• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

6750XT can it be flashed ?

The reason for seeking more is that 6700XT and other 6750XT cards boost higher and I feel face punched by Gigabyte with this card cooler combination … maybe 50-100 Mhz isn’t a lot at at 2.5ghz frequencies so … I should call it a day. I can’t buy a different card now and with watercooling cant find parts to fit so this it
 
Hotspot on RDNA2 is always quite high.
even a 6500XT has a 20+°C delta after a repaste.
my 6900 XT Nitro + runs at 70°C edge and 100°C Hotspot after repasting it.
 
I just played several games with -50mV undervolt and frequencies hoover around 2625-2675 MHz ... still lame for this card which should be around 2750-2800Mhz ... The hotspot difference is much bigger on my side :( its like 30-40C I get 60-62-63C for GPU edge temp and 110C for Hotspot temp ... and I've tried everything except a good waterblock which I can't find. I'm thinking about buying an AIO and convert the CPU block to use it ... but CPUs these days dont consume 250W of power most of them ... so I'm a bit puzzled to what to do next :)

So far I've tried:

- undervolt (max I could get is -50 mV 99% stable)
- MPT tweaks
- Kryonaut thermal compound + pads (for mem)
- lapped the GPU heatsink almost to mirror finish
- switched included coolers with 2x12cm Artic cooling fans with high static pressure (these are running maxed at around 1800rpm)
- made a plastic shroud so air is pushed inside the cooler (since 12cm fans are a bit bigger)
- added washers to bolts to increase GPU mouting pressure (I think this stupid heatsink isnt making 100% contact - but thermal compound shows otherwise)
- I reseated the heatsink maybe 3-4 times using new compound each time and new thermal pads

Each of these maybe helped a bit but card still goes 62-63C for GPU edge and 110C for HOTSPOT. I could understand 110C hotspot but at 2750-2800 MHz not at 2650MHz.
 
so I'm a bit puzzled to what to do next
nothing?
i fully understand the appeal to max everything out (my 9900k ran at 5.2 Ghz all core squeezed to the edge of stability at almost 1.5V)
but we are talking about 3 FPS and a card that will probably not survive 2023.
 
so for example, 1200mV to 1150mV was unstable?! a 50mV undervolt should not be making a card unstable in my limited experience of messing with GPUs like that (since the RX 580 Polaris generation)

that is the biggest delta I've ever seen anyone report on any GPU. are you sure it behaved like this out of the box and these problems are not because you keep dicking around with it?

if it was like this OOTB I'd recommend returning to it's OTTB state and get a refund it as it is not working as it should
 
Why do you think it wont make it past 2023 ? I wont kill it :D

I still have an 1080TI lying around and a VEGA64 (Sapphire) and some older cards. The 1080TI is still going strong in 2022 (in older games) and it has an Artic Cooling big heatsink and never goes above 50C. The Sapphire 64 is the best card when it comes to cooling ... I just wish I could do the same with this one...
 
if it was like this OOTB I'd recommend returning to it's OTTB state and get a refund it as it is not working as it should
my 6700XT barely does -25mv in Warzone...
bad bin, card runs fine at stock. no way to refund since the product works as intended.
 
so for example, 1200mV to 1150mV was unstable?! a 50mV undervolt should not be making a card unstable in my limited experience of messing with GPUs like that (since the RX 580 Polaris generation)

that is the biggest delta I've ever seen anyone report on any GPU. are you sure it behaved like this out of the box and these problems are not because you keep dicking around with it?

if it was like this OOTB I'd recommend returning to it's OTTB state and get a refund it as it is not working as it should
It was performing like this from day one :) I lost the warranty when I lapped the heatsing and took the card apart. Gigabyte has horrible support I asked them about this problem and they just copy paste some generic answer. Don't get me wrong the card is working fine at 2600-2650MHz with no tweaks whatsoever. When I undervolt it by -50mV it still working fine clocks a bit higher (maybe extra 15-20 MHz) but the delta is still big when I give it more power or clocks. At completly stock it reaches 60C GPU and 100-105C for hotspot so still 40C maybe 35C sometimes.

I'm thinking about taking it apart again and maybe try to cut a small amount from the nuts that are on the heatsink that contact the PCB maybe they are not 100% even and GPU die isnt 100% in contact with heatsink so part of it can get hotter ?! But should heat dissipate anyway ? can it really create this "hotspot" ? I believed hot spot is one of the multi sensors inside GPU that reaches max temp.
 
wow that is some serious bad UV potential! I thought I had it bad with the delta on the RX 5700 XT, you guys got it way worse with your 6700 XTs. At least I could UV it to my hearts content (I did have to underclock the GPU too while UV) and you cannot even manage to adjust things that way either. Sorry to hear it's turning out to be such a crap shoot!

if you had really bad contact you would have a bad GPU core temp too but it's just the hotspot. It's the design of the card I think in that sort of scenerio. not necessarily the cooler, the actual PCB and die etc etc kinda parts.
 
Well atleast I found out that GPU Temp should also be bad from bad contact (I was thinking this way also in the past). I guess I will try to undervolt higher and lower clocks to around 2600MHz. I keep telling myself the difference from 100MHz at thousands of MHz is nothing. But the idiot inside wants to push it higher.
 
Well atleast I found out that GPU Temp should also be bad from bad contact (I was thinking this way also in the past). I guess I will try to undervolt higher and lower clocks to around 2600MHz. I keep telling myself the difference from 100MHz at thousands of MHz is nothing. But the idiot inside wants to push it higher.
your card has higher temps than the AMD Reference card. Maybe try the "Washer Mod" where you place a small plastic washer on the GPU mounting screw and maybe increase the pressure on the DIE area.
washer.jpg

But be careful with the mounting pressure!
 
your card has higher temps than the AMD Reference card. Maybe try the "Washer Mod" where you place a small plastic washer on the GPU mounting screw and maybe increase the pressure on the DIE area.
View attachment 262523
But be careful with the mounting pressure!
I've already tried that :) well my washers were 0.5 mm not that big like in the picture but I can try again with wider ones :)

Why is the edge temp GPU temp sitting at around 60C ? can the heatsink be that bad ? what about the screws that are on the GPU die side that are on the heatsink maybe if they aren't perfectly flat (the 4 of them) the heatsink doesnt make 100% contact ?

At least I managed a default clock undervolt at -100mV stable for 2 hours in RDR2 and clocks are 2650MHz stable and the temperature was 60-62C for GPU edge and around 95-100C for HOTSPOT.

I'm already thinking if a heatpipe from the 4 or 5 that it has might be broken ? I mean with no alcohol in it ? (I think its alcohol).

I would buy a heatsink / fan combo or water cooling kit or block + heatsinks for vrms and memory ... but can't find any that matches this card ...
 
Last edited:
5700XT here, more than 2.5years on my system and last few months, GPU hotspot was starting to pass 100C. Got up to 110C. Originally was around 90C when GPU temp was around 60~70C.
I tear down the cooler, put Kryonaught on GPU and changed all thermal pads and the ones on the back plate. Actually on the backplate I installed larger ones (more area covered). Pads are Thermalright Valor Odin.
At first GPU hot spot hasn't improved much, but after a couple of days started to lower and now its below 90C (also tight screws a little bit after 2 days).

I think GPU die/cooler pressure is key to GPU hot spot.

And BTW... RDNA1 GPUs are the first GPUs ever to introduce GPU Hotspot temp.
Prior to them the only GPU temp visible was the classic edge temp. The one we now call "GPU temp".
 
Sadly I've installed Kryonaut from Thermal Grizzly myself ... didn't remove the backplate tho ... you think the backplate could be an issue ???
 
Sadly I've installed Kryonaut from Thermal Grizzly myself ... didn't remove the backplate tho ... you think the backplate could be an issue ???
If backplate does not work well on thermals, VRAM temp has the bigger issue not the GPU so much. Putting more (bigger area) pads on backplate though could improve GPU/VRAM temp by a little if the plate is not plastic and can dissipate heat. Attached small heatsinks on it can improve temps further.

Did you check heatsink pressure on GPU?
 
Don't. If it's not broken, don't fix it.
 
@Lenne can you elaborate a bit further please ? Are you talking about the backplate or the card in general ?

If backplate does not work well on thermals, VRAM temp has the bigger issue not the GPU so much. Putting more (bigger area) pads on backplate though could improve GPU/VRAM temp by a little if the plate is not plastic and can dissipate heat. Attached small heatsinks on it can improve temps further.

Did you check heatsink pressure on GPU?
I've seen there is some kind of "paper" that can be put between the GPU die and heatsink to check contact. It's like a transfer paper from the 90s ? The one we put at school between two pieces of paper and you can write and "copy". I'll have to look into that because I dont have any. I could also check if the 4 bolts of the heatsink are perfectly even ! (the ones that align with the mounting holes where the screws from the backplate side go into).

P.S. what method you recommend to use to check the contact between GPU die and heatsink ?

Removed the backplate it has a 3mm thermal pad behind the gpu (too thick to actually do anything) the rest was clean. Some cheap aluminium I think ... anyway temps went better by 3-4 degress (despite of using MX4 thermal compound since Kryonaut was over). So backplate didnt help at all in my case. Card clocks at around 2650-2675 Mhz now and I call it quits. Had it with this card. The delta is still big at 40C but it works ...

I used separate screw for gpu mounting since the ones from the backplate had limited threads and without a backplate they wouldnt be enough to make good contact. I still used the springs and torqued them down nicely to snug fit.

Gigabyte screwed me with heatsink and thats the reason I got such a big delta T.

If anyone can help me choose some GPU universal block that will fit my card (and the rest of components) I would gladly do it !
 
@Lenne can you elaborate a bit further please ? Are you talking about the backplate or the card in general ?


I've seen there is some kind of "paper" that can be put between the GPU die and heatsink to check contact. It's like a transfer paper from the 90s ? The one we put at school between two pieces of paper and you can write and "copy". I'll have to look into that because I dont have any. I could also check if the 4 bolts of the heatsink are perfectly even ! (the ones that align with the mounting holes where the screws from the backplate side go into).

P.S. what method you recommend to use to check the contact between GPU die and heatsink ?

Removed the backplate it has a 3mm thermal pad behind the gpu (too thick to actually do anything) the rest was clean. Some cheap aluminium I think ... anyway temps went better by 3-4 degress (despite of using MX4 thermal compound since Kryonaut was over). So backplate didnt help at all in my case. Card clocks at around 2650-2675 Mhz now and I call it quits. Had it with this card. The delta is still big at 40C but it works ...

I used separate screw for gpu mounting since the ones from the backplate had limited threads and without a backplate they wouldnt be enough to make good contact. I still used the springs and torqued them down nicely to snug fit.

Gigabyte screwed me with heatsink and thats the reason I got such a big delta T.

If anyone can help me choose some GPU universal block that will fit my card (and the rest of components) I would gladly do it !
You need special tools to evaluate properly the Die/cooler contact. What I meant was to check screw tightness and maybe open the cooler again to see how paste is distributed across die
Sorry for misleading you on that and not being 100% clear from start.

My card has 4mm thermal pads on backplate and they work fine. Backplate is getting hot (around 50C) when playing games and that is whats for other than support the PCB from the cooler weight.
I even took it further, maybe too much, and install heatsinks on it, on more than half the backplate.

Did you ever touch it to see if its warming up?
 
You need special tools to evaluate properly the Die/cooler contact. What I meant was to check screw tightness and maybe open the cooler again to see how paste is distributed across die
Sorry for misleading you on that and not being 100% clear from start.

My card has 4mm thermal pads on backplate and they work fine. Backplate is getting hot (around 50C) when playing games and that is whats for other than support the PCB from the cooler weight.
I even took it further, maybe too much, and install heatsinks on it, on more than half the backplate.

Did you ever touch it to see if its warming up?
Yeah I did and it was warming up but the all the temps are higher with the backplate compared to naked card. So I didnt put it back :) I managed like a said a few more degress maybe 3-4C lower ! still not great. The delta remains 40-45C in some games ... in other games and bechmarks is like 30-35C delta. Card is boosting a tad higher now to 2650-2675 but thats about it. Sometimes it settles at around 2630-2640 mhz depending on the game itself :) older dx11 titles seem to be hotter on this card :)

Now I have a fan that blows over it from 15-20cm away :)

The card doesnt seems to have any issue but it ramps the clocks aggresively to reach 110C hotspot then it just rides that hotspot ... sometimes its 105C sometimes 110C sometimes goes to 95C but mostly around 100C.

My GPU TEMP (edge temp) is 57-59 rarely hits 60C now with more pressure on the heatsink but the delta is 35-40C even more sometimes if GPU is 60-61 I see 110C in some games. Card only throttles by 20-30MHz and sits around 2635-2640 MHz in those cases.

I tried to apply a very thin layer of thermal compound on the heatsink itself and when I applied it with a spatula I somehow seen/feel that one of the heatpipes is not perfectly alligned with the others ... but maybe I'm just halucinating :D cant be sure ... maybe the gap between them made me feel that way.

After weeks of trying I susped the heatsink design / quality assurance very bad ... I've tried to undervolt without more power and raising volts and I've got 1100mV stable in all benchmarks / applications. The gpu ASIC quality I've seen in GPU-Z I think its 87% or 86% that should be very bad ?! Are the ASIC values still valid for 6000 series ???

P.S. opening up the card again showed me that a row of memory wasn't making contact because of 1mm thermal pads (grizzly) that I've used ... atleast I could add another layer on top and now pretty sure it makes contact. Also like I've said I've used regular backplate screws with no thread stop for GPU / heatsink mounting and I could tighten it much better now. I didnt put excessive force and I also used the springs from original screw that made somehow an elastic mount (don't get me wrong theyh are compressed fully).

I'm thinking about a applying a sandwich or just one foil of pure silver to check if temps get better. I would also put thermal compound between the foils.

According to the internet:

Silver​

Silver is a precious metal that’s very ductile and malleable and is also an outstanding conductor of electricity and heat. Its thermal conductivity is 248 [BTU/(hr·ft⋅°F)] or 429 [W/m-K]. Despite its exceptional heat displacement, silver doesn’t find a lot of use in industrial applications, since it’s costly.

Copper​

Pure copper has the best conductivity of any other metal. For that reason, it is widely used for the manufacturing of heat exchangers, air conditioning and refrigerators, and hot water tanks. However, copper is also expensive, which limits its use in commercial applications.

Copper has a thermal conductivity of 232 [BTU/(hr·ft⋅°F)] or 401 [W/m-K] at room temperature.

Silver has better thermal conductivity compared to copper so thats why I think I could use it in my case.

Or maybe I should go copper shim way ?

 
Last edited:
I want to try another bios on the card. Is it possible to flash from a different vendor ? What bios should I choose from Techpowerup collection ? :) Thanks
 
I want to try another bios on the card. Is it possible to flash from a different vendor ? What bios should I choose from Techpowerup collection ? :) Thanks
Nope, there is no crossflashing vendors, post a gpu-z screenshot of your working bios please
 
Back
Top