AMD RX 7000 series GPU Owners' Club

[QuietBob]

@Vya Domus @Zach_01
These are the maximum recorded clocks for the in-game benchmark in native 4K with manually maxed out settings (disabled AA and RT where possible). Uncapped frame rate, no upscaling, no dynamic resolution/quality scaling, no frame generation, with game's default FOV. The tests were taken in the same exact conditions, so the only difference is run-to-run variance:

Game	HWinfo GPU front end clock (effective) MHz	HWinfo GPU shader clocks (effective) MHz	CapFrameX GPU core MHz
Assassin's Creed: Valhalla	2989	2771	2792
Borderlands 3 (dx12)	2689	2500	2494
Deus Ex: Mankind Divided (dx12)	2697	2508	2500
F1 23	2486	2308	2359
Far Cry 6	2887	2678	2671
Forza Horizon 4	3138	2909	2825
Forza Horizon 5	3060	2841	2858
Ghost Recon: Breakpoint (Vulkan)	2776	2581	2696
GTA V	2877	2669	2701
Horizon: Zero Dawn	2830	2632	2651
Immortals: Fenyx Rising	2976	2764	2759
Metro Exodus	2740	2545	2540
Metro Exodus: Enhanced Edition	3156	2927	2497
Rainbow Six: Extraction	3000	2780	2495
Red Dead Redemption 2 (dx12)	2799	2604	2599
Shadow of the Tomb Raider	2720	2528	2526
The Division 2	3064	2844	2466
The Riftbreaker	2850	2644	2659
Watch Dogs: Legion	2776	2577	2531

So yeah, it seems that CapFrameX reports effective shader clock. Thanks for the clarification! But there are some irregularities still

 

[physicsjunky]

When i first put conductonaut on my card the hot spot was doing amazing. Now after about 6 months I have a 30C delta when running full bore oc and uncapped fps. i've tried to fix it but no luck doing so. Only some games I play go past 165 fps. I mainly play War Thunder at 2k almost max settings the die temp can be in the mid 60C's and the hot spot could be high 80's to mid 90's.

 

[remekra]

When i first put conductonaut on my card the hot spot was doing amazing. Now after about 6 months I have a 30C delta when running full bore oc and uncapped fps. i've tried to fix it but no luck doing so. Only some games I play go past 165 fps. I mainly play War Thunder at 2k almost max settings the die temp can be in the mid 60C's and the hot spot could be high 80's to mid 90's.

Use PTM if you can, doesn't pump out and temps are good so I heard.
This looks like a pump out issue, don't know if it's possible with LM but PTM should be better anyway

 

[Vya Domus]

I used the paste that came with my block, it was very high viscosity I could barely even spread it. Temps are the same 6-7 months later, I suspect that the more "liquidy" the paste is the worse it will perform long term.

 

[Kabouter Plop]

6 months on my PTM no issue altho my delta was always high but its only high cos my edge temps are very low like around 39-42c usually hotspot is in the range of 60-70 usually avg at the lower end with peeks towards max around 22-23c watertemp on my liquid devil.

 

[wNotyarD]

Use PTM if you can, doesn't pump out and temps are good so I heard.
This looks like a pump out issue, don't know if it's possible with LM but PTM should be better anyway

I think it is. Weren't there PS5's with their LM pumping out?

 

[Vya Domus]

Weren't there PS5's with their LM pumping out?

It was a different issue, it had to do with the orientation of the console, some people keep their consoles upright so the LM would slowly move to one side of the chip.

 

[physicsjunky]

Use PTM if you can, doesn't pump out and temps are good so I heard.
This looks like a pump out issue, don't know if it's possible with LM but PTM should be better anyway

I did find some LM on the side of the die trying to fix it. Is there a proper way to clean LM off the coldplate?

 

[Zach_01]

When i first put conductonaut on my card the hot spot was doing amazing. Now after about 6 months I have a 30C delta when running full bore oc and uncapped fps. i've tried to fix it but no luck doing so. Only some games I play go past 165 fps. I mainly play War Thunder at 2k almost max settings the die temp can be in the mid 60C's and the hot spot could be high 80's to mid 90's.

Liquid Metal right?
Is there any copper involved in there?
Conductonaut tends to dry out and solidify long term, and needs reapplication before does that, a few times if copper is present in contacting surfaces. Every 2 months at first for 2-3 times.

 

[physicsjunky]

Liquid Metal right?
Is there any copper involved in there?
Conductonaut tends to dry out and solidify long term, and needs reapplication before does that, a few times if copper is present in contacting surfaces. Every 2 months at first for 2-3 times.

I replaced the LM and the cold plate is nickel plated copper. Could my nail polish job be the problem there is no nail polish on the die but a tiny amount on the metal part.

 

[Zach_01]

10hours session...
3 hours of simple usage
7 hours with monitor off

12-13W avg

 

[tpa-pr]

Oh, something interesting to report with the new card too: it seems i'm one of the lucky few with a monitor configuration that avoids the dreaded memory clock bug. This is with 2x 165Hz monitors (one 1080p and one 1440p) with TechPowerUp open and not much else:

Even my 6950 XT had the bug and the memory clock was always forced to max. So this is a definite improvement

 

[Lew Zealand]

I've seen the memory clock bug once or twice on my 7700 XT and the last time it happened (a few days ago) I just set the memory speed in Adrenalin to 2260 instead of 2250 and it clocked down properly. Of course this shouldn't happen at all but the fix is at least easy.

 

[Deleted member 237813]

Some undervolt can raise clocks within the same power level.
Voltage slider in drivers, at least for RDNA3, is a kind of an offset curve setting close to what CurveOptimizer is for Ryzen.
And what (UV) can work with 1 game doesn't mean will work on another with different type load, just keep in mind

thats why i do tuning on game per game basis depending on how many fps my goal is.

its a nice feature

 

[Kabouter Plop]

It was a different issue, it had to do with the orientation of the console, some people keep their consoles upright so the LM would slowly move to one side of the chip.

Many Asus gaming laptops have liquid metal to and often way to many of it that it fries the laptop, i guess liquid metal gets to liquid when it melts, while PTM7950 remains sticky while it melts.
Speaking of which techland support reacted to my e-mail finally after more then a week about the rgb laser show issue in dying light 2

You can fix this by playing on dx11 and using dxvk btw so its an directx issue with the game potentially a driver issue

 

[R-T-B]

So I now own a repasted XFX 7900 XTX Merc10. Just thought I'd mention that.

 

[Kabouter Plop]

So I now own a repasted XFX 7900 XTX Merc10. Just thought I'd mention that.

It likely will get worse again and pumpout i would see if you can get PTM7950 and do one final repaste with that so you never again have to do it again.
MX4 or MX6 is waste of time it can pump out as fast as 1 to 2 weeks

 

[remekra]

I wonder if Sapphire is using PTM on Nitro. Have the card for 8 months now, currently pumping 460W on it with OC and at fans running at max 1600RPM it reaches 90 max on hotspot and edge follows with good delta at 68-70.

 

[R-T-B]

It likely will get worse again and pumpout i would see if you can get PTM7950 and do one final repaste with that so you never again have to do it again.
MX4 or MX6 is waste of time it can pump out as fast as 1 to 2 weeks

I'll monitor it. I'm not concerned with "perfect" thermals mind, just need it to function within parameters. Stock XFX paste and pads are kinda notorious though, or so I am told.

 

[tpa-pr]

I'll monitor it. I'm not concerned with "perfect" thermals mind, just need it to function within parameters. Stock XFX paste and pads are kinda notorious though, or so I am told.

I'm actually surprised at the quality of PowerColor's job this time. Based on the thermals of my 6950 XT their paste jobs are not great. But once I cranked the fan curve on my 7900 XTX from default I struggle to get it hot, even when pumping 10% more power into it so far with testing.

 

[mama]

I'm actually surprised at the quality of PowerColor's job this time. Based on the thermals of my 6950 XT their paste jobs are not great. But once I cranked the fan curve on my 7900 XTX from default I struggle to get it hot, even when pumping 10% more power into it so far with testing.

What are the actual temps?

 

[Kabouter Plop]

I'll monitor it. I'm not concerned with "perfect" thermals mind, just need it to function within parameters. Stock XFX paste and pads are kinda notorious though, or so I am told.

When i got my liquid devil 7900 XTX it went from 70 to 80 to 92 hotspot it wont be a small degradation when it pumps out it will get pretty bad, if the paste you used is hard to spread tho it likely won't pumpout as easily, but if its something like mx4 or mx6 it likely gets worse over a period of 1-2 weeks

 

[R-T-B]

When i got my liquid devil 7900 XTX it went from 70 to 80 to 92 hotspot it wont be a small degradation when it pumps out it will get pretty bad, if the paste you used is hard to spread tho it likely won't pumpout as easily, but if its something like mx4 or mx6 it likely gets worse over a period of 1-2 weeks

I mean 80 to 92 hotspot is still within spec so I wouldn't be worried in my use case.

 

[Kabouter Plop]

I mean 80 to 92 hotspot is still within spec so I wouldn't be worried in my use case.

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.

just for info i ran my liquid devil 7900 XTX with a external mo-ra3 radiator and dual D5 with prety high flow rate
right now its at 60-70 range the hotspot and has't changed since asuming watertemp idles around 22-23c

 

[remekra]

If it will pump out it will get worse overtime. My Gigabyte Gaming OC model started at 88 hotspot and after some time it went to 100, then 105 then in some games it reached 110. Edge stayed the same at 60.