RX 9000 series GPU Owners Club

[AusWolf]

I just got it yesterday from Microcenter, testing today for frames, temps and checking if it pops some coil whine.
I pulled the trigger since my wife supported me mentally to get it and have no remorse. $719.99 is too much imo for a gpu but since i already grabbed the 9800x3d on xmass
im putting together a balanced system to forget myself about upgrading for at least 3 more years to come.
Pc gaming or i should say gpu market is getting stupidly overpriced and games are less and less optimized.
but well this is not the thread for my personal rant
Count me in, last amd gpu i had was the hd5870 so im practically a noob
can you recommend me tests and software that i should be running to test my card.



thanks in advance

View attachment 401329

It's not your personal rant. Everybody with half a brain cell thinks the same. Graphics cards and games have become stupidly expensive.

And congrats for the card! I'm sure you'll enjoy it for years to come.

I'm in the same shoes as you, BTW. I upgraded from a 6750 XT that was already enough for me, only because I managed to grab my card for MSRP on launch day. I'm also not planning to upgrade it for at least 2-3 generations.

Welcome to the party

---

Playing some Ninja Gaiden 2 Black on the 4K TV. Figured I'd load up an overlay to capture some numbers and caught some wild power spikes, decided to record a little bit to share.

Bare in mind my TBP is rated at 240 (9070 NITRO+).
View attachment 401258View attachment 401257
View attachment 401260View attachment 401259

Man, that game looks so... foggy... and brown.

 

[spacer00ster]

Hey guys. Does Speed Way crash for anyone here? It crashes for me EVERYTIME - It loads, I hear what I believe is the benchmark audio (no video), then poof, it crashes to desktop and the driver recovers.

Note that I am on Unraid with GPU passthrough to Windows 11 VM. This weekend I plan to shut down Unraid and boot into a test Win11 disk I have but just wanted to know if anyone has ever experienced issues like this.

Update guys: Speed Way does run perfectly fine on my standalone Windows 11 disk.

Ran Steel Nomad & Speed Way stress tests (20 loops) with -110mv, VRAM 2750MHz and both -30% & +10% PL with my RD 9070XT, no crashes. VRAM didn't go higher than 92 deg, with GPU fan not more than 1500 rpm (super quiet) on auto. Didn't mess with the UV, VRAM or fan any more than this.

 

[AusWolf]

Update guys: Speed Way does run perfectly fine on my standalone Windows 11 disk.

Ran Steel Nomad & Speed Way stress tests (20 loops) with -110mv, VRAM 2750MHz and both -30% & +10% PL with my RD 9070XT, no crashes. VRAM didn't go higher than 92 deg, with GPU fan not more than 1500 rpm (super quiet) on auto. Didn't push the UV, VRAM or fan.

So basically, more confirmation that our high VRAM temps aren't catastrophic (among other things).

 

[rusty caterpillar]

Magic GPU! It smells different for men and women!

Congrats, by the way! Not many of us dare to buy PC hardware with such a bold design.

You can't imagine my joy deshrouding it.

I like the other Sakura better, touched my hot spot immediately.
However are some improvements that Yelston can make.... to the shroud.

You can see the original shroud



And down bellow my idea of improvements




Oh , and very important Ocean scent is not that exciting, definitely the exotic burnt VRAM scent please.

 

[NSR]

Man, that game looks so... foggy... and brown.

Oh that's just the environment design for that particular level.

Some other levels for comparison:

 

[lexluthermiester]

So basically, more confirmation that our high VRAM temps aren't catastrophic (among other things).

Catastrophic? No. Very much less than good. Yes. Find a way to get those temps under 75C or that card will not last long.

 

[AusWolf]

Catastrophic? No. Very much less than good. Yes. Find a way to get those temps under 75C or that card will not last long.

There is no way to do that. Even during idle with two fans pulling fresh air at it, it sits in the low 60s. Based on what I've read here and in reviews, it's pretty much normal.

 

[rusty caterpillar]

There is no way to do that. Even during idle with two fans pulling fresh air at it, it sits in the low 60s. Based on what I've read here and in reviews, it's pretty much normal.

Yes, only with backplate cooling.
Or change thermal pads https://wccftech.com/amd-radeon-rx-...eplacing-thermal-pads-fixes-high-gddr6-temps/
This card has Samsung chips though, plastic shroud doesn't cover the exhausts.

Edit: Is any way you can remove front glass panel of your 280X case? Or is just the Mallet way? really don't think small side gaps intake left by Corsair are enough for your card. But than I never built a system in the 280X.
I would change from D9 to AIO so I can have room for backplate cooling. Thermalright 240 are not expensive, yet efficient.

 

[Kwadratowicz]

There is no way to do that. Even during idle with two fans pulling fresh air at it, it sits in the low 60s.

Do You have bottom fans that are sucking air for gpu?, and If the case is on stock feets?
Cause You can add few cm to clearance for bottom intake, it should help with temps.
With Lian Li Lancool 207, and 2 bottom fans on minimal rpm plus minimal gpu fans rpm I've got around 50C on vram while browsing internet.
Is it Your setup? (I've found picture in Your thread )

Can You check temps with only bottom top fans spinning, and with double the height of the legs of the case?
(I think that front fans can disturb the flow of air. They add resistance for bottom fans, and gpu fans to exhaust hot air. In My opinion).
And with that much clearance You can add heatsink in the back of the card itself
Edit, You can even block front and back with cardboard and see if it helps with temperatures .

 

[rusty caterpillar]

Do You have bottom fans that are sucking air for gpu?, and If the case is on stock feets?
Cause You can add few cm to clearance for bottom intake, it should help with temps.
With Lian Li Lancool 207, and 2 bottom fans on minimal rpm plus minimal gpu fans rpm I've got around 50C on vram while browsing internet.
Is it Your setup? (I've found picture in Your thread )
View attachment 401405
Can You check temps with only bottom top fans spinning, and with double the height of the legs of the case?
(I think that front fans can disturb the flow of air. They add resistance for bottom fans, and gpu fans to exhaust hot air. In My opinion).
And with that much clearance You can add heatsink in the back of the card itself

By the look of it are pushing air to the GPU.
Edit: L 207 has good intake, he got only 2x narrow side strips for intake , rest is covered by glass. Also Lan 207 has 2x160mm in front, he got 12 or 14 cm fans.
On the other hand he has enough clearance for backplate cooling, no need to change to AIO for CPU.

I was pointing to this


Maybe removing glass panel and add a frame with mesh with magnets strips.

Is less complicated than on mine, where I have to build a whole frontal panel.

 

[Kwadratowicz]

L 207 has good intake, he got only 2x narrow side strips for intake , rest is covered by glass. Also Lan 207 has 2x160mm in front, he got 12 or 14 cm fans.

Corsair has option to mount fans on bottom, but it has to little in My opinion clearance at the bottom, and Lancool 207 has 2x140 at front. But I'm thinking of going for down to top air pass it could help with cooling gpu. (Lancool 216 has 2x160mm fans at front. )

 

[Bloste]

By the look of it are pushing air to the GPU.
Edit: L 207 has good intake, he got only 2x narrow side strips for intake , rest is covered by glass. Also Lan 207 has 2x160mm in front, he got 12 or 14 cm fans.
On the other hand he has enough clearance for backplate cooling, no need to change to AIO for CPU.

I was pointing to this

View attachment 401407
Maybe removing glass panel and add a frame with mesh with magnets strips.

Is less complicated than on mine, where I have to build a whole frontal panel.

View attachment 401408

View attachment 401410

Now that's some dedication, man. If I had to do that kind of mod to my case... I'd buy a new one.

 

[rusty caterpillar]

Now that's some dedication, man. If I had to do that kind of mod to my case... I'd buy a new one.

I get you, but buying another case would have set me 150-200£ short, modifying it, the mats cost me around 55£ + labor 6-8Hr at least. I had some MDF pieces around and the paint.
At least I went to lower noise from 43dB to -37dB(which I love) and very decent cooling, Is kind of a Fractal Torrent with AIO support for the top of the case.

I paid dearly for choosing this case but, at least it proved it can be done a lot where the manufacturers are cutting corners. You need the drive, the passion and a tad of sprinkled frustration from GPU market masquerade

Frontal Panel you see is just the tip of the Iceberg(For example are some noise guards panels and cork sheets you don't see), I've done a lot to that case, see here:
https://www.techpowerup.com/forums/...and-bottom-fan-14cm-test.337297/#post-5524770

but is not all, I'll make a video at some point.

 

[iSpeakVeryWell]

Yes, only with backplate cooling.
Or change thermal pads

I tried many things: I replaced thermal pads, I put heatsink on the backplate, I put a fan throwing air at the back of the card. Result? I get the exact same temps than stock. I just wasted money and wasted a perfectly fine application of PTM7950. My recommendation is: don't waste time with that.

 

[rusty caterpillar]

Corsair has option to mount fans on bottom, but it has to little in My opinion clearance at the bottom, and Lancool 207 has 2x140 at front. But I'm thinking of going for down to top air pass it could help with cooling gpu. (Lancool 216 has 2x160mm fans at front. )

Thanks for clarification on 160mm fans, my bad, going from down to top cooling has a major obstruction and turbulence > the GPU in normal horizontal position. IMO front to rear and ceiling is better.
As you point it out, one of the top fans hinders IMO a lot the cooling, he rather put that as intake or remove it as close the gap with a carboard or something, wood panel or al panel.


IMO I would really change that. That top fan takes front air intake right out of the case and can really reduce airflow for CPU cooling if CPU cooler has normal positioning blowing to the rear.
GN mention something similar in one of their videos. They also tested.

I tried many things: I replaced thermal pads, I put heatsink on the backplate, I put a fan throwing air at the back of the card. Result? I get the exact same temps than stock. I just wasted money and wasted a perfectly fine application of PTM7950. My recommendation is: don't waste time with that.

Thanks for sharing.
I beg to differ it worked for me for several cards. The fan in pull not push. if use Heatsink needs to be linked with thermal putty to the backplate. I used mostly thermal putty, later Upsiren 8.
Pads are more expensive, you can mess up the desired thickness, they can leak silicon oil.
Your backplate has no TIM linkage of VRM to the backplate. VRM can get hot as VRAM and is in close proximity to the VRAM. Heat by proximity.


I have less metal in my backplate compared to yours which is a disadvantage but I have a middle plate

All the pads you see and some more was replaced by thermal putty Upsiren UTP 8, it went over 100gr of putty in this card.




HD2 game in 1440P ultra and high almost 2hrs no brake 1300 RPM GPU fans, backplate fans 1000 -1200 in pull, no heatsink, 1100 RPM rear case fan 14 cm fan.
Even if you add 20 C to my VRAM because sensor placement and DDR5X instead of DDR6, I have 75C for VRAM and VRM > PWR4 sensor.




Superposition 8k cinematic loop 22 min, done 3min warm up before so roughly 25 min.



Time Spy and Nomad temps are way lower than Superposition 8K 22 min.

You think I achieve this by some miracle? devil is the detail. I wish I could help you further.

Also AVATARAT can confirm his lower temps with 14cm fan on the back of an 9070XT Prime, I use 2x92mm atm, usually is only one, I'm testing the endurance in long sessions 1vs 2 fans. I can't hear the at that low RPM.

 

[AVATARAT]

Also AVATARAT can confirm his lower temps with 14cm fan on the back of an 9070XT Prime, I use 2x92mm atm, usually is only one, I'm testing the endurance in long sessions 1vs 2 fans. I can't hear the at that low RPM.

Yes, I can confirm that my 140mm fan on the back of the card works very well and brings lower VRAM/Hotspot temperatures. But the airflow in the case must be good because this fan needs cold air to work.

 

[Kwadratowicz]

Small test:
Test Subject: Gigabyte Gaming OC 9070 XT in Lian Li Lancool 207

cpu has curve that over 60C is 30% pwm on fans.
Test was steel nomad 10 pass stresstest (from 20 passes stress test, that's why I don't have data from it).

Test setup 1: Front Fans off, bottom fans 40% pwm, top fan 40% pwm, gpu fans 45% pwm:
Results :

Test seup 2: all case fans set to 40% pwm, and gpu fans set to 45% pwm

Conclusion, With front fans OFF gpu could achieve around 3200MHz boost clock, but cpu was sitting 62C,
With front fans ON gpu was below 3000MHz boost clock, but cpu was colder with 54C.

So it is preferable to not utilize front fans cause of the lower boost clock in gpu.
But results may vary in other cases, and cases .

I get the exact same temps than stock

what about clocks?

 

[rusty caterpillar]

Small test:
Test Subject: Gigabyte Gaming OC 9070 XT in Lian Li Lancool 207
View attachment 401480
cpu has curve that over 60C is 30% pwm on fans.
Test was steel nomad 10 pass stresstest (from 20 passes stress test, that's why I don't have data from it).

Test setup 1: Front Fans off, bottom fans 40% pwm, top fan 40% pwm, gpu fans 45% pwm:
Results :
View attachment 401458
Test seup 2: all case fans set to 40% pwm, and gpu fans set to 45% pwm
View attachment 401457

Conclusion, With front fans OFF gpu could achieve around 3200MHz boost clock, but cpu was sitting 62C,
With front fans ON gpu was below 3000MHz boost clock, but cpu was colder with 54C.

So it is preferable to not utilize front fans cause of the lower boost clock in gpu.
But results may vary in other cases, and cases .

Thanks for the test.

I can't run Steel Nomad X10 with free version am I?

I find that weird behavior coming from your card/case...
1.a Did you touched power, UV, OC for the card before running the tests, even if you revert changes to the stock one prior the tests?
1.b I think with all fans on is a warm cloud around your GPU and is not moving fast enough > frontal fans when on, are pushing the hot air around GPU and the bottom fans can't get rid of them fast enough IDK. Rear fan is not present, I would use one.
2. Also there is possibility Adrenaline is doing something to the voltages of the GPU IMO
3. You should run a third test , bottom fans off, close the gap with carboard to not lose pressure from front fans use a rear fan @1300 RPM.
4. If CPU worked more for some reason in first bunch of tests(some background apps) , the temp maybe was higher and warmed up the backplate of the GPU higher than for the second test but that is far fetched a bit.
5. GN tested with 2x bottom + 2x frontal but no top fans or rear, 3d place in top with the stock fans. Second test with their fans, 2x front + 1 rear case came on second place.

6. Remove that guy, he jinxed the tests or exchange the card with a Sakura 9070XT with Ocean Scent, the croc will like that, he will not jinx anymore.



observation: In both tests, temps of GPU are very similar yet clocks are higher only in one test, maybe you should have show the AVG temps of GPU and not max.

Somebody reported something similar for the Fractal Torrent case where he had to reverse the airflow rear fan intake and frontal 2x 180mm as exhaust, GPU was hotter with the normal config.
What are the RPM of your bottom fans in the test where frontal ones was off? roughly 800-1000RPM ?

In the end of the day you have lower noise with the front off and focused the cooling on the GPU which is good as easts more wattage than the CPU, you can leave it like that if you are happy. More fans is not always better as we know already.

Hope it helps

 

[Krit]

I find that weird behavior coming from your card/case...

False test 100%. I would change upper fan to exhaust. Warm air will escape anyways as it's lighter. Your PC front is too close to the wall it needs space for breathing.

Kwadratowicz you are blocking psu exhaust

​

 

[Kwadratowicz]

False test 100%. I would change upper fan to exhaust. Warm air will escape anyways as it's lighter.

Kwadratowicz you are blocking psu exhaust

I'm not, psu in Lancool 207 is in front, and it is getting fresh air from bottom, and is exhausting it in the back of case, strange layout that I love
I think You mean cpu, the fan at top is exhaust I forgot to mention that.

I can't run Steel Nomad X10 with free version am I?

Unfortunately not.

maybe you should have show the AVG temps of GPU and not max.

I will be testing when Noctua will send me bracket to mounting vertical nh d15 g2 ,
Cause I've got stupid idea in mind

What are the RPM of your bottom fans

Bottom fans around 1500 rpm 40% pwm (120mm)
and front fans and top around 1600rpm 40% pwm(140mm)

1.a Did you touched power, UV, OC for the card before running the tests,

-30mV and 2700 MHz vram, fans from gpu where manually set to 45%.

1.b I think with all fans on is a warm cloud around your GPU and is not moving fast enough > frontal fans when on, are pushing the hot air around GPU and the bottom fans can't get rid of them fast enough IDK. Rear fan is not present, I would use one.

Yes I think it is the case here, that's why I will be doing some strange testing with different fan speeds, I think from 20% and then I will add steel nomad stress test as data too.

2. Also there is possibility Adrenaline is doing something to the voltages of the GPU IMO

I'm not using adrenaline, only stock driver and moreclocktool.

3. You should run a third test , bottom fans off, close the gap with carboard to not lose pressure from front fans use a rear fan @1300 RPM.

Cardboard is waiting for Noctua.

4. If CPU worked more for some reason in first bunch of tests(some background apps) , the temp maybe was higher and warmed up the backplate of the GPU higher than for the second test but that is far fetched a bit.

Between test time was added to cool down everything to similar temperatures +-1C between run starts, but it was only checked for cpu and gpu, not mobo and ram.

5. GN tested with 2x bottom + 2x frontal but no top fans or rear, 3d place in top with the stock fans. Second test with their fans, 2x front + 1 rear case came on second place.

I will be testing with strange fan configuration, that does not make any sense too . for example cpu cooler horizontal with air going from back to front

6. Remove that guy, he jinxed the tests or exchange the card with a Sakura 9070XT with Ocean Scent, the croc will like that, he will not jinx anymore.

It's gecko, not croc -.- It's still not in it's bipedal form

Hope it helps

Yes it is, I've found much more ideas how to do tests in future . just for fun
And it's true that gpu is much more powerhungry I've got my cpu set to 45W tdp with 180W cooler , and It's waiting for 240W cooler

 

[iSpeakVeryWell]

Thanks for clarification on 160mm fans, my bad, going from down to top cooling has a major obstruction and turbulence > the GPU in normal horizontal position. IMO front to rear and ceiling is better.
As you point it out, one of the top fans hinders IMO a lot the cooling, he rather put that as intake or remove it as close the gap with a carboard or something, wood panel or al panel.

View attachment 401428
IMO I would really change that. That top fan takes front air intake right out of the case and can really reduce airflow for CPU cooling if CPU cooler has normal positioning blowing to the rear.
GN mention something similar in one of their videos. They also tested.


Thanks for sharing.
I beg to differ it worked for me for several cards. The fan in pull not push. if use Heatsink needs to be linked with thermal putty to the backplate. I used mostly thermal putty, later Upsiren 8.
Pads are more expensive, you can mess up the desired thickness, they can leak silicon oil.
Your backplate has no TIM linkage of VRM to the backplate. VRM can get hot as VRAM and is in close proximity to the VRAM. Heat by proximity.

View attachment 401430
I have less metal in my backplate compared to yours which is a disadvantage but I have a middle plate

All the pads you see and some more was replaced by thermal putty Upsiren UTP 8, it went over 100gr of putty in this card.




HD2 game in 1440P ultra and high almost 2hrs no brake 1300 RPM GPU fans, backplate fans 1000 -1200 in pull, no heatsink, 1100 RPM rear case fan 14 cm fan.
Even if you add 20 C to my VRAM because sensor placement and DDR5X instead of DDR6, I have 75C for VRAM and VRM > PWR4 sensor.


View attachment 401432

Superposition 8k cinematic loop 22 min, done 3min warm up before so roughly 25 min.

View attachment 401433

Time Spy and Nomad temps are way lower than Superposition 8K 22 min.

You think I achieve this by some miracle? devil is the detail. I wish I could help you further.

Also AVATARAT can confirm his lower temps with 14cm fan on the back of an 9070XT Prime, I use 2x92mm atm, usually is only one, I'm testing the endurance in long sessions 1vs 2 fans. I can't hear the at that low RPM.

Apology, my message came out rude, it was not the goal, I wrote it in a rush before going to work and I came out wrong.

I think it's all very contextual. I probably got at a point where I can't do more with air. I think the AsRock 9070 XT Taichi come ootb with good TIM and pads too. I often forget that not all build are equals. Some build have more headroom than other.

Anyway I had a lot of fun in the process of trying to get better memory temps, and this what matter the most!

Yes, I can confirm that my 140mm fan on the back of the card works very well and brings lower VRAM/Hotspot temperatures. But the airflow in the case must be good because this fan needs cold air to work.

Do you have a photo of the result? I'm curious to see how you managed that!

Also, what is the max temp you get on memory since you added the fan?

what about clocks?

Same clock, same scores. No improvement, no regression.

 

[AVATARAT]

Do you have a photo of the result? I'm curious to see how you managed that!

Also, what is the max temp you get on memory since you added the fan?

ОК, there is one 14" fan on the back of case too.



And the result, average temperatures during the Steel Nomad stress test (20 loops).



Max temps after the end of stress test.



Of course, I manage that power limit and offset depending on the game I'm playing.
Acceptable fan noise on the card is around 2200 - 2500 RPM.

 

[rusty caterpillar]

I will be testing with strange fan configuration, that does not make any sense too . for example cpu cooler horizontal with air going from back to front

No need to test, how they tested, I was mentioning their setups and the results as tests that was already done on 207.
Second config of their test is not at all strange > 2 front fans + 1 rear, most of the cases have that. Well, most of the case have 3 fans in front, but the lower one is obstructed by chambers 40-60% so they pulled it out I guess .

 

[iSpeakVeryWell]

ОК, there is one 14" fan on the back of case too.

View attachment 401529View attachment 401530

And the result, average temperatures during the Steel Nomad stress test (20 loops).

View attachment 401532

Max temps after the end of stress test.

View attachment 401534

Of course, I manage that power limit and offset depending on the game I'm playing.
Acceptable fan noise on the card is around 2200 - 2500 RPM.

Very nice build! I'm still trying to figure out how you managed to make the fan hold there! You have a very great magic trick!

I can't fit a fan there, because my AIO have an offset mount and the tubes and the pump are too close to the GPU.

I made the same test than you and temps are similar. The only thing that really catch my eyes is the difference in clock! There is a pretty huge gap! Probably because you have -15mV more than me and you also have 10W more.

Thank you for the photos and the temps!

My avg are a bit off because I don't know how to run the stress test windowed.

Edit: Found out how to run stress test in windowed mode!

I've ran a second loop, but this time I've tried to match your settings as closely as I could. I'm missing 100MHz compare to you. I'm 4.7C colder on average on the hot spot, I don't understand why the card don't use the headroom, it's way too conservative.

I'm soooo going to flash the BIOS of this thing! (and brick it! )

 

[Krit]

I'm not, psu in Lancool 207 is in front, and it is getting fresh air from bottom, and is exhausting it in the back of case, strange layout that I love
I think You mean cpu, the fan at top is exhaust I forgot to mention that.