connects the two. I inserted the screws and tightened in a ADBC/ diagonal pattern. I fully tightened - via a micro screwdriver -- all screws. Curious, if I were to open it all up again, what would be the mark of bad coverage?

As for voltage, no, I am not checking for lower voltage. I am fine with the voltage/clocks, I just want to reduce heat now. I thought the thermal paste would work, but so far not?
 

Aegir

Limp Gawd
Joined
Apr 15, 2020
Messages
202
Try using more thermal paste, and look at some videos that show how thermal paste spreads under transparent glass.
Based on what I've seen, people don't actually use enough thermal paste. The large dot method seems to be best. Not lines or X.

But you could probably get away with using less paste if you heat it up first so that it flows better, and melts into the microscopic pits and grooves of a heatsink/CPU.
 

pendragon1

Fully [H]
Joined
Oct 7, 2000
Messages
20,345
Curious, if I were to open it all up again, what would be the mark of bad coverage?
the whole die isnt covered and thickness being uneven would be bad/uneven pressure. on a gpu there isnt really a "too much" with non conductive paste.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Gotcha. Ok, i'll try again. This time, I'll take a picture of what I added the first time. Then i'll take a 2nd pic of what I added again
 

THUMPer

2[H]4U
Joined
May 6, 2008
Messages
3,327
Well if you lowered voltage some more then it would produce less heat. Also try a more aggressive fan curve of you haven't. I've had 4 5700s that I've used, all of them had a delta of about 20-25c between the core temp and hot spot.
 
Last edited:

Aegir

Limp Gawd
Joined
Apr 15, 2020
Messages
202
Feel free to post a full description of your technique.
This topic isn't explored very much, so by the end of it, you'll be an authority on the subject. =p

So I'd love to know your results in detail.
 

pendragon1

Fully [H]
Joined
Oct 7, 2000
Messages
20,345
Feel free to post a full description of your technique.
This topic isn't explored very much, so by the end of it, you'll be an authority on the subject. =p

So I'd love to know your results in detail.
what are you talking aboot? there is tonnes of info about repasting a gpu, plenty of threads here about it even. we've been doing it since gpus got heatsinks.
 

Aegir

Limp Gawd
Joined
Apr 15, 2020
Messages
202
what are you talking aboot? there is tonnes of info about repasting a gpu, plenty of threads here about it even. we've been doing it since gpus got heatsinks.
Yeah, but more information is good. I like to know details. There is a lot of misinformation, and having statistics and details noted helps.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Well if you lowered voltage some more then it would produce less heat. Also try a more aggressive fan curve of you haven't. I've had 4 5700s that I've used, all of them had a delta of about 20-25c between the core temp and hot spot.
Lowered voltage isn't stable at the already lowered clocks I am at. And the fan curve is already custom/aggressive enough.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
So, I just did this all over again, with the 2nd time going alot quicker (obviously). I noticed when I removed the large heatsink that there was paste across the entire GPU die. It seems pretty well coated. And look at the raised edges on the heatsink, thats a perfect indication that enough paste was added the first time.

However, as you all said, you can (mostly) never add enough. So I added maybe 30% more than what I added last time. I redid the FurMark bench (30 min) and the temps were exactly the same. In-application said the max was 63C. Open HW Monitor said Core max was 63C and Junction was 73C - a whole 1C improvement for core and a 4C improvement for junction. I don't have time tonight to play COD, but I have a feeling it will be more of the same.

Maybe the thermal paste that was stock was pretty decent already? There surely can't be this huge of a delta between Noctua NT-H1 and TG Kyro. Maybe my case already has excellent airflow (2xNoctua 140mm intakes, 1x120 Noctua exhaust) that heat wasn't a big issue?
 

Attachments

Aegir

Limp Gawd
Joined
Apr 15, 2020
Messages
202
It might just be the thermal capacity or heat->air speed of the heatsink itself then.

Even if you transfer heat from the GPU to the heatsink at a decent speed, the bottleneck might actually be getting heat from the heatsink to the air.

And also, if you DID notice a slight heat decrease on the second try, it doesn't mean it will run cooler.
It might just imply that it will cool *faster,* in the sense that once the usage of the GPU stops, it will go from 73C back down to 30C faster than before.

Beyond just that, it might be *boosting* for longer, and simply running as fast as it can until it hits the thermal limit, which means you'll actually get better performance.
On the downside, it means that unless you set the thermal cap or GPU usage cap lower, it will always boost until it hits about 73C, and thus, you'll never get it cooler.
 

Furious_Styles

[H]ard|Gawd
Joined
Jan 16, 2013
Messages
1,830
So, I just did this all over again, with the 2nd time going alot quicker (obviously). I noticed when I removed the large heatsink that there was paste across the entire GPU die. It seems pretty well coated. And look at the raised edges on the heatsink, thats a perfect indication that enough paste was added the first time.

However, as you all said, you can (mostly) never add enough. So I added maybe 30% more than what I added last time. I redid the FurMark bench (30 min) and the temps were exactly the same. In-application said the max was 63C. Open HW Monitor said Core max was 63C and Junction was 73C - a whole 1C improvement for core and a 4C improvement for junction. I don't have time tonight to play COD, but I have a feeling it will be more of the same.

Maybe the thermal paste that was stock was pretty decent already? There surely can't be this huge of a delta between Noctua NT-H1 and TG Kyro. Maybe my case already has excellent airflow (2xNoctua 140mm intakes, 1x120 Noctua exhaust) that heat wasn't a big issue?
That looks fine to me, you perhaps used just a tad too much, but that won't matter if you have it screwed down evenly. And from what I've found there is very little difference between NT-H1 (+ others) and the stock paste unless you get unlucky and the job from the factory was really poor.
 

THUMPer

2[H]4U
Joined
May 6, 2008
Messages
3,327
Yeah you really gotta weigh the pros and cons. If it's hitting thermal limit, then something may be wrong. But those temps don't really look that bad. I thought you said it was hitting 93 c hot spot?
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
93C hot spot yes. The temps were always fine, I just wanted them lower for a lower fan noise (while keeping clocks/voltage the same).

Curious, there's no aftermarket noctua fans I can replace the oem shroud with, is there? Has a review been done about noise/cooling benefits?
 

Furious_Styles

[H]ard|Gawd
Joined
Jan 16, 2013
Messages
1,830
93C hot spot yes. The temps were always fine, I just wanted them lower for a lower fan noise (while keeping clocks/voltage the same).

Curious, there's no aftermarket noctua fans I can replace the oem shroud with, is there? Has a review been done about noise/cooling benefits?
As I mentioned in a previous post I tried what you were doing with a radeon VII and specifically the washer mod which also supposedly got 10C lower temps. It did nothing for me (even tried different thickness washers) but what did work is the ol' 2 noctua (or any good fans) zip tied to the oem heatsink. I simply removed the plastic shroud+fans and used about 10 zip ties and got something like 15-20C lower temps and much quieter.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Since there is only 1 fan header, I would need a Y adapter for the 2 fans. And since there is an impediance (the heatsink) the correct Noctua fan would be the new NF-A12x25. So bill of materials:
Total of $62 for something that worked with the Radeon VII, but no reviews yet of it working with a 5700xt. Oh and zipties. Hmmm

edit: I've actually reached out to Noctua about their opinion on which fans I should use (92mm/120mm). And as a reminder, to cool the same item to the same C, the 120mm fans will have an easier time than the 92mm? So they will be quieter, right? The only issue against them is size (if they can fit in case) and marginal cost increase?

https://noctua.at/en/which_fan_is_right_for_me
 
Last edited:

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Curious, I started reducing gpu voltage, this time using MSI Afterburner instead of AMD Wattman. Wattman, surprise, was buggy AF and only allowed me 1900 clock, 1080mv before crashing in COD. Using MSI AB, I can reduce the mv all the way down to 1050mv and this passes both COD session and 30 min furmark. I will attempt to reduce the voltage further.

My question here is why hasn't there been a huge reduction in temps? For my CPU OC, reducing the voltage even a little results in quite a bit lower temps. For my 5700xt, going from 1080mv to 1050mv results in Core temps from 74 to 73, and Junction from 90 to 87. A 1C core, and 3C junction improvement. Shouldn't I be expecting more?
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Oh i'm def going to keep going. After all, the case has great cooling, clocks are set, so now you just decrease voltage. I was just expecting better temp improvement.
 

Rev. Night

Limp Gawd
Joined
Mar 30, 2004
Messages
175
Ok so I tested again tonight with 15 min+ COD sessions. I started at 1040 and dropped it down to 1000mv in MSI AB. I'm 1) shocked its stable at 1000mv and 2) dissappointed by the not huge reduction at temps. My temps at 1000mv are 73C core, 83C junction, an improvement of 2C core and 7C junction from 1080mv. 10% reduction of mv isn't a 10% reduction of temp?

Even though I have both temps and voltage being monitored by two programs (MSI AB and Open HW), I still feel like something is not correct. How can such a huge decrease in voltage not cause a huge decrease in temps? If I look at the chart log of MSI AB, and not just Max values, it decreases by a further 1-2C, but nothing major.
 

Attachments

Last edited:

THUMPer

2[H]4U
Joined
May 6, 2008
Messages
3,327
stop complaining about your temps then. It doesnt sound like you have enough air blowing on your card. if your fans are at 20% the gpu will be hotter than if it were at 40%. I think you may be at the limit of the cooler at that fan RPM. You are seeing a reduction in temps still.
 
Last edited:
Top