AMD RX 7000 series GPU Owners' Club

[R-T-B]

Yeah well consider fact its being cooled with waterblock and overkill watercooling system i would expect much lower temps and my hotspot kept degrading, there even reports of the asrock variant aqua having user reports of 110c hotspot i would hardly consider that normal.

Yeah I understand now, didn't know you meant under water block.

I'll have some ptm pads ready, wanted to play with them for some time anyways.

 

[Ferrum Master]

Yeah I understand now, didn't know you meant under water block.

I'll have some ptm pads ready, wanted to play with them for some time anyways.

Do as Uncle Plop says and put PTM or Kryosheet as me. Are you using both win and Linux?

 

[Kabouter Plop]

Yeah I understand now, didn't know you meant under water block.

I'll have some ptm pads ready, wanted to play with them for some time anyways.

PTM7950 is for gpu core btw not for memory chips, anyway pumpout is a real issue, anyway seems like @Vya Domus found a solution in paste that came with the block that was hard to spread, either way if you where desperate enough to try liquid metal i would never consider that instead just go for PTM7950
I got mine from ebuy7 its a little expensive maybe but do consider fact chances you have to repaste again is close to 0%
NVIDIA has this stuff on their RTX 4090 founders edition as well.

Anyway kryosheet works to but do consider it conducts electricity, which is why i would personally go PTM7950

Im pretty sure if i kept on MX4 on a little longer my hotspot would eventually also reach 110c which is ridiculous.

 

[tpa-pr]

So I decided to tinker with an undervolt in between script work today. And even with me pumping 450W into this card (+15% PL) I still couldn't get the hotspot over 90C. The hottest part of the card was the memory junction which got to 90C for a second, then dropped again. This card is utterly amazing.

I'm not going to even keep the undervolt because it's unnecessary, the card plays everything I want beyond my expectations. But boy I felt like i've won the lottery here. It seems PowerColor did a good job pasting it and the cooler is way overkill.

 

[Vya Domus]

If you want to see what the actual maximum hotspot temp is run the AIDA64 GPGPU benchmark.

 

[Kabouter Plop]

3dmark does not record hotspot and hotspot temp is disabled by default these days in radeon software unless you check it in performance metrics here

Then show it in an overlay using ctrl+shift+o which is the default key for performance overlay.

 

[AusWolf]

I had a go at overclocking my 7800 XT with some strange results.

Raising or lowering the power limit works wonderfully, the card is super quiet and super efficient with -10%, and my clocks are much more stable around 2400 MHz with +15%.

However, if I do anything with the GPU clock, I immediately lose half of my VRAM clock - it goes up to 1200-ish MHz instead of 2425, even with the power limit raised.

I have no idea why this is. I haven't even touched any undervolt.

 

[tpa-pr]

If you want to see what the actual maximum hotspot temp is run the AIDA64 GPGPU benchmark.

I did run OCCT if that counts? It still doesn't get hot.

3dmark does not record hotspot and hotspot temp is disabled by default these days in radeon software unless you check it in performance metrics here

Then show it in an overlay using ctrl+shift+o which is the default key for performance overlay.

I was monitoring it via HWiNFO64 on my second screen

 

[Vya Domus]

However, if I do anything with the GPU clock, I immediately lose half of my VRAM clock - it goes up to 1200-ish MHz instead of 2425, even with the power limit raised.

Strange, try and set the memory clock manually ?

I did run OCCT if that counts? It still doesn't get hot.

I've found AIDA64 to be more demanding. They're both unrealistic though.

 

[AusWolf]

Strange, try and set the memory clock manually ?

It's working fine now. Now I really don't understand what's going on. Maybe the driver had a moment?

With that said, any setting gives me clocks of 2480-2530 in Unigine Superposition, and the exact same score as the +15% power limit with no overclock. I guess it's true - overclocking is a fool's errand these days.

 

[Kabouter Plop]

it's a rdna3 thing if you push core clocks to high it leaves not enough power for the memory so it will downclock the memory, which probably result of you lowering your power limit.
edit: it also can happen when you use third party tool instead to overclock like msi afterburner.

 

[R-T-B]

Do as Uncle Plop says and put PTM or Kryosheet as me. Are you using both win and Linux?

Linux is on hiatus until the HDR support improves, but probably will be using it eventually.

I will certainly be sourcing genuine honeywell PTM7950 pads (I know a legit, if slightly expensive way to get them) when I order them. No desire to go with conductive pastes on this or muck about with others.

 

[AusWolf]

edit: it also can happen when you use third party tool instead to overclock like msi afterburner.

I don't use third party tools for basic things that are included in the driver. I haven't had MSi Afterburner installed since the Middle Ages, probably.

it's a rdna3 thing if you push core clocks to high it leaves not enough power for the memory so it will downclock the memory, which probably result of you lowering your power limit.

That makes sense.

Why my GPU clock doesn't change is another matter. I guess, even the factory boost algorithm pushes the card to its limits, so there's no room left for raising GPU clocks.

 

[Kabouter Plop]

That is why you undervolt so you can boost higher on core clocks at a lower voltage budget, undervolting makes it so it uses less voltage to hit targets, but if its stable is another question this requires lot of testing and will vary per game, usually benchmarks it is easy to do ridicilous undervolts while for most games may not even tolerate a slight undervolt.

 

[tpa-pr]

Strange, try and set the memory clock manually ?


I've found AIDA64 to be more demanding. They're both unrealistic though.

Here we go, stock with my fan curve:

 

[Deleted member 237813]

people suddenly care about temps they could never reaqd out before stra nge how that works.....people also think they know better than the people who designed and built the cards. lmao

 

[AusWolf]

That is why you undervolt so you can boost higher on core clocks at a lower voltage budget, undervolting makes it so it uses less voltage to hit targets, but if its stable is another question this requires lot of testing and will vary per game, usually benchmarks it is easy to do ridicilous undervolts while for most games may not even tolerate a slight undervolt.

That sounds like way too much rocket science for my taste. I guess I'll leave it at +15% power, then. It does the job anyway.

people suddenly care about temps they could never reaqd out before stra nge how that works.....people also think they know better than the people who designed and built the cards. lmao

Why so sour? It's not like people know better. They just like tinkering. Everybody needs a hobby.

 

[tpa-pr]

people suddenly care about temps they could never reaqd out before stra nge how that works.....people also think they know better than the people who designed and built the cards. lmao

I do care in the sense that no matter what I do this card stays 20C+ cooler than my older 6950 XT! I'm impressed!

 

[Deleted member 237813]

i meant that poeple are worrying for no reason at all just because we now have sensor to read temps we couldnt many years agon and no one gave a shit. you get it? people should only worry with temps if it gets really hot for no reason at all or when it really has an issue which is rare. if the hot spot is 105 celsius who cares? these thing got tested for thousands of hours. every gpu gets tested before getting into the box. some get tested for 24 hours straight some only 1-2 hours but if it passes it gets boxed and sold.

or even better hur durhh my gpu reaches 80 celsius is that okay ? my cpu kis at 90 celsius hurr durhhh damn so many useless forum help posts .

cpu doesnt gives a shit if it runs 50 celsius or 100 celsius for a decade straight.

I do care in the sense that no matter what I do this card stays 20C+ cooler than my older 6950 XT! I'm impressed!

whats do you get out of it that its cooler? it doesnt matter like at all. these chips are designed to run hot if needed. it doesnt matter its just in your head a psychological effect but in the end as i said it doesnt matter at all.

 

[remekra]

It does matter as Edge temp is basically useless. Fan curve goes with the hotspot temp, not edge. If I would not care about hotspot I would still be stuck with broken 7900XTX that had edge always at max 60 while hotspot went to 105. It was not a normal behaviour and fans started to go into jet mode with 80% speed.
Also at least from the Gaming OC model, after card reached 100C on hotspot your manual fan curve was ignored.
So yeah hotspot is a lot more useful than edge.

 

[tpa-pr]

whats do you get out of it that its cooler? it doesnt matter like at all. these chips are designed to run hot if needed. it doesnt matter its just in your head a psychological effect but in the end as i said it doesnt matter at all.

I mean, I don't want to start an argument so I'll make this my last reply on the subject, but I find it interesting and neat. The new generation chip runs much cooler than my older generation chip, that's an improvement. It gets you thinking about what they did differently with this chip (or what the manufacturer did differently with the cooler, the paste etc). I'm pretty passionate about technology so I like watching numbers change and wondering or working out why they did. That's all.

It's the same reason why I was so excited the GPU boosted to 3.2Ghz. I haven't had a GPU go that high before, the number is bigger and that's interesting and neat!

 

[Kabouter Plop]

problem is that some vendors especially powercolor use cheap paste that does not even ressist pumpout, it won't last and thermals you get in first month will be worse in 2e month, its still better then what you can get by repasting with cheap paste like mx4 tho or mx6, so you are better of not repasting, unless you do your research and use something that can ressist pumpout which does exist even when ignoring kryosheet or ptm7950 by the looks of it, but most that had high hotspot did not solve it until they went kryosheet or ptm7950

This has nothing to do with remaining within spec yes its 110c and its within spec, but without pumpout you never even reach anywhere near those hotspot temps, my hotspot is still in the 60-70 range after more then 6 months at 22-23c watertemp at idle

also regardless at 110c hotspot card will just throtle harder, you will lose performance overtime due pumpout cos the chip will throtle it self to protect it self harder and harder.
Which is why i went PTM7950 cos i do not want to repaste it again honestly within its lifetime or atleast not within next 2 years.

 

[Vayra86]

i meant that poeple are worrying for no reason at all just because we now have sensor to read temps we couldnt many years agon and no one gave a shit. you get it? people should only worry with temps if it gets really hot for no reason at all or when it really has an issue which is rare. if the hot spot is 105 celsius who cares? these thing got tested for thousands of hours. every gpu gets tested before getting into the box. some get tested for 24 hours straight some only 1-2 hours but if it passes it gets boxed and sold.

or even better hur durhh my gpu reaches 80 celsius is that okay ? my cpu kis at 90 celsius hurr durhhh damn so many useless forum help posts .

cpu doesnt gives a shit if it runs 50 celsius or 100 celsius for a decade straight.


whats do you get out of it that its cooler? it doesnt matter like at all. these chips are designed to run hot if needed. it doesnt matter its just in your head a psychological effect but in the end as i said it doesnt matter at all.

It matters for several reasons, to an extent I might add. I don't completely disagree with you (people do overreact on seeing temps above 80C), but there's also just the aspect of optimizing your stuff and seeing how it affects your system.

But there are key reasons too:
- hotspot temp >110C is throttling. If you can remain below the throttle point consistently, you've removed a limitation for better clocks (ideally, you want to OC on a voltage limitation, not a temp limit).
- Noise. High hotspot temps will rev fans up.
- Summer ambient temps. If you have high hotspot temps in winter, you'll have a lot more throttling in summer.

 

[Kabouter Plop]

That sounds like way too much rocket science for my taste. I guess I'll leave it at +15% power, then. It does the job anyway.


Why so sour? It's not like people know better. They just like tinkering. Everybody needs a hobby.

I would try out some undervolt while benchmarking just run stock for playing games, which what i do usually.

 

[Deleted member 237813]

I mean, I don't want to start an argument so I'll make this my last reply on the subject, but I find it interesting and neat. The new generation chip runs much cooler than my older generation chip, that's an improvement. It gets you thinking about what they did differently with this chip (or what the manufacturer did differently with the cooler, the paste etc). I'm pretty passionate about technology so I like watching numbers change and wondering or working out why they did. That's all.

It's the same reason why I was so excited the GPU boosted to 3.2Ghz. I haven't had a GPU go that high before, the number is bigger and that's interesting and neat!

sure enjoy it. i personally couldnt care less if my gpu runs with 1 mhz if the fps is good thats all that matters, how its achieved i dont care much lol.

- hotspot temp >110C is throttling

thats why there is no reason to fear temps modern hardware will regulate itsself if it hits a limit. and also while its throttling you wont lose anything that is more than measurable. if a gpu runs with 2900mh or 2700mhz it doesnt matter you wont feel the difference in gameplay.

and if you throttle then your system is i dont even know how anyone can have a system with bad air flow or anything. you only need 2 fans thats all 1 front 1 back. people with 10 fans are running worse than they would with 3-4 which is the sweetspot even then it only matter if yaou have an 4090 running wild and a modern high end intel cpu. otherwise it doesnt matter.

shit with a 4070 ypou could get away with 1 fan on the back. especially with mesh front. but as weall know more fans and rgb gets you more fps. thats why i attached freaking xenon lights to my pc