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

AMD RX 7000 series GPU Owners' Club

Im running 2700 core at 15% power / 1030mv. Fast memory timings at 2514.
You're extremely lucky if you get below 90 °C hotspot with +15% power.
 
You're extremely lucky if you get below 90 °C hotspot with +15% power.
After some more testing - it hit 100 in some intense jedi survivor moments. It averages around 95-97 though. Its still too high for me. I will drop the core clock (and voltage) down as id like to stay at 90.

Does a backplate have any effect on the hotspot temp? I've usually taken it off to get as much air around the back of the card but if it isn't making much of a difference, then it might be best to leave it on
:confused:
 
After some more testing - it hit 100 in some intense jedi survivor moments. It averages around 95-97 though. Its still too high for me. I will drop the core clock (and voltage) down as id like to stay at 90.

Does a backplate have any effect on the hotspot temp? I've usually taken it off to get as much air around the back of the card but if it isn't making much of a difference, then it might be best to leave it on
:confused:
Why not just run the card at stock? Is the difference that meaningful?

The backplate is mostly for structural rigidity, but it can also help with VRAM temperatures on some models. It does nothing with GPU/hotspot temperature.
 
Why not just run the card at stock? Is the difference that meaningful?

The backplate is mostly for structural rigidity, but it can also help with VRAM temperatures on some models. It does nothing with GPU/hotspot temperature.
There is a slight performance improvement, around 5. But if it's free, I'll take it. Stock is usually terrible when optimizing performance/wattage anyway.

Is VRAM temperature the same as the memory temperature in the AMD overlay?
 
There is a slight performance improvement, around 5. But if it's free, I'll take it. Stock is usually terrible when optimizing performance/wattage anyway.

Is VRAM temperature the same as the memory temperature in the AMD overlay?
nah vram temps and memory temps are two different things...
if you dont know what vram is check vrm on motherboards.
I would give you an answer but i want you to learn it yourself.
 
So, I have the Powercolor Red Devil GRE. I upped the mem to 2500 from 2250 and got an instant system crash. Are all GRE's gimped on memory or something?

Also, in GPU-Z, when I click the bios lookup, I get sent to the Powercolor Hellhound page! WTF?
 
So, I have the Powercolor Red Devil GRE. I upped the mem to 2500 from 2250 and got an instant system crash. Are all GRE's gimped on memory or something?

Also, in GPU-Z, when I click the bios lookup, I get sent to the Powercolor Hellhound page! WTF?
The weakest thing about the GRE is the memory but I thought they fixed that. Did you download the latest driver/software package from AMD?
 
The weakest thing about the GRE is the memory but I thought they fixed that. Did you download the latest driver/software package from AMD?
I saw about the memory on HUB, but I thought it was patched. I'm on 24.7.1. and have the latest everything being a new build.
 
Is VRAM temperature the same as the memory temperature in the AMD overlay?

Yes.. I believe its the same as Memory Junction Temp.
 
There is a slight performance improvement, around 5. But if it's free, I'll take it. Stock is usually terrible when optimizing performance/wattage anyway.
Performance is never free. It (almost) always comes with more heat.

Is VRAM temperature the same as the memory temperature in the AMD overlay?
Yes.

nah vram temps and memory temps are two different things...
if you dont know what vram is check vrm on motherboards.
I would give you an answer but i want you to learn it yourself.
I'm talking about VRAM, short for video random access memory, not VRM which is short for voltage regulation module.
 
Have being testing AFMF 2 and it's much improved, it doesn't seem to turn on/off intermittently anymore and quite honestly it's making FSR3 redundant.
 
Have being testing AFMF 2 and it's much improved, it doesn't seem to turn on/off intermittently anymore and quite honestly it's making FSR3 redundant.
I tried it in jedi survivor and ghost of Tsushima - i could see textures shaking or leaving a small visible trail when panning. It also wasn't very smooth.

EDIT: Heilos just got delivered. I'll apply it after work and hope for improvements.
 
i could see textures shaking or leaving a small visible trail
That's probably FSR, FG should not create temporal artifacts and ghosting. I've seen some UI artifacts but it's much better than what FSR3 mods usually look like when it comes to UI elements.
 
Right, so the initial results are in.

The Nitro is at 1040mv, 2625 core, 2564 FAST mem, 15% power and the fan capped at 65%.

I applied some NT-H1 2 days ago, after i scrapped off the hardened paste. That was hitting roughly 75 gpu and 104 hotspot.

With Heilos and about 45 mins of Jedi Survivor, it's sitting at 70 and the hotspot isnt going over 98.

Id say thats a nice improvement over what i had previously. Its not amazing but certainly better. Mind you, the card is in vertical mode, with the case on its side, ill drop 10+ from the hotspot easily.
 
Last edited:
Have being testing AFMF 2 and it's much improved, it doesn't seem to turn on/off intermittently anymore and quite honestly it's making FSR3 redundant.
The YT has an AM5 7950X + RX 7900XTX combo on CBP2077 with Path Tracing, prolly using the Preview AFMF driver, pretty amazing job by the driver team. I can't wait for it to make it to the regular driver release....

My apologies IF it has been posted before....
 
I've a question about RX7000's undervolt behavior.

Does the I/O voltage (VDDCI) on RDNA3 serves as the baseline for core voltage, just like all previous AMD GPUs?

For example, if the VDDCI is internally set at 800mV, would I be able to apply a voltage lower than 800mV on the core? Or would the core simply refuse to run below 800mV?

The reason I'm asking is because More Power Tool is incompatible with RDNA3 GPUs, and More Clock Tool appears to have very basic functionality with no option to tweak VDDCI, unlike RDNA2 and RDNA1 cards.

And I'm just curious if it's even possible to run RDNA3 based GPUs at super low voltages for purposes other than gaming?

I'd be absolutely grateful if you could spare a moment of your time and test if it's possible to run an RDNA3 card at voltages as low as 650-700mV.

Thanks in advance!
 
I'd be absolutely grateful if you could spare a moment of your time and test if it's possible to run an RDNA3 card at voltages as low as 650-700mV.
You could just simply limit the frequency until it runs at whatever voltage you want it to run at, like this :

Screenshot 2024-08-13 130143.png
 
  • Like
Reactions: Bad
I've a question about RX7000's undervolt behavior.

Does the I/O voltage (VDDCI) on RDNA3 serves as the baseline for core voltage, just like all previous AMD GPUs?

For example, if the VDDCI is internally set at 800mV, would I be able to apply a voltage lower than 800mV on the core? Or would the core simply refuse to run below 800mV?

The reason I'm asking is because More Power Tool is incompatible with RDNA3 GPUs, and More Clock Tool appears to have very basic functionality with no option to tweak VDDCI, unlike RDNA2 and RDNA1 cards.

And I'm just curious if it's even possible to run RDNA3 based GPUs at super low voltages for purposes other than gaming?

I'd be absolutely grateful if you could spare a moment of your time and test if it's possible to run an RDNA3 card at voltages as low as 650-700mV.

Thanks in advance!
Is this helps you in any way?

7900XTX
Its readings after several hours of light every day usage and some idling

1723543623471.png


1723543831046.png
 
  • Like
Reactions: Bad
Are there any bios settings to play with to increase stability/voltage? I'm on a gigabyte B550 motherboard.

I've come to realise, all the settings people recommended are useless to me. Using power +15 on my end makes my card super hot and clocks lower.

I get higher results with default power and an undervolt.
 
Idk where to ask, so I ask here.

Are the any 7900 models(Gre/Xt/Xtx) that come with PTM from the factory? Some guy here on TPU claimed Powercolor? uses them.
 
Powercolor Hellhound 7800XT uses PTM, or the equvalent TIM, so I'd guess any higher end Powercolor will use it.
 
Are there any bios settings to play with to increase stability/voltage? I'm on a gigabyte B550 motherboard.

I've come to realise, all the settings people recommended are useless to me. Using power +15 on my end makes my card super hot and clocks lower.

I get higher results with default power and an undervolt.
The +15% is fine to use though, as long as you limit your core clock. You'll more consistently hit that clock then, unless you set it too high. But I agree the space for improvement there is minimal. 0% is fine, heck even -15% is fine, but I found those both to be somewhat less stable.

I run my 7900XT with 2500 mhz max core clock, +15% power target and 1025mv, 2600 mhz on memory. It remains at or under 92C hotspot temp even in the worst case situations then, actual power usage peaks at 300W (while at stock and 0% power target but no max clock limitation, you will see 315W+ consistently). If I back the core clock down to 2400 mhz, which still doesn't cost much in perf, I can see 260-280W peak power most of the time, with a very rare bump towards 300W.
 
You could just simply limit the frequency until it runs at whatever voltage you want it to run at, like this :

That's exactly what I was looking for! I guess that means the base voltage is finally no longer affected by VDDCI with RDNA3.

However, I'm noticing the core frequency is only about 1,200 MHz (as per GPUz).

Hypothetically, if I want to raise the frequency to 1,500 MHz but want to keep the voltage at a constant ~700mV, do I simply raise the minimum frequency to 1,500 MHz? Or do I've to lower the voltage slider until I either achieve the desired V/F curve or the card gets unstable?

Thanks again in advance!
 
That's exactly what I was looking for! I guess that means the base voltage is finally no longer affected by VDDCI with RDNA3.

However, I'm noticing the core frequency is only about 1,200 MHz (as per GPUz).

Hypothetically, if I want to raise the frequency to 1,500 MHz but want to keep the voltage at a constant ~700mV, do I simply raise the minimum frequency to 1,500 MHz? Or do I've to lower the voltage slider until I either achieve the desired V/F curve or the card gets unstable?

Thanks again in advance!
Its a bit complicated...
In fact the voltage regulation of RX7000 series is a lot more complicated than RX6000

And in my case the undervolt doesnt work for "idle" state.
Its the same 0.68V on avg no matter the frequency and voltage I choose.
I believe those bars are for 3D loads at least or any load and not for straight idling.
I guess GPU chips have limits on lower possible voltage before they call themselves on sleep or shutdown state.
Like CPUs can park/sleep or shutoff different cores with 0.2V for example. GPUs have only 1 big core and its awake all times.
(my understanding on the subject)

1723560237360.png

1723560412417.png

1723560621731.png
 
  • Like
Reactions: Bad
Hypothetically, if I want to raise the frequency to 1,500 MHz but want to keep the voltage at a constant ~700mV, do I simply raise the minimum frequency to 1,500 MHz?
The voltage will never be constant, it's just going to hit a maximum according to the frequency.
 
  • Like
Reactions: Bad
Back
Top