• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD RX 7000 series GPU Owners' Club

Just go PTM7950 and be done with it.
 
Checking in again with my undervolt, these settings appear to be the final iteration. It seems either my card (or RDNA3 in general) is very touchy with its VRAM so that can't be pushed very hard at all. I can't push the clocks any higher without getting timeouts and I can't push the power draw any lower without getting timeouts. So these settings are it, and they work very nicely!

Oh, also: Stable Diffusion makes for an excellent test of undervolt stability. It pushes your card (both GPU and VRAM) really hard when you're generating images so you can monitor heat and find crashes quite quickly. And you can easily tell if something is wrong because you will get blank/corrupted images if you're pushing things too far (this is very apparent with VRAM tweaking).
1715725567835.png
 
  1. PTM7950
  2. CARBONAUT
  3. KRYOSHEET
Carbonaut is for sure going to be the worst, not even worth considering. There seems to be no real advantage from kryosheet vs high end paste thermally but it's also going to be electrically conductive, so there is a potential to screw things up. PTM7950 seems to be the best choice.
 
My GPU is already set up to use electrically conductive TIMs. I previously used liquid metal but after a good 6ish months it screwed up my thermals and I'm not sure how it happened.
Could it be my coating around the die preventing decent contact to the coldplate? I check to see if it was taller then die but i couldn't tell.
It is Nickel plated Copper BTW.
 
Something is not right, 110C with LM should not happen.
 
Checking in again with my undervolt, these settings appear to be the final iteration. It seems either my card (or RDNA3 in general) is very touchy with its VRAM so that can't be pushed very hard at all. I can't push the clocks any higher without getting timeouts and I can't push the power draw any lower without getting timeouts. So these settings are it, and they work very nicely!

Oh, also: Stable Diffusion makes for an excellent test of undervolt stability. It pushes your card (both GPU and VRAM) really hard when you're generating images so you can monitor heat and find crashes quite quickly. And you can easily tell if something is wrong because you will get blank/corrupted images if you're pushing things too far (this is very apparent with VRAM tweaking).
View attachment 347430
I'm not sure about your card, but as I've seen with my 7800 XT, RDNA 3 is heavily power limited. Any slight touch of the VRAM eats into the whole card's power allocation.
 
Something is not right, 110C with LM should not happen.
That was a heavy overclock that was at acceptable temps for the first 5 months. After that my idle temps shot up from 33-42C to 42-50C and temps went whack, I tried redoing the LM application a few times it didn't work. Since then I took 7000 Grit sand paper and got rid of the hardened LM and went back to thermal past (XTM-50).
The thermal paste brought idle temps back down too 33-42C but temps while gaming with an UV and no OC will still skyrocket with 165 FPS cap and Ultra setting=110C on the hotspot. I used to hit power limit (430W) wayy before any temp limits.

I no longer push an overclock UNLESS there is some headroom.
 
Last edited:
I am not promoting this channel but some nice assessment on the eco system from a new user.

 
Decided to try and get a better pic of my powercolor hellhound 7900gre
Everything has been stable and running great so far to and temps seem great.
1715867648973.jpeg
 
Decided to try and get a better pic of my powercolor hellhound 7900gre
Everything has been stable and running great so far to and temps seem great.
View attachment 347610
Nice! It seems we share excellent taste in CPU cooler, I have the TR PA120 SE as well! :toast: But it's in my 2nd rig, when the AIO in my main rig shows signs of dying, I'll swap it for whichever TR flagship cooler that is available at the time. It's great to see TR making a comeback after all these years, they used to be my go to brand back in the days. Oh yeah, your Hellhound RX 7900GRE looks great too....:D
 
Nice! It seems we share excellent taste in CPU cooler, I have the TR PA120 SE as well! :toast: But it's in my 2nd rig, when the AIO in my main rig shows signs of dying, I'll swap it for whichever TR flagship cooler that is available at the time. It's great to see TR making a comeback after all these years, they used to be my go to brand back in the days. Oh yeah, your Hellhound RX 7900GRE looks great too....:D
Thank you :D.
I got the CPU cooler for free and it's been great
 
That was a heavy overclock that was at acceptable temps for the first 5 months. After that my idle temps shot up from 33-42C to 42-50C and temps went whack, I tried redoing the LM application a few times it didn't work. Since then I took 7000 Grit sand paper and got rid of the hardened LM and went back to thermal past (XTM-50).
The thermal paste brought idle temps back down too 33-42C but temps while gaming with an UV and no OC will still skyrocket with 165 FPS cap and Ultra setting=110C on the hotspot. I used to hit power limit (430W) wayy before any temp limits.

I no longer push an overclock UNLESS there is some headroom.

So what are your hotspot temps at stock ? and at how much total board power ?
 
I played CP 2077 for 20 minutes at stock settings and here is the result.
image_2024-05-30_125823303.png
Screenshot 2024-05-30 123110.png

My case fans were set to full speed to help with temps.
War Thunder_2024.05.30-13.12.png
 
Last edited:
332 board power and hitting 110c hotspot is not normal i believe stock is 355 tbp yours is throttling by the looks of it.
 
Last edited:
Yep it sure is. Whenever the gpu goes over 300 Watts that's when the hotspot temp shoots up.
 
Was it this bad on day 1 when you just did repaste as well ?
 
No the gpu temps where amazing for the first 8 months even with LM but somthing messed up after that.
 
Wonder if PTM7950 can help but you need the right thickness for the pads and good contact with gpu core as well or it wont make a difference, if it does work i would never remove it.
 
yeah i tried fixing by repasting it so many times with both LM and my only paste XTM-50. It looks like pump out to me.
 
If you get instantly bad temps after re applying it might be contact issue, if you can fix that go PTM7950 it should no longer pump out.
 
Welp, just replaced my MSI RX 6600 Mech 2 with ...................... a WHITE rabbit! :p

Bought the Sapphire Pure RX 7900 GRE 16GB two days ago and it arrived today.

I'm in the process of getting all the parts for an all white build by October and I just couldn't resist installing these into my current B550 A-Pro / R5 5600 ... :D

My all white build will most likely consist of ...

NZXT H6 Flow
Gigabyte Aorus B650E Elite X AX Ice
AMD Ryzen 7 7800X3D or a Ryzen 9 7950X3D
Thermalright Peerless Assassin 120 SE RGB (or a white AIO - haven't decided)
32GB Crucial Pro 6000Mhz DDR5
2x Phanteks M25-140D RGB Fans
3x Phanteks M24-120D RGB Fans (or a white AIO top mounted)
Be Quiet Pure Power 12 M 1000 Watt (it's black but it'll be tucked behind the motherboard)
 
The 7800XT is absolutely the worst OC experience I ever had with a GPU. Touching the voltage slider even slightly is a sure way to have your game crash after 10+ hours of apparent stability. This experience is pretty much the same for several other people I've been in contact, so I highly doubt it's pure silicon lottery- there simply has to be something inherently broken with the entire 7800 XT lineup.

I could run certain tests with insane UV settings, but there are several games I played that completely bork themselves with even the slightest change to voltage or increase in frequency limit.

Especially annoying is seeing people share their "100% stable" settings at like, 3000MHz @1020mV or something crazy like that. You just havent played a game that will inevitably crash with those settings.

I crashed in Dark and Darker with 2300MHz limit and 1100mV. Just let that sink in... The voltage offset of -50mV is problematic for the lower part of V/F curve apparently.

Rant over.
 
I crashed in Dark and Darker with 2300MHz limit and 1100mV. Just let that sink in... The voltage offset of -50mV is problematic for the lower part of V/F curve apparently.

This seems to be the case on my 7700 XT as well. Now I don't crash all about in games and mine seems to undervolt fine as I use the 1083 mV curve, but if I'm going to play at a lower core clock speed, I bump the undervolt up to 1100 to be safe as it can crash there during a low GPU demand part of the game. However I can UV to the 1050 curve with the max OC on the card at around 2850 MHz in a continual high-demand game, but if the GPU load drops by too much and the card clocks dow, there's real chance of video crashout.

Note that the card isn't using 1050mV at 2850MHz, rather it's using 1150mV or thereabouts as the UV settings on RDNA3 and 2 are just a curve. But AMD likes to describe it as a mV setting which IMO is misleading but whatever, it's just semantics. Likewise, running the 7700 XT with the 1100mV curve at 2300 MHz powers the GPU cores at less than 1000mV.
 
The 7800XT is absolutely the worst OC experience I ever had with a GPU. Touching the voltage slider even slightly is a sure way to have your game crash after 10+ hours of apparent stability. This experience is pretty much the same for several other people I've been in contact, so I highly doubt it's pure silicon lottery- there simply has to be something inherently broken with the entire 7800 XT lineup.

I could run certain tests with insane UV settings, but there are several games I played that completely bork themselves with even the slightest change to voltage or increase in frequency limit.

Especially annoying is seeing people share their "100% stable" settings at like, 3000MHz @1020mV or something crazy like that. You just havent played a game that will inevitably crash with those settings.

I crashed in Dark and Darker with 2300MHz limit and 1100mV. Just let that sink in... The voltage offset of -50mV is problematic for the lower part of V/F curve apparently.

Rant over.
The only way to OC a 7800 XT is to max out the power slider. The rest of the stuff either does nothing, or affects your stability, so shouldn't be touched.

If you want a good OC card, get a 6500 XT. With that, you can max out all the sliders and still be rock stable and quiet. :p
 
Does anybody here have a 7600 (XT) here? If so, how does it do with video playback power consumption? Is it as bad as chiplet-based RDNA 3 cards?

I'm thinking about upgrading my spare 6500 XT to a 7600 because it has a better video decoder, better performance, and I could do it with minimal investment, but there's no point if it eats up to 50 W playing a video like the rest of the 7000-series.
 
Back
Top