The Radeon RX9070XT / RX9070 Owners Thread

[hornetstinger]

I did a bit of snooping on reddit and found other people who are having similar issues to me with the 9070 xt reaper (high pitched ringing sound from the fans). Got a replacement on the way and sending the current one back so we'll see if it's a faulty card, or if Powercolor just don't really care about this stuff (and it's me being overly sensitive to this sort of thing).

Either way, if it still annoys me, I might upgrade to a larger ITX case and just go for a 9070 XT which is more catering towards those who want silence. I have an Ncase M1 so the Reaper was the only choice for me but if I upgrade to a larger case, then I supposed I could go for osmething like a Sapphire Pure.

Fans on my Sapphire pure are silent, until I manually increase the rpm.

Bit of VRM squel if I unlock framerate though

 

[Kirkatron]

am i the only one whos using a mild [-40 to -45mv usually] undervolt and not reducing the power limit? i actually set my power limit to +10% my thinking is where ive left the actual clocks alone im making it use slightly less power per cycle [by undervolting] which both leaves more power available for more cycles per second and same with the heat generated if your generating slightly less at x speed you now have more thermal headroom.....
the idea being that in theory at least it can now use its boost clocks a lot mroe, so not getting any more actual clock out of it but being able to use the boost for longer more sustained periods of time before having to clock down for heat or power reasons
seems everyone ive read so far have been pulling hte power limits down but im sticking mine up as far as it will allow me lol [+10%]

I have +10% and use FPS limit over power limit. Don't see the point in limiting the power as it could end up with a poor experience in demanding areas of a game that temp need extra power to produce the same FPS. Some people are losing a fair chunk of performance with their power limits and seem ok with it, like 10fps in Steel Nomad 70 vs 80 fps. Maybe they have thermal issues or care about saving a few pennies on the electric bill, i'm not sure. For me, I want the most performance the card will give available at any time for a smooth gaming experience.

 

[Tommy_Here]

Hi All,

Just thought I'd chirm in this discussion as I also have a 9070 Xt (Red Devil) and I've always wondered how my daily settings compared to a Max OC.. anyway here's my results using Steel Nomad Stress Test. I will be using my Daily 250w profile as the baseline.

*Key note - on my 250w & 360w profile, I used 2714 Memory & -40MV. 330w Was ran out of the box settings for the Red devil.

250w Profile: https://www.3dmark.com/3dm/147502415?
Performance - Best 68.96 fps
GPU Clock - Max 2589mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 81 Degrees (17 Delta)
Fan Speed - Max 35% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 999 RPM
Board Power - Max 273w

330w Profile (Stock) - https://www.3dmark.com/3dm/147506128?
Performance - Best 72.14 fps
GPU Clock - Max 2827mhz
Memory Clock - 2500mhz
GPU Temp - Max 66 Degrees
GPU Temp Hot Spot - Max 92 Degrees (26 Delta)
Fan Speed - Max 50%
Fan Speed RPM - Max 1588RPM
Board Power - Max 348W
Performance Increase - 4%
Power increase - 32%

360w Profile - https://www.3dmark.com/3dm/147504504?
Performance - Best 75.10 fps
GPU Clock - Max 2960mhz
Memory Clock - 2700mhz
GPU Temp - Max 73 Degrees
GPU Temp Hot spot - Max 102 degrees (29 Delta)
Fan Speed - Max 42%
Fan Speed RPM - Max 1298rpm
Board Power - Max 467w
Performance Increase - 8%
Power Increase - 44%

Conclusion - I can see why I use the 250w profile - it offers the lowest Delta between GPU temp and Hot spot, the GPU & Memory Temps remain lower than Stock & Max OC whilst also being quieter with the fans hardly spinning (Max 999rpm after a stress test) all whilst offering 92% performance of the Max OC card.

the RX 9070 non XT get's praised for it's powererformance whilst the XT is known for drawing much higher Power but it was my objective to get as much performance out of my card as I could at a more reasonable 250w which is more in line with acceptable imo.

Hope this insight could help anyone out - Would be good to see any 9070 Non XT post there score here and Power draw so i can see the difference

 

[Dg834man]

I have +10% and use FPS limit over power limit. Don't see the point in limiting the power as it could end up with a poor experience in demanding areas of a game that temp need extra power to produce the same FPS. Some people are losing a fair chunk of performance with their power limits and seem ok with it, like 10fps in Steel Nomad 70 vs 80 fps. Maybe they have thermal issues or care about saving a few pennies on the electric bill, i'm not sure. For me, I want the most performance the card will give available at any time for a smooth gaming experience.

Further on from this from my own experience large undervolt plus negative power limit, i get quite hilarious levels of clock stretching, like gpu clock - 3349mhz gpu clock effective - 2589

 

[eeii]

Hi All,

Just thought I'd chirm in this discussion as I also have a 9070 Xt (Red Devil) and I've always wondered how my daily settings compared to a Max OC.. anyway here's my results using Steel Nomad Stress Test. I will be using my Daily 250w profile as the baseline.

*Key note - on my 250w & 360w profile, I used 2714 Memory & -40MV. 330w Was ran out of the box settings for the Red devil.

250w Profile: https://www.3dmark.com/3dm/147502415?
Performance - Best 68.96 fps
GPU Clock - Max 2589mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 81 Degrees (17 Delta)
Fan Speed - Max 35% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 999 RPM
Board Power - Max 273w

330w Profile (Stock) - https://www.3dmark.com/3dm/147506128?
Performance - Best 72.14 fps
GPU Clock - Max 2827mhz
Memory Clock - 2500mhz
GPU Temp - Max 66 Degrees
GPU Temp Hot Spot - Max 92 Degrees (26 Delta)
Fan Speed - Max 50%
Fan Speed RPM - Max 1588RPM
Board Power - Max 348W
Performance Increase - 4%
Power increase - 32%

360w Profile - https://www.3dmark.com/3dm/147504504?
Performance - Best 75.10 fps
GPU Clock - Max 2960mhz
Memory Clock - 2700mhz
GPU Temp - Max 73 Degrees
GPU Temp Hot spot - Max 102 degrees (29 Delta)
Fan Speed - Max 42%
Fan Speed RPM - Max 1298rpm
Board Power - Max 467w
Performance Increase - 8%
Power Increase - 44%

Conclusion - I can see why I use the 250w profile - it offers the lowest Delta between GPU temp and Hot spot, the GPU & Memory Temps remain lower than Stock & Max OC whilst also being quieter with the fans hardly spinning (Max 999rpm after a stress test) all whilst offering 92% performance of the Max OC card.

the RX 9070 non XT get's praised for it's powererformance whilst the XT is known for drawing much higher Power but it was my objective to get as much performance out of my card as I could at a more reasonable 250w which is more in line with acceptable imo.

Hope this insight could help anyone out - Would be good to see any 9070 Non XT post there score here and Power draw so i can see the difference

I noticed this on mine in borderlands 4 after 300 watts the efficiency drops off a cliff. Currently playing dieing light which is 10 years old and capped at 100 uses about 100 watts. Once I’ve finished with that I’ll start on hogwalts legacy (currently free on epic games store) which will be a much sterner test to see what tuning looks like.

 

[Andromeda1983]

For the price you would have paid, i think i would put the effort in, even if just to sell on.

Update: Corsair sending me a replacement AX1600i...

 

[freddie64]

Update: Corsair sending me a replacement AX1600i...

Great result and fair play to Corsair for the quick turnaround.

 

[Andromeda1983]

Yeah im really quite surprised, happily for once

 

[Kirkatron]

Hi All,

Just thought I'd chirm in this discussion as I also have a 9070 Xt (Red Devil) and I've always wondered how my daily settings compared to a Max OC.. anyway here's my results using Steel Nomad Stress Test. I will be using my Daily 250w profile as the baseline.

*Key note - on my 250w & 360w profile, I used 2714 Memory & -40MV. 330w Was ran out of the box settings for the Red devil.

250w Profile: https://www.3dmark.com/3dm/147502415?
Performance - Best 68.96 fps
GPU Clock - Max 2589mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 81 Degrees (17 Delta)
Fan Speed - Max 35% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 999 RPM
Board Power - Max 273w

330w Profile (Stock) - https://www.3dmark.com/3dm/147506128?
Performance - Best 72.14 fps
GPU Clock - Max 2827mhz
Memory Clock - 2500mhz
GPU Temp - Max 66 Degrees
GPU Temp Hot Spot - Max 92 Degrees (26 Delta)
Fan Speed - Max 50%
Fan Speed RPM - Max 1588RPM
Board Power - Max 348W
Performance Increase - 4%
Power increase - 32%

360w Profile - https://www.3dmark.com/3dm/147504504?
Performance - Best 75.10 fps
GPU Clock - Max 2960mhz
Memory Clock - 2700mhz
GPU Temp - Max 73 Degrees
GPU Temp Hot spot - Max 102 degrees (29 Delta)
Fan Speed - Max 42%
Fan Speed RPM - Max 1298rpm
Board Power - Max 467w
Performance Increase - 8%
Power Increase - 44%

Conclusion - I can see why I use the 250w profile - it offers the lowest Delta between GPU temp and Hot spot, the GPU & Memory Temps remain lower than Stock & Max OC whilst also being quieter with the fans hardly spinning (Max 999rpm after a stress test) all whilst offering 92% performance of the Max OC card.

the RX 9070 non XT get's praised for it's powererformance whilst the XT is known for drawing much higher Power but it was my objective to get as much performance out of my card as I could at a more reasonable 250w which is more in line with acceptable imo.

Hope this insight could help anyone out - Would be good to see any 9070 Non XT post there score here and Power draw so i can see the difference

But why not just set an FPS limit rather than a power limit when we all know there are areas of games that can suddenly require a bump in power, especially when using RT or changing environments. I see mine spike above 300w a lot during a game of BF6 when frame limited. Had I been power limited I'd have seen a dip in FPS instead, not ideal.

I think in your case, pardon the pun, you fall into the thermal issue category I mentioned. Which I can see a reason for using a power limit. 75 FPS for max OC in Steel Nomad on a 9070 XT seems pretty poor although looks like you have poor case cooling as over 100 hotspot is insane. Why did you use a lower fan speed for max power when it looks like you have thermal issues?
It's usually around 80 FPS on the 360w cards.

 

[Tommy_Here]

But why not just set an FPS limit rather than a power limit when we all know there are areas of games that can suddenly require a bump in power, especially when using RT or changing environments. I see mine spike above 300w a lot during a game of BF6 when frame limited. Had I been power limited I'd have seen a dip in FPS instead, not ideal.

I think in your case, pardon the pun, you fall into the thermal issue category I mentioned. Which I can see a reason for using a power limit. 75 FPS for max OC in Steel Nomad on a 9070 XT seems pretty poor although no looks like you have poor case cooling as over 100 hotspot is insane. Why did you use a lower fan speed for max power when it looks like you have thermal issues?
It's usually around 80 FPS on the 360w cards.

Id rather keep the noise to a minimum + I feel 250w is the sweet spot for my use case - I have a Corsair 4500D High airflow which does offer good airflow it’s just a conservative fan curve - I’ll try again with a more aggressive fan curve plus my card doesn’t like to go below- 40 mv for 100% stability across all games which seems worse than a lot of cards on here?

 

[Kirkatron]

Id rather keep the noise to a minimum + I feel 250w is the sweet spot for my use case - I have a Corsair 4500D High airflow which does offer good airflow it’s just a conservative fan curve - I’ll try again with a more aggressive fan curve plus my card doesn’t like to go below- 40 mv for 100% stability across all games which seems worse than a lot of cards on here?

Yeah -40mV isn't loads but probably not the worst either. Not sure how big a difference it makes overall. I'm -95mV at the moment and it seems to be getting better with age. I've got a 5000D airflow, older than your case, not sure how similar they are for cooling but my card stays pretty cool.
I like quiet too but I've always used a frame cap rather than power cap to achieve that. I suppose it's swings and roundabouts as they both have their pros and cons.

 

[Aurhinius]

Joined the club with a 9070XT in the misses PC Gigabyte Gaming OC to replace a 3080.

 

[eeii]

Joined the club with a 9070XT in the misses PC Gigabyte Gaming OC to replace a 3080.

Let us know how you get on.

 

[Tommy_Here]

Hi All,

Just thought I'd chirm in this discussion as I also have a 9070 Xt (Red Devil) and I've always wondered how my daily settings compared to a Max OC.. anyway here's my results using Steel Nomad Stress Test. I will be using my Daily 250w profile as the baseline.

*Key note - on my 250w & 360w profile, I used 2714 Memory & -40MV. 330w Was ran out of the box settings for the Red devil.

250w Profile: https://www.3dmark.com/3dm/147502415?
Performance - Best 68.96 fps
GPU Clock - Max 2589mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 81 Degrees (17 Delta)
Fan Speed - Max 35% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 999 RPM
Board Power - Max 273w

330w Profile (Stock) - https://www.3dmark.com/3dm/147506128?
Performance - Best 72.14 fps
GPU Clock - Max 2827mhz
Memory Clock - 2500mhz
GPU Temp - Max 66 Degrees
GPU Temp Hot Spot - Max 92 Degrees (26 Delta)
Fan Speed - Max 50%
Fan Speed RPM - Max 1588RPM
Board Power - Max 348W
Performance Increase - 4%
Power increase - 32%

360w Profile - https://www.3dmark.com/3dm/147504504?
Performance - Best 75.10 fps
GPU Clock - Max 2960mhz
Memory Clock - 2700mhz
GPU Temp - Max 73 Degrees
GPU Temp Hot spot - Max 102 degrees (29 Delta)
Fan Speed - Max 42%
Fan Speed RPM - Max 1298rpm
Board Power - Max 467w
Performance Increase - 8%
Power Increase - 44%

Conclusion - I can see why I use the 250w profile - it offers the lowest Delta between GPU temp and Hot spot, the GPU & Memory Temps remain lower than Stock & Max OC whilst also being quieter with the fans hardly spinning (Max 999rpm after a stress test) all whilst offering 92% performance of the Max OC card.

the RX 9070 non XT get's praised for it's powererformance whilst the XT is known for drawing much higher Power but it was my objective to get as much performance out of my card as I could at a more reasonable 250w which is more in line with acceptable imo.

Hope this insight could help anyone out - Would be good to see any 9070 Non XT post there score here and Power draw so i can see the difference

I've just done another test for Giggles at 230w to see what performance I could get

I scored 1 in Steel Nomad Stress Test

Intel Core i9-9900KF Processor, AMD Radeon RX 9070 XT x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Performance - Best 68.07 fps
GPU Clock - Max 2584mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 78 Degrees (14 Delta)
Memory Temp - Max 84 degrees
Fan Speed - Max 30% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 782 RPM
Board Power - Max 249w
99% performance of 250w using 8% less power... I know people will want to push their hardware to the max and that used to be me, but lately efficiency and noise I am most keen about... overall 10% less performance for 56% less power against the 360w power settings.

 

[Kirkatron]

I've just done another test for Giggles at 230w to see what performance I could get

I scored 1 in Steel Nomad Stress Test

Intel Core i9-9900KF Processor, AMD Radeon RX 9070 XT x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Performance - Best 68.07 fps
GPU Clock - Max 2584mhz
Memory Clock - 2700mhz
GPU Temp - Max 64 Degrees
GPU Temp Hotspot - Max 78 Degrees (14 Delta)
Memory Temp - Max 84 degrees
Fan Speed - Max 30% (Using very conservative profile for silent operation)
Fan Speed RPM - Max 782 RPM
Board Power - Max 249w
99% performance of 250w using 8% less power... I know people will want to push their hardware to the max and that used to be me, but lately efficiency and noise I am most keen about... overall 10% less performance for 56% less power against the 360w power settings.

You'd need to do frame time analysis for any of this to be meaningful and on more games, as average FPS in a single benchmark is only part of the story.
Frame time analysis will show the and 0.1% and 1% mins will be effected more than the average.
I'm 20% higher than this average which suggests my mins will be quite a lot higher as the peaks will be similar on these lightly loaded scenes.
I wonder if anyone has recorded the frame times. Also wonder how a 9070 none XT does with XT bios at same power. On your 9900K I reckon a 9070 wouldn't be far off as you'll be CPU limited in more modern games.

 

[Aresias]

Hi,

I just arrived and I’m not going to read through all the pages of the thread, but here are my observations on the 9070 XT as well as the OC/UV settings.
Please feel free to contradict me or add to this if you’ve noticed anything different.

Goal: Achieve maximum stable performance for real gaming (not just chasing high scores in unstable benchmarks).

Rig:

GPU: PowerColor Hellhound RX 9070 XT
CPU: R7 7700 with +200 frequency and -36 curve optimizer (not 100% stable in P95 tests but never crashed for a year)
RAM: 2×16GB 6000 CL30 with “aggressive timings” in BIOS
PSU : 850W
Display: LG OLED C2 42" 120Hz, 4K display playing most games upscaled from 1440p

The 9070 XT is power-limited when trying to increase frequency, even with +10% Power Limit.
Undervolting the card allows higher frequencies, but by default, even at +0 frequency offset, the card tries to go too high, which results in crashes: My card is only stable at -40mV if I leave the frequency offset at 0, and it still hits the power limit even at +10% PL.
The best compromise for maximum performance seems to be both undervolting and applying a negative frequency offset, which allows reaching higher stable frequencies and more consistently. Even the default +0 frequency is too much and unrealistic for this card.
Memory overclocking is tricky due to ECC; my card seems limited around 2700MHz, I would need to check what memory is used.

That said, here’s the setting that seems to offer the best performance compromise so far after a few hours of stability testing on 3DMark:
20 loops Solar Bay Extreme, 20 loops Time Spy Extreme, 20 loops Steel Nomad, and a few hours of gaming:

-125MHz Max Frequency
-100mV Voltage
2685MHz Memory with fast timings
+10% PL

It is not my final settings since I need to play more games and for a longer time.

Observations :
Most games run at around 3200–3250MHz and the power limit is never reached.
3DMark Nomad around 3025Mhz average frequency, 76fps.
2700Mhz memory frequency definitively give me worse results but I need more testing.
I set my custom fan curve so I never reach more than 85°C Hotspot and my memory reach 76°C maximum.

Is it better or worse to use Fast Timings for my card ?

 

[t2na]

I just arrived and I’m not going to read through all the pages of the thread

Poor excuse tbh, 222 pages of absolute gold.

 

[O.Giroud]

If -100mv is stable for everyday gaming you have won the silicon lottery !

 

[mark 11:11]

i accidentally after doing a benchmark run the other day trying to see if i could beat my old scores and i tried i think it was -95mv
and proceeded to completely forget to set it back to the -45mv i usually leave it set on for daily use
think thats nearly a week before i realised it was set that far
however ive not done anything that would have really asked the gpu to even wake up in that week no gaming or anything i might have done like 10 minutes of gta v actually but not 100% sure i even did that
however the fact it took me nearly or even possibly more than a week to notice..
probably proves absolutely nothing LOL!
but for web browsing and occasionally badly attempting to play guitar using a vst............
oh actually i ran furmark a few times first thing in the morning for like half an hour to try warm myself up a bit [partly as a joke partly because i have no idea lol it fairly certain its achieving nothing]
but yeah i doubt id be able to run -100mv and be completely stable in games so yeah id say they won the silicone lottery

 

[Aresias]

I’ve done further testing and made some very interesting discoveries:

1) The heavier the benchmark, the easier it is to undervolt because the GPU frequency doesn’t go as high.

→ For example, Speed Way allows for insane undervolts that crash almost instantly in most games when using the same settings.
Whereas in Wild Life Extreme, the fps are mostly between 300 and 400, and I crash within a few seconds with my -100mV setting.

2) Every time my card crashes, the same strange phenomenon occurs: suddenly the card jumps to 3300 MHz (usually a jump of more than 100 MHz at once, sometimes 300Mhz !) and holds that frequency for a few seconds, even during scene changes that normally cause large FPS variations the card keep boosting at 3300Mhz. This sustained 3300 MHz spike systematically causes a crash.

→ I ran many tests on Wild Life Extreme at -90, -85, and -80 mV, and I even tried setting an FPS limit of 240 in the AMD Driver to see whether going too high in FPS could be the reason for the crash.
At one point, I managed to complete 20 loops at -80 mV without issues. Then I did more tests and tried another 20 loops run at -80 mV, and this time I crashed on the 17th loop.
After analyzing the data again, I saw that I crashed for the same reason: the 17th loop was the only one where, at the end, the card spiked to 3300 MHz for a few seconds before crashing.

This is disappointing because if i could set a real max frequency number slightly lower than 3300Mhz i could probably be stable à -100mV in every scenarios.
But it's OK my solution could be to have -100mV for heavy lower than 150 fps games and another profile at maybe -75mV for light games.
But again today I played for two hours and didn’t have any issues with my -100mV offset.