Overclock.net banner
1 - 20 of 22 Posts

PotatoVonEpicus

· Registered
Joined
·
756 Posts
Discussion starter · #1 · (Edited)
Some time ago I bought XFX 6750XT and generally was quite satisfied with it. It's a good bin, it can run 2825 MHz stable on core with stock voltage, can max out memory up to it's superficial limit and run fast timings. It doesn't really coil whine that much and cooler is not too loud, even though it seemed fairly loud when I first let it run.

But it suffers from quite annoying issue and it is... You guessed it, overheating. I run the card on 2725 MHz with -50mv and 280w power limit ( which it never hits ) and it reaches up to 85c core and 96c memory temps in around 30 minutes.

I should have returned the card and bought something else, since I bought it new. But I was after the good bin and decided to keep it, firstly I replaced thermal paste and that didn't help at all. It was odd indeed, I took it apart again and squeezed memory pads down a little bit so core die makes better contact with heatsink and that dropped temps quite a bit ( which is 85c mentioned above, it would be like 10c higher, this increased memory temps by 2c ). So clearly contact between core and heatsink was not perfect due to thermal pads being too high.

While 85c core is still too hot for my liking, memory reaching 96c is definitely too hot for my liking. So I am guessing I should get some better thermal pads and see if that fixes memory temps. But even if it does I would like to drop core temp down a bit. I don't think it should be running 85c while drawing 260w.

I tested temps with realistic gaming load by just standing still in Hunt Showdown looking at same spot for 30 mins.
Wheel Vehicle Plant Wood Motor vehicle
Tested with 3200x1800 max settings.

Top case set as pull - push ( intake ) 6 fans through 360mm Liquid Freezer II ( CPU drawing around 50w, shown in image above ). Case front set as push with 3 fans, so airflow is quite decent in the case.

And this is exact graphics card model that I have - XFX SPEEDSTER MERC 319 AMD Radeon™ RX 6750 XT BLACK Gaming Graphics Card with 12GB GDDR6 HDMI 3xDP, AMD RDNA™ 2

Suggestions are welcome.
 
Some time ago I bought XFX 6750XT and generally was quite satisfied with it. It's a good bin, it can run 2825 MHz stable on core with stock voltage, can max out memory up to it's superficial limit and run fast timings. It doesn't really coil whine that much and cooler is not too loud, even though it seemed fairly loud when I first let it run.

But it suffers from quite annoying issue and it is... You guessed it, overheating. I run the card on 2725 MHz with -50mv and 280w power limit ( which it never hits ) and it reaches up to 85c core and 96c memory temps in around 30 minutes.

I should have returned the card and bought something else, since I bought it new. But I was after the good bin and decided to keep it, firstly I replaced thermal paste and that didn't help at all. It was odd indeed, I took it apart again and squeezed memory pads down a little bit so core die makes better contact with heatsink and that dropped temps quite a bit ( which is 85c mentioned above, it would be like 10c higher, this increased memory temps by 2c ). So clearly contact between core and heatsink was not perfect due to thermal pads being too high.

While 85c core is still too hot for my liking, memory reaching 96c is definitely too hot for my liking. So I am guessing I should get some better thermal pads and see if that fixes memory temps. But even if it does I would like to drop core temp down a bit. I don't think it should be running 85c while drawing 260w.

I tested temps with realistic gaming load by just standing still in Hunt Showdown looking at same spot for 30 mins.
Tested with 3200x1800 max settings.

Top case set as pull - push ( intake ) 6 fans through 360mm Liquid Freezer II ( CPU drawing around 50w, shown in image above ). Case front set as push with 3 fans, so airflow is quite decent in the case.

And this is exact graphics card model that I have - XFX SPEEDSTER QICK 319 AMD Radeon™ RX 6700 XT BLACK Gaming Graphics Card with 12GB GDDR6 HDMI 3xDP, AMD RDNA™ 2

Suggestions are welcome.
Hot-spot 96-106C for AMD cards is normal..
Anyway 2700Mhz+ card is considered hard OC for 6950XT so your temps are pretty good !
 
  • Rep+
Reactions: Mergatroid
Discussion starter · #4 ·
Hot-spot 96-106C for AMD cards is normal..
Anyway 2700Mhz+ card is considered hard OC for 6950XT so your temps are pretty good !
6950XT does have more cores, thus higher power. Even if die itself is bigger it's much more likely that 6950XT runs hotter compared to lower power draw GPUs. Hotspot up to 105c is normal, yes. Core near 90c / mem 96c. Less so.
 
What are your ambient temps? Have you tried playing around with the fan curve?
 
6950XT does have more cores, thus higher power. Even if die itself is bigger it's much more likely that 6950XT runs hotter compared to lower power draw GPUs. Hotspot up to 105c is normal, yes. Core near 90c / mem 96c. Less so.
True. Forgot to address memory temp - 96C is too much. Replace thermal pads..
 
I don't think you can do much more on the core unless you deshroud and put bigger fans on the heatsink (or just plain out undervolt and underclock). The vram temps do look on the higher side, so you could be right in assuming the thermalpads are not that good.
What kind of cooling you have inside your PC case?
 
Discussion starter · #9 ·
I don't think you can do much more on the core unless you deshroud and put bigger fans on the heatsink (or just plain out undervolt and underclock). The vram temps do look on the higher side, so you could be right in assuming the thermalpads are not that good.
What kind of cooling you have inside your PC case?
It doesn't undervolt below 1150mv with any reasonable core clock. I don't quite remember PC fan specs, but I have decent fans and even with glass side panel removed temp drops just by 2 - 3c. So I think airflow is good.

I could deshroud and get some better fans for the heatsink, but I am not sure how helpful that would be. It seems to me like that heatsink itself is not that good.
 
It doesn't undervolt below 1150mv with any reasonable core clock. I don't quite remember PC fan specs, but I have decent fans and even with glass side panel removed temp drops just by 2 - 3c. So I think airflow is good.

I could deshroud and get some better fans for the heatsink, but I am not sure how helpful that would be. It seems to me like that heatsink itself is not that good.
Merc not good heatsink ? I wouldnt agree. Just replace thermal pads and tighten well the GPU and everything..
I had 109C temp on my https://www.xfxforce.com/shop/xfx-s...tm-rx-6700xt-core-gaming-graphics-card-with-12gb-gddr6-hdmi-3xdp-amd-rdna-tm-2- while stress testing and even 100% fan didnt help . After a year using it - I decided f*** warranty - Im just going to open it.
Changed the paste (I used Kryonaut) Re-tightened everything and - voila - core temps were fine 96C MAX. I lowered the fan curve and its been fine since...
 
I did a quick google search for 6750XT reviews and found that your "Core temp", assuming that is "Hotspot", seems within average. Example from TechPowerUp. If your Core Temp = GPU edge temp, then we have a real problem. I would return or RMA if possible. 85c (edge) is too high, even if clocks aren't being reduced but I don't think that is the case if you're hitting ~2800Mhz.

Rectangle Font Parallel Screenshot Number


The STRIX is a good example to share. Massive overkill heatsink, very expensive, probably the only card that will manage decent temps at an acceptable noise level.

My advice, unless you're seeing performance degrade, don't put too much effort into it. Usually AMD refreshes run hotter and draw more power than their original, counterpart, for marginally higher clocks. Or in other words, they're just pre-OC cards because the silicon is "slightly" refined. Deshroud would probably net lower GPU hotspot temps, closer to STRIX level but seems like a waste on an already expensive card unless you have fans laying around and time to kill. Memory there really is only one solution -> better cooling at the modules. Pads or shims, not much else you can do.
 
Discussion starter · #12 ·
I did a quick google search for 6750XT reviews and found that your "Core temp", assuming that is "Hotspot", seems within average. Example from TechPowerUp. If your Core Temp = GPU edge temp, then we have a real problem. I would return or RMA if possible. 85c (edge) is too high, even if clocks aren't being reduced but I don't think that is the case if you're hitting ~2800Mhz.

View attachment 2607692

The STRIX is a good example to share. Massive overkill heatsink, very expensive, probably the only card that will manage decent temps at an acceptable noise level.

My advice, unless you're seeing performance degrade, don't put too much effort into it. Usually AMD refreshes run hotter and draw more power than their original, counterpart, for marginally higher clocks. Or in other words, they're just pre-OC cards because the silicon is "slightly" refined. Deshroud would probably net lower GPU hotspot temps, closer to STRIX level but seems like a waste on an already expensive card unless you have fans laying around and time to kill. Memory there really is only one solution -> better cooling at the modules. Pads or shims, not much else you can do.
In picture in initial post you can see GPU temp ( edge ) and hotspot ( 109c ). I did not expect it to reach such high temperature but before making a post I wanted to let it run for a bit. I can't return it at this point and for the most part it reaches this temperature in very specific scenarios ( high power draw games, which use game engines that chug on power ). I am not sure if I can RMA after breaking the sticker to take the card apart in EU. I know XFX is fine with RMA in US and Canada due to their laws, not sure about EU.

I decided to take it apart assuming that there was an issue with application of thermal paste or some sort of contact issue. Mainly because it's rather good bin and I wanted to keep it, but these temperatures are quite a bit too high for my liking. Also GPU idles at 40c, which I thought is a bit high as well. Looking at thermal paste after removing heatsink contact between heatsink and die looks fine after squeezing thermal pads down a little.

Appears to be core issue.
 
In picture in initial post you can see GPU temp ( edge ) and hotspot ( 109c ). I did not expect it to reach such high temperature but before making a post I wanted to let it run for a bit. I can't return it at this point and for the most part it reaches this temperature in very specific scenarios ( high power draw games, which use game engines that chug on power ). I am not sure if I can RMA after breaking the sticker to take the card apart in EU. I know XFX is fine with RMA in US and Canada due to their laws, not sure about EU.

I decided to take it apart assuming that there was an issue with application of thermal paste or some sort of contact issue. Mainly because it's rather good bin and I wanted to keep it, but these temperatures are quite a bit too high for my liking. Also GPU idles at 40c, which I thought is a bit high as well. Looking at thermal paste after removing heatsink contact between heatsink and die looks fine after squeezing thermal pads down a little.

Appears to be core issue.

Thanks for clarifying. You should reach out to confirm before assuming it’s not eligible for RMA.

My impression was EU > NA for consumer rights on warranties.
 
Discussion starter · #14 ·
Thanks for clarifying. You should reach out to confirm before assuming it’s not eligible for RMA.

My impression was EU > NA for consumer rights on warranties.
That is true and I should reach out to confirm this, but from google search results it seems that XFX warranty is void in EU if you break the sticker. Found multiple posts, although for older graphics cards.

Also when I took the card apart it didn't quite hit such high temps, probably had lower ambient temp and game was running for shorter amount of time. So generally it was just a bit too warm for me, but not straight up too hot to need RMA. I just thought the issue was thermal paste, otherwise I wouldn't have taken it apart.

And the issue still could be in the heatsink itself.

With that aside I have tested some other things as well. Such as more undervolting, the funny thing here is that I can undervolt more while in game on 100% load. Can go as low as 1080mv while running 2725 MHz. But idle instantly crashes, there are some weird things going on with voltage curve.
 
Discussion starter · #16 ·
Took card apart again, inspected heatsink. No issues, decided to move thermal pads further away from core and generously applied thermal paste. And, temps dropped down to 76c core and 82c memory in same 30 minutes temp test ( see image in the spoiler ). There was some issue and probably still is some issue with mounting, also crappy blue pads tend to rip very easily. Should definitely get better quality pads.
Plant Automotive tire Tree Wood Biome

I also have a few AIO pumps ( full CPU blocks ) just lying around, found one that actually would work with holes on the PCB without needing any modifications. Other than needing to find some spacers and to actually mounting it. It covers core and almost fully covers memory chips as well, really quite convenient find.

I could just slap that on and slap 240mm rad on it. Although VRMs would have no cooling.
 
I'm getting this model in a week or so. Hopefully I don't have these issues. Did you by chance measure the thickness of the thermal pads?
 
Discussion starter · #18 ·
I'm getting this model in a week or so. Hopefully I don't have these issues. Did you by chance measure the thickness of the thermal pads?
I haven't. Moving pads further away from the core area temporary fixed this issue. Later on I will buy some quality pads, will have to measure pad thickness before doing so. Generally I am very satisfied with how good binned this GPU is.
 
Got the card and been messing around with it some. Damn it's big. Longer than my old RX 480 with a Morpheus II. No idea how I'm going to take it out as the PCI release latch is totally unreachable now.

Anyway, was running some Unigine benchmarks and a few games. Have a custom fan curve and the highest edge temp I saw was 67C, junction 90C and VRAM 64C. This is with the fans at 50% and a 32C room temperature. Of course my display is only 60hz, so maybe the card is just not getting pushed enough (I run things with vsync and a 60 frame cap). This thing is running cooler than my old RX 480 with the Morpheus II and just as quiet. Going to try running things with uncapped fps tomorrow and see if it runs any different.

What I did notice was that the stock power limit on the card is 220W. I was under the impression these cards are 250W stock. Maybe it's just this particular model. Highest the core has boosted is 2680mhz with stock settings.
 
Discussion starter · #20 ·
Got the card and been messing around with it some. Damn it's big. Longer than my old RX 480 with a Morpheus II. No idea how I'm going to take it out as the PCI release latch is totally unreachable now.

Anyway, was running some Unigine benchmarks and a few games. Have a custom fan curve and the highest edge temp I saw was 67C, junction 90C and VRAM 64C. This is with the fans at 50% and a 32C room temperature. Of course my display is only 60hz, so maybe the card is just not getting pushed enough (I run things with vsync and a 60 frame cap). This thing is running cooler than my old RX 480 with the Morpheus II and just as quiet. Going to try running things with uncapped fps tomorrow and see if it runs any different.

What I did notice was that the stock power limit on the card is 220W. I was under the impression these cards are 250W stock. Maybe it's just this particular model. Highest the core has boosted is 2680mhz with stock settings.
220w stock + 15% for around 250w. You can use MPT to push past that. My card runs hot only while running high power draw applications. You should be able to do - 50mv on stock clocks.
 
1 - 20 of 22 Posts