Question 6900XT Red devil acting weird

Kuplungúr

Prominent
Apr 19, 2022
18
1
515
Is my 6900XT faulty? Tomorrow I think I will bring it back to the vendor, but here's what's up:
  • I bought this to increase my 4K framerates while playing on tv (changing from 2080S but DDUd everything, all clean, all fine). However from the very first time I installed it it is very strange to borderline not working. First of all it is loud. Not "fan noise loud" but "im processing" noise loud. Very. Loud.
  • then there are the insane frame drops/utilization changes, essentially making in RDR2 fps counts 120 while looking at one direction, dropping to 50 (while GPU usage also dropping to 50...) while taking 180 turn. In some occasions the monitor switched off and on again, Adrenaline showing that usage and power consumption both hitting 0 for a short time
  • In Valorant (which works on a smartfridge basically) there were heavy framerate issues and when I used Adrenaline's anti-lag and other features it either crashed, lagged or my pc froze (multiple times).
  • Fans switching on and off in very strange patterns while GPU is hot. During RDR2 gpu was around 80 and crosslink (or connector, sry I'm using Adrenaline in Hungarian) was 100°C. Hardly normal I suppose.
(pc is an Asus Z390-H gaming with 32GB G.skill trident, i7-9700k, some M.2. and an SSD, Corsair TX750M)

I tested it out on my old pc, an i7-4790+16GB DDR3+750W Cooler Master. DDUd there as well and tested Witcher 3 and Valorant. Well, neither used the CPU more than 40-50% but in Valorant we were hitting anything between 50-200fps total randomly, heavy stutter which was basically game breaking in a fast fps, you can imagine. In W3 it was not only awful (something was odd, I not remember it looking like that) but FPS was between 40 and 50. In ultra, I give them that but in 1080p. I remember that I played W3 on that pc with GTX970 in 50-60FPS in high and when I switched to the 2080+9700K it was 130... So anything under 100 was unacceptable, but then again GPU usage for the 6900XT was 30% at most... Why, I cannot tell.

So all in all:
  • heavy fps drops, sometimes even 100% drop from one second to the other
  • varying GPU usage seemingly at random
  • freezing the pc while turning certain features on
  • tested in two PCs

Any ideas other than this is a faulty Red Devil? All I know is that this particular card should be even better than the rest of the 6900XTs and despite having very little experience with AMD GPUs in the past 10 or so years, they surely can't be nVidias fierce competition featuring cards and fps numbers like this. I mean this is something beyond stuff like "you should have clean installed a new windows" or "this is a new card in an older system". 6900XT RD ultimate, 1000 USD card at least. Cannot even make Valorant run steadily.
 
Have you seen this post regarding Adrenaline? https://community.amd.com/t5/gaming-discussions/the-stutter-is-real-on-the-6900xt/m-p/527747#M21366

See if you can rule that out. If performance immediately improves without it then it's something in Adrenaline i guess you could investigate.

Also see here regarding overclock which op in this thread helped his situation underclocking.

 
None of my stuff is overcloked and despite that this 6900XT requires 3x8 pin connectors, the total wattage never went above 300Ws which can be easily taken. Don't tell me that a 3080ti (working fine with a TX750M) eats that much less power. And until/unless GPU usage is closing to 100% this couldn't be the issue for stutter in Valorant, could it? How much power does it want to consume to run a game that runs on mediocre laptops as well? 😀
I will check the forums, thank you!
Might worth mentioning that we ran a 3D mark spy on the secondary rig and it utilzed the GPU well and without problems we got similar results to everyone else using similar setups. (close to 19000 in graphics)
 
None of my stuff is overcloked and despite that this 6900XT requires 3x8 pin connectors, the total wattage never went above 300Ws which can be easily taken. Don't tell me that a 3080ti (working fine with a TX750M) eats that much less power. And until/unless GPU usage is closing to 100% this couldn't be the issue for stutter in Valorant, could it? How much power does it want to consume to run a game that runs on mediocre laptops as well? 😀
I will check the forums, thank you!
Might worth mentioning that we ran a 3D mark spy on the secondary rig and it utilzed the GPU well and without problems we got similar results to everyone else using similar setups. (close to 19000 in graphics)
19000 on timespy with 6900xt? mine 6800 non xt has 18k in graphics (2x8pin 750w psu)
18328 graphic score to be specific

looking at 3d mark for 6900xt, you should be getting around 25k with OC model
 
19000 on timespy with 6900xt? mine 6800 non xt has 18k in graphics (2x8pin 750w psu)
18328 graphic score to be specific

looking at 3d mark for 6900xt, you should be getting around 25k with OC model
If you check here, nobody had vastly different results to mine. (i'm in line 6 btw at the moment of posting)
 
oh then you are CPU bottlenecked, that explains wild GPU utilisation fluctuations, that doesnt explain black screen tho
Ok, let's separate the two things now, because I'm might be confusing you:
  • the W3 and 3D mark tests were done on my older pc. No black screen there but stutter and utilization fluctuation is not explained by CPU bottleneck as cpu util was also low there. Also, the same CPU with GTX970 had comparable results to the 6900XT.
  • The black screen was on my main, featuring the 9700K processor in RDR2. That game also does not use more than 25-40% of the CPU in any case.
  • 3D markt during GPU testing also lets the CPU rest, it is probably the build, the motherboard, the older ram, etc which lead to the lower results. Still no answer to why any game has worse results with the 6900xt than they had with an older gpu.
 
your PC platform is a bit old tbh....might be wort tryin to disable message signaled interrupts, amd and nvidia is enabling it lately to reduce driver latency, could be that your CPU is not supporting it right

its done in windows registry
first open device manager to get hardware id of your GPU
dIXRD4v.png

mine is VEN_1002&DEV_73BF

then open regedit
go to this adress: HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Enum\PCI
in there open your gpu hardware id, go deeper until you see device parameters, then open interrupt management and messagesignaledinterruptproperties, here change MSIsupported to 0 then close regedit and reboot
R5VELU9.png


reinstalling gpu drivers will set it back to 1 unless you use some very old drivers
 
your PC platform is a bit old tbh....might be wort tryin to disable message signaled interrupts, amd and nvidia is enabling it lately to reduce driver latency, could be that your CPU is not supporting it right

its done in windows registry
first open device manager to get hardware id of your GPU
dIXRD4v.png

mine is VEN_1002&DEV_73BF

then open regedit
go to this adress: HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Enum\PCI
in there open your gpu hardware id, go deeper until you see device parameters, then open interrupt management and messagesignaledinterruptproperties, here change MSIsupported to 0 then close regedit and reboot
R5VELU9.png


reinstalling gpu drivers will set it back to 1 unless you use some very old drivers
Okaaaay, but I guess this does not explain the issues on my "not very dated" system with the 9700k and the Z390. Is that the PSU? Or the hardware graphics scheduling or whatever it was up above? I'm clueless. Why is that not working like every single upgrade I ever had: plug in, change drivers if necessary and enjoy? Sorry if I'm a douche here but my weekend was ruined by this and I'm in constant mental anguish...
 
Okaaaay, but I guess this does not explain the issues on my "not very dated" system with the 9700k and the Z390. Is that the PSU? Or the hardware graphics scheduling or whatever it was up above? I'm clueless. Why is that not working like every single upgrade I ever had: plug in, change drivers if necessary and enjoy? Sorry if I'm a douche here but my weekend was ruined by this and I'm in constant mental anguish...
6900xt has transient power spikes up to 500watts, thats reference not OC model (2x8pin)
03-Peak-Power.png
 
adrenaline has 1000ms default setting, so ye those wont be visible there
Ok just one more thing: these transient spikes, I guess most other GPUs also have these. How can we tell that there's just no way that 2x8 pin coming from the PSU (3x8 in this case) is enough or not? Is there a rule of thumb for how much transient wattage the PSU can provide given its on paper wattage? Like if you have a 750W don't expect anything above 400 or similar? Also, what happens if these transient spike requirements are not met by the PSU? Can it generate errors like this?
I'm asking this because my friend owns a very similar setup with the same TX750M and he has a 3080ti (2x8 pin) and has no problem whatsoever.
 
A rule of thumb for transient spikes is, if you are lets say 400w usage, multiply by 2 and thats your transient power spike. Also, even a 750w power supply can do 800-900w for a short time before tripping if it isnt a <Mod Edit> quality.
 
Last edited by a moderator: