The Radeon RX9070XT / RX9070 Owners Thread

Kirkatron · 2025-03-17T20:11:08+0000

Some Steel Nomad Testing Results to show the effect of changing the different settings.

For me the Core Offset does nothing when setting to a positive value but helps stabilise the above 3.4Ghz boost when set to negative.

I threw in a couple of 50% static fan profiles to show how cooling effects performance.

Mem above 2760 Fast timings had no extra performance and above 2800 had a negative impact.

First scored entry is trying to simulate a reference clocked none OC card with reference power limit.

Run on the system in my Sig. Steel Nomad required a higher core voltage to be stable than the games I currently play.

NOPE means it wasn't stable enough to run multiple runs.

Steel Nomad on 9070 XT Nitro+

Core Clock Offset	Core Voltage Offset	Mem Frequency	Mem Timings	Power Offset	Fan Profile	Score	FPS
0	-90mV	2500	Standard	+10%	Default	NOPE	NOPE
0	-100mV	2500	Standard	0	Default	NOPE	NOPE
0	-100mV	2500	Standard	+10%	Default	NOPE	NOPE
-200Mhz	-100mV	2730	Fast	+10%	Default	NOPE	NOPE
-200Mhz	-100mV	2730	Standard	+10%	Default	NOPE	NOPE
-90Mhz	0	2500	Standard	-9%	Default	6942	69.42
+450Mhz	0	2500	Standard	0	Default	7126	71.26
0	0	2500	Standard	0	Default	7142	71.42
0	0	2500	Fast	0	Default	7145	71.45
0	0	2630	Standard	0	Default	7152	71.52
0	0	2630	Fast	0	Default	7203	72.03
0	0	2760	Standard	0	Default	7225	72.25
0	-30mV	2500	Standard	0	Default	7231	72.31
0	0	2760	Fast	0	Default	7271	72.71
0	0	2500	Standard	+10%	Default	7279	72.79
0	-30mV	2500	Standard	+10%	Default	7351	73.51
0	-60mV	2500	Standard	0	Default	7353	73.53
0	-80mV	2500	Standard	0	Default	7387	73.87
0	0	2760	Fast	+10%	Default	7401	74.01
0	-90mV	2500	Standard	0	Default	7417	74.17
-200Mhz	-100mV	2500	Standard	0	Default	7454	74.54
0	-60mV	2500	Standard	+10%	Default	7463	74.63
0	-30mV	2760	Standard	+10%	Default	7486	74.86
0	-30mV	2760	Fast	+10%	Default	7508	75.08
0	-80mV	2500	Standard	+10%	Default	7541	75.41
-50Mhz	-90mV	2500	Standard	+10%	Default	7568	75.68
0	-60mV	2760	Standard	+10%	Default	7576	75.76
-200Mhz	-100mV	2500	Standard	+10%	Default	7580	75.80
0	-60mV	2760	Fast	+10%	Default	7650	76.50
0	-80mV	2760	Standard	+10%	Default	7668	76.68
0	-80mV	2760	Fast	+10%	Default	7687	76.87
0	-80mV	2800	Fast	+10%	Default	7693	76.93
-200Mhz	-100mV	2630	Fast	+10%	Default	7693	76.93
-50Mhz	-90mV	2760	Standard	+10%	Default	7719	77.19
0	-60mV	2760	Fast	+10%	50%	7729	77.29
-50Mhz	-90mV	2760	Fast	+10%	Default	7735	77.35
0	-80mV	2760	Fast	+10%	50%	7784	77.84

Funky-Melon · 2025-03-17T20:13:17+0000

Wow, I actually got one ordered! I'm so happy yet inside I hate myself because I know I've overpaid and gone against everything I stood for

Oh well, stuff morals, shiny new toy will arrive Wednesday and my little 3060 TI can retire

hornetstinger · 2025-03-17T20:14:37+0000

Funky-Melon said:
Wow, I actually got one ordered! I'm so happy yet inside I hate myself because I know I've overpaid and gone against everything I stood for

Oh well, stuff morals, shiny new toy will arrive Wednesday and my little 3060 TI can retire

210% difference according to techpowerup.

LeMson · 2025-03-17T20:14:39+0000

Funky-Melon said:
Wow, I actually got one ordered! I'm so happy yet inside I hate myself because I know I've overpaid and gone against everything I stood for

Oh well, stuff morals, shiny new toy will arrive Wednesday and my little 3060 TI can retire

Same ;( Roll on Wednesday

eeii · 2025-03-17T20:21:16+0000

Kirkatron said:
Some Stell Nomad Testing Results to show the effect of changing the different settings. For me the Core Offset does nothing when setting to a positive value but helps stabilise when set to negative.
I threw in a couple of 50% static fan profiles to show how cooling effects performance.
Mem above 2760 Fast timings had no extra performance and above 2800 had a negative impact.
First scored entry is trying to simulate a reference clocked none OC card with reference power limit.
Run on the system in my Sig. Steel Nomad required a higher core voltage to be stable than the games I currently play.
NOPE means it wasn't stable enough to run multiple runs.

Steel Nomad on 9070 XT Nitro+

Core Clock Offset Core Voltage Offset Mem Frequency Mem Timings Power Offset Fan Profile Score FPS
0 -90mV 2500 Standard +10% Default NOPE NOPE
0 -100mV 2500 Standard 0 Default NOPE NOPE
0 -100mV 2500 Standard +10% Default NOPE NOPE
-200Mhz -100mV 2730 Fast +10% Default NOPE NOPE
-200Mhz -100mV 2730 Standard +10% Default NOPE NOPE
-90Mhz 0 2500 Standard -9% Default 6942 69.42
+450Mhz 0 2500 Standard 0 Default 7126 71.26
0 0 2500 Standard 0 Default 7142 71.42
0 0 2500 Fast 0 Default 7145 71.45
0 0 2630 Standard 0 Default 7152 71.52
0 0 2630 Fast 0 Default 7203 72.03
0 0 2760 Standard 0 Default 7225 72.25
0 -30mV 2500 Standard 0 Default 7231 72.31
0 0 2760 Fast 0 Default 7271 72.71
0 0 2500 Standard +10% Default 7279 72.79
0 -30mV 2500 Standard +10% Default 7351 73.51
0 -60mV 2500 Standard 0 Default 7353 73.53
0 -80mV 2500 Standard 0 Default 7387 73.87
0 0 2760 Fast +10% Default 7401 74.01
0 -90mV 2500 Standard 0 Default 7417 74.17
-200Mhz -100mV 2500 Standard 0 Default 7454 74.54
0 -60mV 2500 Standard +10% Default 7463 74.63
0 -30mV 2760 Standard +10% Default 7486 74.86
0 -30mV 2760 Fast +10% Default 7508 75.08
0 -80mV 2500 Standard +10% Default 7541 75.41
-50Mhz -90mV 2500 Standard +10% Default 7568 75.68
0 -60mV 2760 Standard +10% Default 7576 75.76
-200Mhz -100mV 2500 Standard +10% Default 7580 75.80
0 -60mV 2760 Fast +10% Default 7650 76.50
0 -80mV 2760 Standard +10% Default 7668 76.68
0 -80mV 2760 Fast +10% Default 7687 76.87
0 -80mV 2800 Fast +10% Default 7693 76.93
-200Mhz -100mV 2630 Fast +10% Default 7693 76.93
-50Mhz -90mV 2760 Standard +10% Default 7719 77.19
0 -60mV 2760 Fast +10% 50% 7729 77.29
-50Mhz -90mV 2760 Fast +10% Default 7735 77.35
0 -80mV 2760 Fast +10% 50% 7784 77.84

Well done this is excellent work. When mine finally turns up I’ll give this a go.

Funky-Melon · 2025-03-17T20:29:00+0000

hornetstinger said:
210% difference according to techpowerup.

Wow, I didn't realise it was that high.

I hope this works out, last time I had an AMD card it crashed continuously and the one before that released the magic smoke on first boot. From then on I stayed with Nvidia and it's been a pretty faultless experience.

Feel's good giving Nvidia the finger though, based on Jenson's ego, lies and poor 5000 release.

eeii · 2025-03-17T20:32:38+0000

Funky-Melon said:
Wow, I didn't realise it was that high.

I hope this works out, last time I had an AMD card it crashed continuously and the one before that released the magic smoke on first boot. From then on I stayed with Nvidia and it's been a pretty faultless experience.

Feel's good giving Nvidia the finger though, based on Jenson's ego, lies and poor 5000 release.

You will need to do a minimum of using DDU to get rid of the drivers. Some forum members have had todo a clean install.

Funky-Melon · 2025-03-17T20:35:25+0000

eeii said:
You will need to do a minimum of using DDU to get rid of the drivers. Some forum members have had todo a clean install.

I was tempted to just reinstall Windows from scratch to mitigate any issues.

fs123 · 2025-03-17T20:46:29+0000

Case1987 said:
I used FSR3 in the game settings and I'm getting FSR4,the problem is it says I have a 4090 in the benchmarks

I can't get the FSR3 inputs to work. The optiscaler menu comes up but changing upscalers has no effect. Only XeSS as source API is working for me. I'm on Win11 so maybe that is causing the issue. FSR4 still looks great though but would prefer to use the fsr3 inputs.

Mercurial · 2025-03-17T21:12:06+0000

Regarding Optiscaler; how likely is it that NV will accidentally "break" support for it from DLSS if it gains considerable traction (though I guess you can use Xess/FSRxx in it's stead)?

They do have form in this regard from the past.

NB. I've not yet played with Optiscaler so this is just based on what i've cursively read.

fs123 · 2025-03-17T21:32:02+0000

Mercurial said:
Regarding Optiscaler; how likely is it that NV will accidentally "break" support for it from DLSS if it gains considerable traction (though I guess you can use Xess/FSRxx in it's stead)?

They do have form in this regard from the past.

NB. I've not yet played with Optiscaler so this is just based on what i've cursively read.

Optiscaler and other mods such as CyberFSR and DLSS2FSR3 have been around for over 3 years now. I doubt Nvidia cares or if they can do anything about it. They have to have their dll files in the game directory and that is where Optiscaler hooks into the game. They wouldn't integrate DLSS code into a game because it's proprietary and even game devs can't look at it's code.

Mercurial · 2025-03-17T21:35:24+0000

Makes sense, thank you for the explanation. Sounds like there is nothing to lose by trying it and a considerable gain if it works.

ben90 · 2025-03-17T21:45:36+0000

Decided to buy one on Thursday on the 2pm drop, card arrived Friday morning at 7.15am! I set a limit of £650 in the rumour thread a month or two ago so I was at my limit! I think it's too much really but I've been waiting for 3 months to upgrade so that I can build my girlfriend a PC of her own. Absolutely 0 chance I would buy an Nvidia card at this point so I couldn't be bothered to wait any longer, especially as I don't expect to see msrp any time soon.

Case1987 · 2025-03-17T21:48:48+0000

fs123 said:
I can't get the FSR3 inputs to work. The optiscaler menu comes up but changing upscalers has no effect. Only XeSS as source API is working for me. I'm on Win11 so maybe that is causing the issue. FSR4 still looks great though but would prefer to use the fsr3 inputs.

In optiscaler I just used FSR3.x,then changed FSR settings to FSR4.0.0.I'm in Windows 11 also,but it doesn't really matter if it's working,it's so much better than FSR3

Case1987 · 2025-03-17T21:50:41+0000

Mercurial said:
Makes sense, thank you for the explanation. Sounds like there is nothing to lose by trying it and a considerable gain if it works.

It works and it's amazing imo,the improvement is insane

Bidley · 2025-03-17T22:15:38+0000

What people should remember regarding Cyberpunk and the difference between FSR3 and FSR4 via Optiscaler... CDPR's implementation is awful, and is actually worse than the FSR2 in the game. And FSR2 is bad in that game to begin with. They added FSR3 months after 3.1 released. Which makes sense as they didn't want 3.1's decoupled frame gen to use with older RTX cards and DLSS upscaling.

It makes sense as it's an Nvidia-sponsored game, and FSR4 is a massive leap ahead of FSR3.1 in general, but Cyberpunk isn't a great scenario for FSR native.

Thank god for modders

Tythus · 2025-03-17T22:40:53+0000

Bidley said:
What people should remember regarding Cyberpunk and the difference between FSR3 and FSR4 via Optiscaler... CDPR's implementation is awful, and is actually worse than the FSR2 in the game. And FSR2 is bad in that game to begin with. They added FSR3 months after 3.1 released. Which makes sense as they didn't want 3.1's decoupled frame gen to use with older RTX cards and DLSS upscaling.

It makes sense as it's an Nvidia-sponsored game, and FSR4 is a massive leap ahead of FSR3.1 in general, but Cyberpunk isn't a great scenario for FSR native.

Thank god for modders

I hope that the vendor neutral endpoint is eventually adopted from MS

tommybhoy · 2025-03-17T22:42:16+0000

fs123 said:
Cyberpunk uses XeSS inputs to get FSR4 working.
I just booted up CP2077 and it remembered that FSR4 was enabled. It works like this : Enable XeSS in the game settings and open Optiscaler menu, set upscaler to fsr3.x (FSR4.0.0). Now when I change upscaler in the game settings to fsr2 or fsr3 it uses those upscalers but when I switch to Xess it uses FSR4.

Generally you shouldn't need to change the upscaler in the game settings once you have Optiscaler working. You can change the upscaler from within the Optiscaler menu to test out the different upscalers for comparison reasons.

↓

Case1987 said:
Once you set it that's it

Lol, thanks @fs123 for trying, thanks @Case1987 for the info I was looking for.

Tythus · 2025-03-17T22:43:23+0000

tommybhoy said:
↓

Lol, thanks @fs123 for trying, thanks @Case1987 for the info I was looking for.

worst case is on update you might have to jiggle it back

Phreaky · 2025-03-17T22:45:49+0000

Kirkatron said:
Some Steel Nomad Testing Results to show the effect of changing the different settings.

For me the Core Offset does nothing when setting to a positive value but helps stabilise the above 3.4Ghz boost when set to negative.

I threw in a couple of 50% static fan profiles to show how cooling effects performance.

Mem above 2760 Fast timings had no extra performance and above 2800 had a negative impact.

First scored entry is trying to simulate a reference clocked none OC card with reference power limit.

Run on the system in my Sig. Steel Nomad required a higher core voltage to be stable than the games I currently play.

NOPE means it wasn't stable enough to run multiple runs.

Steel Nomad on 9070 XT Nitro+

Core Clock Offset Core Voltage Offset Mem Frequency Mem Timings Power Offset Fan Profile Score FPS
0 -90mV 2500 Standard +10% Default NOPE NOPE
0 -100mV 2500 Standard 0 Default NOPE NOPE
0 -100mV 2500 Standard +10% Default NOPE NOPE
-200Mhz -100mV 2730 Fast +10% Default NOPE NOPE
-200Mhz -100mV 2730 Standard +10% Default NOPE NOPE
-90Mhz 0 2500 Standard -9% Default 6942 69.42
+450Mhz 0 2500 Standard 0 Default 7126 71.26
0 0 2500 Standard 0 Default 7142 71.42
0 0 2500 Fast 0 Default 7145 71.45
0 0 2630 Standard 0 Default 7152 71.52
0 0 2630 Fast 0 Default 7203 72.03
0 0 2760 Standard 0 Default 7225 72.25
0 -30mV 2500 Standard 0 Default 7231 72.31
0 0 2760 Fast 0 Default 7271 72.71
0 0 2500 Standard +10% Default 7279 72.79
0 -30mV 2500 Standard +10% Default 7351 73.51
0 -60mV 2500 Standard 0 Default 7353 73.53
0 -80mV 2500 Standard 0 Default 7387 73.87
0 0 2760 Fast +10% Default 7401 74.01
0 -90mV 2500 Standard 0 Default 7417 74.17
-200Mhz -100mV 2500 Standard 0 Default 7454 74.54
0 -60mV 2500 Standard +10% Default 7463 74.63
0 -30mV 2760 Standard +10% Default 7486 74.86
0 -30mV 2760 Fast +10% Default 7508 75.08
0 -80mV 2500 Standard +10% Default 7541 75.41
-50Mhz -90mV 2500 Standard +10% Default 7568 75.68
0 -60mV 2760 Standard +10% Default 7576 75.76
-200Mhz -100mV 2500 Standard +10% Default 7580 75.80
0 -60mV 2760 Fast +10% Default 7650 76.50
0 -80mV 2760 Standard +10% Default 7668 76.68
0 -80mV 2760 Fast +10% Default 7687 76.87
0 -80mV 2800 Fast +10% Default 7693 76.93
-200Mhz -100mV 2630 Fast +10% Default 7693 76.93
-50Mhz -90mV 2760 Standard +10% Default 7719 77.19
0 -60mV 2760 Fast +10% 50% 7729 77.29
-50Mhz -90mV 2760 Fast +10% Default 7735 77.35
0 -80mV 2760 Fast +10% 50% 7784 77.84

I'm lazy so just tried your last one without really thinking and got Timespy from 28k to 31k (GPU score) no bother. Nice work!