RX 9000 series GPU Owners Club

[spacer00ster]

@TSiAhmat and @spacer00ster, you were both added to the list

Btw, spacer00ster, I added the two cards you purchased, at least for now, you tell me. Really looking forward to hearing if you notice a big difference between the two. Please share your thoughts when you get the chance!

Just as a side note; for now, the ratio is close to 2:1 in terms of people having the XT vs the non-XT, and honestly, I thought it would be more like 3:1 or higher. So I'm pretty happy to see the non-XT isn’t being forgotten at all.

I thought the same because of the power consumption differences between the 2 models which helped my initial decision of going with the non-XT. But notable reviewers kept recommending the XT due the minor +$50 with the noticeable added performance.

For me, that saved $50 of the non-XT would disappear in about a month or 2 in gaming on my electricity bill. In a year, it adds up big time. As I mentioned in my 1st post, I now know about undervolting and power limiting, so I'll do some tweaking to get the XT power draw down.

 

[SirKeldon]

I thought the same because of the power consumption differences between the 2 models which helped my initial decision of going with the non-XT. But notable reviewers kept recommending the XT due the minor +$50 with the noticeable added performance.

For me, that saved $50 of the non-XT would disappear in about a month or 2 in gaming on my electricity bill. In a year, it adds up big time. As I mentioned in my 1st post, I now know about undervolting and power limiting, so I'll do some tweaking to get the XT power draw down.

Yeah, to be honest, I initially tried to grab the XT on launch day, but after 5–10 minutes, all stock was gone. I actually failed the checkout process on two different websites. But after the initial frustration, I had a few minutes to think it through a bit more carefully. I was coming from a GTX 1080 too, just like you, so going with a similar TBP and better performance than my about-to-return 7800 XT felt like a no-brainer. When I saw the Pulse at MSRP, I didn't hesitate this time.

Like you said, my power bill will definitely thank me in the long run. It's not just about saving $50 upfront, but also cutting down power usage for just 10% less performance under stock circumstances.

As I mentioned in the RX 9070 OC thread, I've managed to tune it to: -70mV, -40MHz, and -2% PL w/total game stability. That gives me around a 5–6% performance boost at a 215W TBP, cutting the gap with the XT at stock and coming close to the power draw of my old GTX 1080, but with a 115% increase in performance. Pretty crazy!

 

[spacer00ster]

Thanks for the tips. The situation is a bit sketchy atm because I'm in transition: I just rammed this card into my "old" rig, based on 5600X and the now-sold RTX 4070. It's a pretty basic ATX case with 3 fans...but also I don't think it’s really bad re: airflow, and I also keep it open for now. The ambient at the time was about 25C indoors. I did try to up the fan curve a bit, but didn't really have time to do any proper testing. Also, I just got the 9800X3D and so will build the new rig in the next week or so, most likely with a slightly bigger case and maybe an extra fan.

But even so, I'd have thought that a GPU should respect my current kind of setup regarding temps...they don't say it's a requirement for some super special fan arrangements or case just to keep the temps reasonable.

I saw this quick underclock guide and maybe could try it, but isn't messing around with that invalidating warranty or something?

If the situation persists after my new case build, I will definitely contact the shop.

Hey @cinemaware, have you tried undervolting yet? I'm waiting on the same card you have and I've been looking some forum posts. Many users state that this card does get warm, higher than most other brands but I've not seen any messages with users undervolting and limiting power. If you tried that could you post your findings please?

 

[SirKeldon]

The following is just a set of observations for those who, like me, pay close attention to temperatures (maybe a bit too much )

I've been monitoring the behavior of my RX 9070 during extended gaming sessions. While core and hotspot remain within reasonable ranges, VRAM consistently stands out as the hottest component on the card.

I’ve not been reaching, nor even approaching, 82% of the thermal limit during these sessions, so nothing truly concerning yet. But given how close VRAM can get to its thermal threshold — and considering long-term reliability (electromigration, etc.), as discussed in other threads — I think it's worth keeping an eye on. Not panicking over, just addressing it. Probably says more about my obsession with temps after years of watercooling than anything else, as mentioned before.

With a light undervolt and a slightly more aggressive fan curve, I'm seeing 76–78 °C sustained, with 80 °C max in God of War. Without those tweaks, VRAM would hit 86–88 °C in long sessions. That brought thermal headroom closer to a ~70% average, which I’m more comfortable with.

After switching to a 120Hz monitor, idle VRAM temps rose noticeably due to memory clocks jumping to 2505 MHz (instead of ~900 MHz at 60Hz). Even basic tasks like GPU-accelerated browsing or YouTube playback caused a +5–8 °C increase. Disabling Zero RPM and setting the fans to a 15% minimum brought "light tasking" back down to a max of 54–56 °C.

These small tweaks cost nothing and help keep temps under control with no noticeable noise impact.

 

[cinemaware]

Yesterday I took my 9070 XT (which was getting up to 102C VRAM, in 15C (EDIT: typo, meant 25C obvs) ambient with default fan curve) to the service centre of the shop where I bought it. I thought that they will take it in for some stress tests, but straight away the guy just offered me a refund. A replacement wasn't possible because, surprise surprise, they don't have any in stock.

It was a bit of a strange situation. Because of the language barrier (I'm in Taiwan and my native language skills are quite low) I couldn't really dig as deep as possible when asking him about this quagmire, but on the other hand the guy spoke decent English. I asked whether this is normal or not, whether I am overreacting, and his replies were that a) "You’re not the 1st one with this problem" b) "we tried to ask AMD what to do about it, got no real answers" c) when asked what should I do - wait or refund, "if I was AMD rep, I'd tell you to wait for BIOS update or something, but personally I'd say get the refund".

I was really conflicted at this point, but ended up refunding. The situation is a total AFU, since of course there are no other cards I can buy atm, and quite possibly the prices will rise well over the MSRP in the foreseeable future. But...I always eventually planned to have this card only as a placeholder till I can get 5070Ti or 5080, which were my 1st choices, and still are because reasons. So, I took a gamble taht in the next few months I can find one of them at MSRP. Probably a fool's move, but on the other hand didn't want to end up with a potentially faulty card on my hands. I could've easily sold it with a markup here too, but am trying not to be a dick, even if the rules of this game seem to have been thrown out overall.

The added insult to this injury was that when I went to the actual shop to process the refund, the shelves were full of Gigabyte's 5070Ti/5080 models. The clerk in this one didn't speak English, but when I tried to tell him I'd like to get one instead of refund I was met with a deadpan "these are only for people who build whole computers". This has been standard BS practice here since the shortage, (eg I could only buy my 9800X3D in combo with a B850 motherboard) but still...they could've bent the stupid rule a bit for somebody who just got a
possibly dud card in their shop.

So now I'm stuck with a power build sporting 1000W PSU, World's Best Gaming Processor, etc....and 1050Ti

Also mad at Trump, Lisa, Jacket Man, etc, but most of all every bleedin' reseller who takes advantage of this situation. Y'all really managed to suck the fun out of this hobby.

What a time to be alive, eh

@spacer00ster sorry mate, never tried undervolting because of this "possible RMA" situation. I believe most people can control this with fan curve, eg see @SirKeldon's post above. I also managed to bring it down to 80-90 zone, but my fans were loud as hell and I also live in a very hot country, so in the long run this felt like too much hassle.

 

[Event Horizon]

The following is just a set of observations for those who, like me, pay close attention to temperatures (maybe a bit too much )

I've been monitoring the behavior of my RX 9070 during extended gaming sessions. While core and hotspot remain within reasonable ranges, VRAM consistently stands out as the hottest component on the card.

I’ve not been reaching, nor even approaching, 82% of the thermal limit during these sessions, so nothing truly concerning yet. But given how close VRAM can get to its thermal threshold — and considering long-term reliability (electromigration, etc.), as discussed in other threads — I think it's worth keeping an eye on. Not panicking over, just addressing it. Probably says more about my obsession with temps after years of watercooling than anything else, as mentioned before.

With a light undervolt and a slightly more aggressive fan curve, I'm seeing 76–78 °C sustained, with 80 °C max in God of War. Without those tweaks, VRAM would hit 86–88 °C in long sessions. That brought thermal headroom closer to a ~70% average, which I’m more comfortable with.

After switching to a 120Hz monitor, idle VRAM temps rose noticeably due to memory clocks jumping to 2505 MHz (instead of ~900 MHz at 60Hz). Even basic tasks like GPU-accelerated browsing or YouTube playback caused a +5–8 °C increase. Disabling Zero RPM and setting the fans to a 15% minimum brought "light tasking" back down to a max of 54–56 °C.

These small tweaks cost nothing and help keep temps under control with no noticeable noise impact.

That's pretty much how I've setup the fan curve for my 9070 as well. Do note that fan stop prolongs the life of the fans, so by disabling that we increase the wear and tear on the fan. Might be a good idea to use the default profile for desktop use and only activate my custom profile (fan stop off + more aggressive curve) for gaming sessions.

 

[AusWolf]

Is this "faffing" around with Optiscaler?

TAA. Portal RTX / HL2 RTX only have TAA and NiS.

Edit: And maybe DLSS, I don't know, but you can't use that on a Radeon card.

 

[Zach_01]

I thought the same because of the power consumption differences between the 2 models which helped my initial decision of going with the non-XT. But notable reviewers kept recommending the XT due the minor +$50 with the noticeable added performance.

For me, that saved $50 of the non-XT would disappear in about a month or 2 in gaming on my electricity bill. In a year, it adds up big time. As I mentioned in my 1st post, I now know about undervolting and power limiting, so I'll do some tweaking to get the XT power draw down.

What is your KW/h cost if I may ask?

 

[LittleBro]

About those VRAM temps ...

They are not critical, yet. During summer, with dusted GPU, it may be a problem.

I still don't get this:
7800XT has same VRAM technology, with barely any (not even 100MHz) clock uplift compared to 9070 (XT). 7800 XT can go above 90°C in really rare cases, at least mine. 9070s goes into this range regularly and near 110 rarely.

 

[TSiAhmat]

About those VRAM temps ...

They are not critical, yet. During summer, with dusted GPU, it may be a problem.

I still don't get this:
7800XT has same VRAM technology, with barely any (not even 100MHz) clock uplift compared to 9070 (XT). 7800 XT can go above 90°C in really rare cases, at least mine. 9070s goes into this range regularly and near 110 rarely.

Mine also doesn't hit 90 C° in the new case (86 C° in Memtest), a lot of 9070 cards have very passive fan curves & the hot spot is quiet low (maybe because of the Honeywell pads), which dictates how fast the fans go. I don't think these cards scale as bad in the Summer as we might think. Also Tjmax for Gddr6 Vram is 120+ C°

 

[Jtuck9]

Yesterday I took my 9070 XT (which was getting up to 102C VRAM, in 15C ambient with default fan curve) to the service centre of the shop where I bought it. I thought that they will take it in for some stress tests, but straight away the guy just offered me a refund. A replacement wasn't possible because, surprise surprise, they don't have any in stock.

It was a bit of a strange situation. Because of the language barrier (I'm in Taiwan and my native language skills are quite low) I couldn't really dig as deep as possible when asking him about this quagmire, but on the other hand the guy spoke decent English. I asked whether this is normal or not, whether I am overreacting, and his replies were that a) "You’re not the 1st one with this problem" b) "we tried to ask AMD what to do about it, got no real answers" c) when asked what should I do - wait or refund, "if I was AMD rep, I'd tell you to wait for BIOS update or something, but personally I'd say get the refund".

I was really conflicted at this point, but ended up refunding. The situation is a total AFU, since of course there are no other cards I can buy atm, and quite possibly the prices will rise well over the MSRP in the foreseeable future. But...I always eventually planned to have this card only as a placeholder till I can get 5070Ti or 5080, which were my 1st choices, and still are because reasons. So, I took a gamble taht in the next few months I can find one of them at MSRP. Probably a fool's move, but on the other hand didn't want to end up with a potentially faulty card on my hands. I could've easily sold it with a markup here too, but am trying not to be a dick, even if the rules of this game seem to have been thrown out overall.

The added insult to this injury was that when I went to the actual shop to process the refund, the shelves were full of Gigabyte's 5070Ti/5080 models. The clerk in this one didn't speak English, but when I tried to tell him I'd like to get one instead of refund I was met with a deadpan "these are only for people who build whole computers". This has been standard BS practice here since the shortage, (eg I could only buy my 9800X3D in combo with a B850 motherboard) but still...they could've bent the stupid rule a bit for somebody who just got a
possibly dud card in their shop.

So now I'm stuck with a power build sporting 1000W PSU, World's Best Gaming Processor, etc....and 1050Ti

Also mad at Trump, Lisa, Jacket Man, etc, but most of all every bleedin' reseller who takes advantage of this situation. Y'all really managed to suck the fun out of this hobby.

What a time to be alive, eh

@spacer00ster sorry mate, never tried undervolting because of this "possible RMA" situation. I believe most people can control this with fan curve, eg see @SirKeldon's post above. I also managed to bring it down to 80-90 zone, but my fans were loud as hell and I also live in a very hot country, so in the long run this felt like too much hassle.

What 9070xt did you have?

Maybe we'll start seeing more of these issues bubbling up to the surface?!

GPU Die Defects Found in PowerColor RX 9070 XT Graphics Card

Surface defects on the GPU die of a PowerColor Radeon RX 9070 XT Hellhound have been linked to excessive thermal hotspots, according to analysis following Igor's Lab's investigation. Microscopic examination revealed 1,934 pits across the silicon die surface despite proper thermal interface...

www.techpowerup.com

 

[rusty caterpillar]

Yesterday I took my 9070 XT (which was getting up to 102C VRAM, in 15C ambient with default fan curve) to the service centre of the shop where I bought it. I thought that they will take it in for some stress tests, but straight away the guy just offered me a refund. A replacement wasn't possible because, surprise surprise, they don't have any in stock.

It was a bit of a strange situation. Because of the language barrier (I'm in Taiwan and my native language skills are quite low) I couldn't really dig as deep as possible when asking him about this quagmire, but on the other hand the guy spoke decent English. I asked whether this is normal or not, whether I am overreacting, and his replies were that a) "You’re not the 1st one with this problem" b) "we tried to ask AMD what to do about it, got no real answers" c) when asked what should I do - wait or refund, "if I was AMD rep, I'd tell you to wait for BIOS update or something, but personally I'd say get the refund".

I was really conflicted at this point, but ended up refunding. The situation is a total AFU, since of course there are no other cards I can buy atm, and quite possibly the prices will rise well over the MSRP in the foreseeable future. But...I always eventually planned to have this card only as a placeholder till I can get 5070Ti or 5080, which were my 1st choices, and still are because reasons. So, I took a gamble taht in the next few months I can find one of them at MSRP. Probably a fool's move, but on the other hand didn't want to end up with a potentially faulty card on my hands. I could've easily sold it with a markup here too, but am trying not to be a dick, even if the rules of this game seem to have been thrown out overall.

The added insult to this injury was that when I went to the actual shop to process the refund, the shelves were full of Gigabyte's 5070Ti/5080 models. The clerk in this one didn't speak English, but when I tried to tell him I'd like to get one instead of refund I was met with a deadpan "these are only for people who build whole computers". This has been standard BS practice here since the shortage, (eg I could only buy my 9800X3D in combo with a B850 motherboard) but still...they could've bent the stupid rule a bit for somebody who just got a
possibly dud card in their shop.

So now I'm stuck with a power build sporting 1000W PSU, World's Best Gaming Processor, etc....and 1050Ti

Also mad at Trump, Lisa, Jacket Man, etc, but most of all every bleedin' reseller who takes advantage of this situation. Y'all really managed to suck the fun out of this hobby.

What a time to be alive, eh

@spacer00ster sorry mate, never tried undervolting because of this "possible RMA" situation. I believe most people can control this with fan curve, eg see @SirKeldon's post above. I also managed to bring it down to 80-90 zone, but my fans were loud as hell and I also live in a very hot country, so in the long run this felt like too much hassle.

I feel for you, at this point you have to wait for another model to show up and get that.
Store was happy to refund you as they can sell it as open box pretty fast.
TBH when you said you taking it to store I thought you gonna get a replacement and try your luck with another one.
Do you use AC in your room? I mean in the summer when gets 32-35 C outside.
RMA is never an option IMO.

 

[Jtuck9]

TAA. Portal RTX / HL2 RTX only have TAA and NiS.

Edit: And maybe DLSS, I don't know, but you can't use that on a Radeon card.

It would be cool to get properly implemented AMD path tracing variants somewhere down the line

 

[LittleBro]

Has anyone tried replacing thermal pads yet?

Is someone unnecessary rich enough willing to try this? For the sake of ... TPU community.

 

[cinemaware]

I feel for you, at this point you have to wait for another model to show up and get that.
Store was happy to refund you as they can sell it as open box pretty fast.
TBH when you said you taking it to store I thought you gonna get a replacement and try your luck with another one.
Do you use AC in your room?
RMA is never an option IMO.

I would take a replacement, but they just didn't have one. The alternative was keeping the card and then maybe trying to send it off to Gigabyte for repairs in some later date or just selling it "as is". In any case I’d have to mention it to a prospective buyer and that would bring the value down.

I use AC but not all the time. And even yesterday, when natural ambient was ~18-20C (cold spell here) the VRAM temps in CP2077 were 96C with default fan.

What 9070xt did you have?

Gigabyte Gaming OC. I think the guys at the service centre were saying that it happens with other models too, though.

 

[Jtuck9]

Gigabyte Gaming OC. I think the guys at the service centre were saying that it happens with other models too, though.

Interesting. I remember that report that leaked out before release mentioning the high VRAM temperature on a test card, although I'm not sure if the specific model was ever confirmed?! Doesn't surprise me regarding the Gigabyte card though as it's the highest performing card, no?

As an aside, I find the temperatures I've seen people mention when running their 9800X3D chips on coolers like the PA 120 mini kind of off putting, yet still within tolerance. (Granted these are SFF cases and probably (mostly) stress test temperatures...) Don't AMD themselves recommend liquid cooling 9950X's?

 

[cinemaware]

Doesn't surprise me regarding the Gigabyte card though as it's the highest performing card, no?

Gaming OC is actually the basic, MSRP-level model. Still, I had the same model 3070 and 4070, and never a problem (not that it proves or disproves anything).

 

[rusty caterpillar]

Gaming OC is actually the basic, MSRP-level model. Still, I had the same model 3070 and 4070, and never a problem (not that it proves or disproves anything).

I don't like PCB designs of Giglebit but, that is me. I would try other AIB later on when the price gets lower.

I don't even know if Gaming OC has thermal pads for back plate or not. The weight of the card is not specified on their website, we only have the thickness 56 MM, thinner than a 9070 PC HH card.

Vex made a review of the Gaming OC but he carefully left out VRAM temps and I'm not wondering why:

 

[Jtuck9]

Gaming OC is actually the basic, MSRP-level model. Still, I had the same model 3070 and 4070, and never a problem (not that it proves or disproves anything).

Ahh, I did wonder as I typed...

I don't like PCB designs of Giglebit but, that is me. I would try other AIB later on when the price gets lower.

Yea, I thought Sapphire generally had a decent reputation, 12V-2*6 connector aside for the Nitro +. I read that Powercolor were the go to cards last generation, not sure how much that will vary over time?!

 

[rusty caterpillar]

Ahh, I did wonder as I typed...


Yea, I thought Sapphire generally had a decent reputation, 12V-2*6 connector aside for the Nitro +. I read that Powercolor were the go to cards last generation, not sure how much that will vary over time?!

On 9000 you'll see many similarities between Sapphire and Asrock and PC in the cooler design.



So we are left with other factors to consider:
1. weigh of the heatsink
2 quality of fins soldering to the heat pipes which is influencing heat transfer - is very hard to establish that
3. T pads on back plates or the lack of it
4. Our own setups impact - how far is the case side panel from the card(Hot air recirculating), rear case fan efficiency etc

 

[Jtuck9]

On 9000 you'll see many similarities between Sapphire and Asrock and PC in the cooler design.

View attachment 393103

So we are left with other factors to consider:
1. weigh of the heatsink
2 quality of fins soldering to the heat pipes which is influencing heat transfer - is very hard to establish that
3. T pads on back plates or the lack of it
4. Our own setups impact - how far is the case side panel from the card(Hot air recirculating), rear case fan efficiency etc

Yea, it is interesting. And whether that even justifies an AIB is something to consider.

Reminds me of something I read in that book "Stasiland" once...

 

[SirKeldon]

@cinemaware your story is a sad one, I'm really sorry to read this. I mean, we all saw somewhat high temps, but yours were near the limit straight out of the box, and even with a strong fan curve, they stayed high, definitely not OK. I won't step in the refund process since I don't know Taiwan's consumer laws, but here in Spain, if you bought the card separately, you're entitled to an RMA replacement of that item if is found defective. However, if you're not experiencing any glitches or malfunction, it would probably be sent back and deemed normal (maybe thermal conditions in their labs are different, who knows). Even so, if the law allows it, I'd ask for a revision of the card, especially if you're still under warranty. You were very close to the thermal throttle point, and that's definitely not ideal.

@Event Horizon can you share your curve as well? I think it would be helpful to have some examples around. You're right that keeping the fans always spinning will lead to more wear and tear, but it wasn't that long ago when GPU fans were always on by default, before 0-RPM modes even existed. Also, it's worth noting that fans constantly stopping and starting, even with built-in safeguards, can also cause wear over time. Depending on the scenario, that might happen anyway. Fan technology has improved a lot though, fans are usually rated for around 50K hours, modern ones probably even more, so even if they degrade faster, they're still much cheaper to replace than VRAM chips, worth the shot IMHO.

@LittleBro you're right, I'm also concerned about future scenarios, especially now that summer is coming in the northern hemisphere. On air-cooled GPUs, ambient temps play one of the biggest roles, if not the biggest sometimes. I haven't tried repasting yet, and probably won't unless it's absolutely necessary and my warranty isn't at risk. That said, I guess my case is one of the less worrying ones since in-game temps stay in the high 70s, not the high 80s or low 90s like many others.

@Jtuck9 the case you mentioned is definitely concerning. Let's hope it remains anecdotal and doesn't become the norm. In the end, cinemaware's card, despite running hot, still performed fine after nearly a month under those conditions.
---
As a final note, who knows, maybe a future driver update helps scale temps better. That said, even with the fans at 700 RPM (15%) and the system totally idle, my VRAM temperature wouldn't go below 48°C. I know it's reading from a junction point, but still, I think it's worth keeping an eye on things without panicking. Better safe than sorry, so tune up your curves if you feel the need

 

[spacer00ster]

What is your KW/h cost if I may ask?

Hey.

1st block 250kWh - $0.1719/kWh
2nd block 250-700kWh - $0.307/kWh
Tail block 700kWh+ - $0.4835/kWh

My bills are normally in the $300-$400 range. And I think this month or next month the rates will increase.

 

[spacer00ster]

Has anyone tried replacing thermal pads yet?

Is someone unnecessary rich enough willing to try this? For the sake of ... TPU community.

LOL not rich but IF (and I hope not) the VRAM temps of my incoming Gigabyte XT Gaming OC card are high then I'll definitely get some thermal pads or PTM7950. It's very difficult here in Bermuda to purchase from a US store, ship to a freight forwarder, pay forwarder shipping rates then pay customs duties/fees etc., to then end up returning the item - you don't get back all those shipping/duty fees. I can live with the total cost of the card I got at the standard Gigabyte price of $730 +extra costs, but getting any card at scalped prices just exponentially increases the other costs. Most times, everything you buy in a store here is priced double.

 

[rusty caterpillar]

I think all 9070 and 9070XT have the SK Hynix GDDR6 VRAM, don't remember any GPU tear up having something else.

Couldn't find any spec sheet from the manufacturer.

However found this: Thermal Resistance of GDDR6 and 7.
GDDR 7 is only on Nvidia. Just to clarify.




Even maybe 74 % is just marketing, is very little we can do to compare and find the real data for Thermal Resistance GDDR6 vs GDDR7.
From the that marketing statement we can draw a conclusion it might be a gross difference.

I don't see other solution than using high efficiency thermal pads/putty and cool the card from the other side as well.
Fan curves will help as much as you can bare the noise, ATM ! the older the GPU will get the less the fan curve will make a dent.