AMD RX 7000 series GPU Owners' Club

[kapone32]

Thanks a lot for sharing your observations! I'm really interested in the power consumption figures. Your data shows a minimum TBP of 25w, while in TPU's testing idle power was 12-16w. Was something running in the background in your case, an open web browser maybe?

Also, could you check with YouTube at 720p to 4K? AMD claims to have improved video playback consumption in their recent drivers.

I always have Steam, GOG and Epic running. Of course there are the Windows background programs like Weather and other News media pop ups. I also have Skype running as well.

Once I adjusted the power limits I do see that HWinfo says a max board draw of 530W. When Gaming I only see a max of 364 but when doing benchmarks spikes to 410 W can be seen. One issue is that the AMD stress test is broken and will push the card for 10 seconds (60 second test) and report idle for the next 50. I do have the overlay applied from AMD software and see a board power of 33 watts as we speak.

I did try Youtube in Watching WEC 8 hrs of Bahrain at 1080P and a 4K Hardware Canucks CES MSI tour (No AMD laptops ). The power draw was around 65 watts.. I do believe that the card has great potential for OC with a sweet Undervolt but you need to tame the card. Maybe that is why the new blocks look like square reservoirs instead of channelled acrylic and the heatsinks are so huge on AIB cards. The fact that there is 1 driver is good in my opinion. It means all the work that Sony, Microsoft and AMD are doing on hardware and software enhancements for said hardware will be realized for this card and it stands to reason that the next Console will be AM5 with RDNA whatever.

 

[kurosagi01]

Thought i'll share my hwinfo stats after just about an hour of COD MW2 with my undervolt settings.
The max recorded temps seems perfectly up to spec? I can't find any information what the spec is for the temperature range.

 

[phanbuey]

Thought i'll share my hwinfo stats after just about an hour of COD MW2 with my undervolt settings.
The max recorded temps seems perfectly up to spec? I can't find any information what the spec is for the temperature range.
View attachment 278001

that's downright cold. The spec is up to 110C before throttle.

 

[kurosagi01]

Testing out target boost clock at 2400mhz at the moment now, using NFS Unbound since that game seems to like crashing the PC with any of my undervolt settings.
So it crashed after 2 races at 1050mv(Seems like this is my cap), I've increased the voltage to 1055mv which seems Ok after 2 races but I wasn't too happy about the frame rate as I was getting 80-90fps while racing.
I've increased it to 1060mv and I was getting 100+fps on average while racing.
I'm using the same fan speed and power limit at 15% like I am with my 1075mv config at 2760mhz.

 

[AusWolf]

Wonder how many people who've spent the past week deriding derbauer as a FUD spreader are going to eat their words okay in all seriousness, it was as hypothesized, the lack of fluid in certain batches of cooler.

It is a minor performance penalty when throttling @ 110C, but it's not a placebo - yes he should have published clocks, but even in derbauer's own temp graphs it's obvious. Visibly rising temp then abruptly flatlining at 110C is not normal unthrottled behaviour for edge temp or hotspot temp on any GPU, period. Also 100% fan speed earrape

I'm just glad that [according to all sources so far] I won't have to worry about it in the XT MBA. ETA got moved up a week, good thing since I sold my TUF today and am now back to my old GPU.

Hopefully this drama is over and they can now focus on driver performance and addressing the multi monitor idle.

@ 3:10

Like I said, I don't want to fight any battle on any side... I just want to know how much performance is lost due to these thermal issues.

I know the internet is full of trolls, but is it so hard to believe that some of us just genuinely want to learn?

Der8auer's video didn't give me the answers I was looking for, that's why I asked here.

 

[HTC]

Stupid question: have any of the proud new owners of XTX cards tried changing the cable to see if there are any differences in your temps, as described in the video posted in the 1st page? I'm referring to ANY such card: even those that DON'T seem to have this issue.

Who knows: maybe swapping the cable MIGHT result in lower hot spot temps AND a small boost in performance, as opposed to a HUGE DROP in those hot spot temps and zero throttling.

Just thought of another stupid question: does swapping the cable have ANY EFFECT @ ALL in idle power consumption? Just wondering.

 

[jesdals]

Any lucky 7900 XTX owners that use Eyefinity setups? I would like to hear about experience with 3x displayport on the reference with a dongle/usb c combo? And how about gaming in my favorit resolution 7680*1440 is it great or dosnt the 24gb memory scale well with the multi monitor performance?

 

[kapone32]

Stupid question: have any of the proud new owners of XTX cards tried changing the cable to see if there are any differences in your temps, as described in the video posted in the 1st page? I'm referring to ANY such card: even those that DON'T seem to have this issue.

Who knows: maybe swapping the cable MIGHT result in lower hot spot temps AND a small boost in performance, as opposed to a HUGE DROP in those hot spot temps and zero throttling.

Just thought of another stupid question: does swapping the cable have ANY EFFECT @ ALL in idle power consumption? Just wondering.

I tried a DP 1.2 cable and while I did not notice an increase in GPU temps that cable felt really warm after 3 runs of 3D Mark Timespy. When I changed it to a DP 1.4 cable the issue went away. I only used it for 15 minutes though so I can't say definitively that the cable is the cause for the Hotspot temps. As I have said before a quirk that I have found is my Memory is higher than my Hotspot.

Testing out target boost clock at 2400mhz at the moment now, using NFS Unbound since that game seems to like crashing the PC with any of my undervolt settings.
So it crashed after 2 races at 1050mv(Seems like this is my cap), I've increased the voltage to 1055mv which seems Ok after 2 races but I wasn't too happy about the frame rate as I was getting 80-90fps while racing.
I've increased it to 1060mv and I was getting 100+fps on average while racing.
I'm using the same fan speed and power limit at 15% like I am with my 1075mv config at 2760mhz.
View attachment 278004

How is TDC limit 448 A when my 7900XTX is at 290? Can we get some more to compare?

 

[kurosagi01]

I tried a DP 1.2 cable and while I did not notice an increase in GPU temps that cable felt really warm after 3 runs of 3D Mark Timespy. When I changed it to a DP 1.4 cable the issue went away. I only used it for 15 minutes though so I can't say definitively that the cable is the cause for the Hotspot temps. As I have said before a quirk that I have found is my Memory is higher than my Hotspot.


How is TDC limit 448 A when my 7900XTX is at 290? Can we get some more to compare?

I'm not sure why the TDC limit is 448A, even on the default config it's reporting 448A on Hwinfo.

 

[kapone32]

I'm not sure why the TDC limit is 448A, even on the default config it's reporting 448A on Hwinfo.

That seems very excessive but your card seems to be working great though. How are you enjoying yours otherwise though?

 

[kurosagi01]

That seems very excessive but your card seems to be working great though. How are you enjoying yours otherwise though?

Getting a lot higher frames than I was with my 3080, that is without DLSS as well which is great. The fan noise of the reference cooler isn't that bad either with my current undervolt, can barely hear it over the other fans installed.
Some games still need optimising though since its a new card and drivers. NFS Unbound is still a bit choppy every now and then, which I think Criterion needs to patch the game as well.
Can't really complain when I have 2 high-end rigs in the house now, which my partner will be able to enjoy on her desk setup, but for now its the 4K gaming machine in the living room.

 

[kapone32]

Getting a lot higher frames than I was with my 3080, that is without DLSS as well which is great. The fan noise of the reference cooler isn't that bad either with my current undervolt, can barely hear it over the other fans installed.
Some games still need optimising though since its a new card and drivers. NFS Unbound is still a bit choppy every now and then, which I think Criterion needs to patch the game as well.
Can't really complain when I have 2 high-end rigs in the house now, which my partner will be able to enjoy on her desk setup, but for now its the 4K gaming machine in the living room.

F me 4K. This card laughs at 4K. I have been running some benchmarks and Strange Brigade as an example was like 267 FPS. The Division 2 will sit at 170 FPS all day and TWWH3 is so butter smooth in battles it is addicting. Redout does not give me much over the 6800XT and Just Cause 4 is higher but only by about 10 to 15 FPS but I agree wholeheartedly with your statement, let Drivers come.

 

[kurosagi01]

F me 4K. This card laughs at 4K. I have been running some benchmarks and Strange Brigade as an example was like 267 FPS. The Division 2 will sit at 170 FPS all day and TWWH3 is so butter smooth in battles it is addicting. Redout does not give me much over the 6800XT and Just Cause 4 is higher but only by about 10 to 15 FPS but I agree wholeheartedly with your statement, let Drivers come.

I would like to try the 7900XT in 4K but I don't think i'll be upgrading to 4K monitor anytime soon.
Neither do I fancy moving my PC downstairs,while hers is being used as the 4K machine haha.
I've just swapped my Iiyama with an AOC monitor as the Iiyama is going to be sent for warranty work, feels good to be back in curved VA panel with 1000R and same resolution but with 165hz range.

 

[MrPerforations]

Hello Club,
i got me the AMD™ Sapphire™ RX7900XTX Nitro+ card to replace my AMD™ RX Vega 56 Crossfire system.
no been to impressed with the benchmarks i have run, but i found i didn't have any of the 4k benchmarks saved for the AMD™ RX Vega 56 Crossfire.
i just did the superposition 1080p extreme bench for the thread.
i got 15000 for the first run, then i noticed resizable bar was disabled and went and enabled that as well as turned on POB in bios, i gain a 17000 score.
it benched on par with a Nvidia™ 4080, which is right.
i have a water block coming tomorrow, the Alphacool™ block.
i will let you know about the temps and then i can play some OC.

 

[AusWolf]

Hello Club,
i got me the AMD™ Sapphire™ RX7900XTX Nitro+ card to replace my AMD™ RX Vega 56 Crossfire system.
no been to impressed with the benchmarks i have run, but i found i didn't have any of the 4k benchmarks saved for the AMD™ RX Vega 56 Crossfire.
i just did the superposition 1080p extreme bench for the thread.
i got 15000 for the first run, then i noticed resizable bar was disabled and went and enabled that as well as turned on POB in bios, i gain a 17000 score.
it benched on par with a Nvidia™ 4080, which is right.
i have a water block coming tomorrow, the Alphacool™ block.
i will let you know about the temps and then i can play some OC.

Congrats!

But why did you buy a Nitro+ if you were planning on watercooling it? You could have saved a few bucks with an MBA card.

 

[MrPerforations]

thanks,
the 3x8 pins board, and last time i got the best one of all the card of that type, it was a good idea as it worked out to be the best performance and cheap. at the time it was better than a 2080 and half the price.
also, im not sure if i want to keep this atm.
my regret i have is not waiting for the ASRock™ cards to come in stock or the MSI™ ones to be released, i bet they beat this one up.

been running the AMD™ adrenaline overlay to monitor the fps , ect, ect.
i noticed that my CPU is utilizing only about 5% at 4k , is that right please?
the GPU is at 100%
gotta be able to save money if it is right for sure.

 

[tabascosauz]

my regret i have is not waiting for the ASRock™ cards to come in stock or the MSI™ ones to be released, i bet they beat this one up.

What makes you say that though? So far AIB design Sapphire, XFX and ASRock (Taichi and Aqua) all look like they use almost-reference PCBs that are only slightly extended to accommodate 3 x 8-pin. The variance in OC headroom looks like it all comes from the GPU itself at this time.

 

[kapone32]

thanks,
the 3x8 pins board, and last time i got the best one of all the card of that type, it was a good idea as it worked out to be the best performance and cheap. at the time it was better than a 2080 and half the price.
also, im not sure if i want to keep this atm.
my regret i have is not waiting for the ASRock™ cards to come in stock or the MSI™ ones to be released, i bet they beat this one up.

The Pulse is actually more expensive than the Nitro right now but you got basically the best AMD GPU vendor. Sapphire is older than EVGA and they only make GPUs. What you get is usually one of the best binned GPUs and super nice build quality. You will see when you go to take your card apart how much engineering has gone into it vs the competition.

 

[MrPerforations]

been waiting for Buildzoids™ review of the boards.
i don't know electronics but i think the Sapphire™ has ceramics on it, but the ASRock™ has 22 phases vs 20 phase, but like i said i dont know.

 

[kapone32]

What makes you say that though? So far AIB design Sapphire, XFX and ASRock (Taichi and Aqua) all look like they use almost-reference PCBs that are only slightly extended to accommodate 3 x 8-pin. The variance in OC headroom looks like it all comes from the GPU itself at this time.

The 3x8 pin is the only reason I got an AIB card. I wanted to get the Aqua but my friend at the PC store told me that there were actually only 5 Aqua units allocated for Canada (his store).

 

[MrPerforations]

yer, i planned on the ASRock™ RX7900XTX Taichi and a water block and make the same for less.
thus , i might return the Sapphire™ yet.

 

[kapone32]

yer, i planned on the ASRock™ RX7900XTX Taichi and a water block and make the same for less.
thus , i might return the Sapphire™ yet.

Don't As Rock is the newest GPU vendor in the space and they make nice cards but the only real advantage they have is they only really compete with Gigabyte when it comes to pricing. The only card that is arguably better than your Nitro would bw the Toxic (Not released), Strix, TRIO OC and Aorus but if you are putting a block on that it becomes moot. Here is something you have on the Nitro that other cards don't have. Your card has an ARGB Header and a 4 Pin PWM header on the board. The ARGB is not what I mean. You will be able to connect your Rad fans to your GPU and still be able to use AMD software to adjust the fan curve. That is something to have vs 2 more VRM phases. The quality of Sapphire's components should also nto be ignored.

 

[tabascosauz]

The quality of Sapphire's components should also nto be ignored.

That is something I have always assumed too, but the only meaningfully unique PCB I've seen so far is the TUF. Everything else largely builds on the same reference design; even the TUF uses the exact same MPS parts in the VRM and same Vcore phase count. Sapphire in particular seems to be using the same cap supplier as MBA, but I guess that's not surprising since there were so many rumors that Sapphire manufactures AMD's reference boards. The only ASRock pictures I see for Taichi and Aqua are silhouettes, but the size of the PCB and layout of major components looks identical.

Of course, maybe it's a vote of confidence in AMD's reference design. If all the AIBs feel confident enough in it even with 3 x 8-pin, that's a good sign considering the 450W the Nitro was pulling on what is essentially a reference PCB.

Anyway, I agree, it's a bit of a stretch to expect ASRock or MSI to come up with anything radically better than the rest. Instead of worrying about PCB components it's a bit more concerning to me the variance in GPU silicon quality and voltage requirements (XTX and XT undervolting results on reddit seem to be wildly all over the place).

Nitro in the middle.

I'm not sure whether the fan stop actually kicks in or not for me, since I don't really look and all the fans in general kind of overpower the GPU noise for me.
As for attempting to see how low I can undervolt the vcore with stock target boost clock, i've not really tried.
The way I've set the target core clock to 2760mhz,is based on the auto value entered when i've adjusted the power limit to 15%(the core clock value changes automatically changed to 2800mhz when I increased the power limit).
So I did 2400(default boost clock)*1.15(15% power limit increase), which gives 2760mhz, so i've set the target as that instead.
I have only gone by observation for the stability,on COD it was perfectly fine on 1050mv with 15% power limit and the auto core clock value, but on NFS Unbound it hard crashed my PC, so i've adjusted it to 1075mv and it has been rock solid so far.
You may be able to get much lower undervolt on the vcore if you are targeting the 2400mhz boost clock with 15% power limit increase, maybe even a 11-15c junction temp drop too with custom fan profile.
I know a lot of the tech tubers like to set fan speed at 100%, but nobody likes to listen to a jet plane all the time.

Ok back to the fan stop issue as I can comment on it now.......it's frustrating as it's unnecessarily conservative. I still haven't managed to figure out just what the fan stop criteria is. It's clearly not GPU edge temp alone.

Because of how boosty RDNA is and the video playback power consumption it's exceedingly difficult to get the fans to actually stop. The video power "bug" has been addressed as the review samples were pulling like 80W on Youtube, but 40-50W is not what I'd call stellar compared to ~20W on Ampere since it affects fan speed. It doesn't even heat up the GPU at all, but it keeps it barely above the fan stop threshold, just enough to spin the fans for no good reason and piss me off.

I tried making a custom fan profile in Radeon Software but that tuning section is buggy as hell and seems to set whatever it wants after a reboot.

I can have a browser open and still be at 20-25W and have fan stop.............as long as none of those tabs are Youtube. Even as a background tab, as long as the browser window itself is in focus, power consumption shoots up to 50W.

Having been away from Radeon for so long I'm not sure what to expect in terms AMD fixing this behaviour. From what I've read, it is probably a bad idea to have any expectations at all.

 

[AusWolf]

Ok back to the fan stop issue as I can comment on it now.......it's frustrating as it's unnecessarily conservative. I still haven't managed to figure out just what the fan stop criteria is. It's clearly not GPU edge temp alone.

Because of how boosty RDNA is and the video playback power consumption it's exceedingly difficult to get the fans to actually stop. The video power "bug" has been addressed as the review samples were pulling like 80W on Youtube, but 40-50W is not what I'd call stellar compared to ~20W on Ampere since it affects fan speed. It doesn't even heat up the GPU at all, but it keeps it barely above the fan stop threshold, just enough to spin the fans for no good reason and piss me off.

I tried making a custom fan profile in Radeon Software but that tuning section is buggy as hell and seems to set whatever it wants after a reboot.

I can have a browser open and still be at 20-25W and have fan stop.............as long as none of those tabs are Youtube. Even as a background tab, as long as the browser window itself is in focus, power consumption shoots up to 50W.

Having been away from Radeon for so long I'm not sure what to expect in terms AMD fixing this behaviour. From what I've read, it is probably a bad idea to have any expectations at all.

If your card is anything like the RDNA 1/2 cards that I've had, then the fan stop works based on GPU temperature. The cutoff point is usually around 50 °C which I suppose, can easily be reached with a multi-monitor setup watching Youtube with the power issues RDNA 3 has with current drivers.

 

[tabascosauz]

If your card is anything like the RDNA 1/2 cards that I've had, then the fan stop works based on GPU temperature. The cutoff point is usually around 50 °C which I suppose, can easily be reached with a multi-monitor setup watching Youtube with the power issues RDNA 3 has with current drivers.

It's certainly not 50C. Looks like 40C, it might be after 60 seconds spent below 40C or something like that

Both CODs were okay but Genshin and Sniper Elite 5 are microstuttering like mad. Too short to be picked up by Latencymon but obvious to anyone with a pair of functioning eyes. Been hitting buttons in Radeon Software to see if they do anything, reinstall didn't do anything either. Full DDU'd before I started so not sure what more I can do aside from clean install and literally reinstall all my games

We needed the new drivers yesterday, AMD