RDNA2 RX6000 Series Owners Thread, Tests, Mods, BIOS & Tweaks !

So I finally switched to the ''evil'' side of the force and ordered the RX 6900 XT, the Asus TUF Gaming OC model (I hope it's not the worst from non reference designs lol)
2080 Ti was not enough for 4K anymore, especially with Reshade and graphics mods..

Also my mobo from MSI (Z390) got BIOS update for BAR/SAM support, I'm excited to test it when my new GPU arrives in next 2-3 days

 

Correction the good side of the force.

 

Neither is the "good side"....good cop...bad cop.
I'm back on topic now

 

It is more about luck with chip. And how much you can undervolt it. Reference design is already good and AIBs cards decrease noise levels at same power draw.

 

https://www.3dmark.com/compare/fs/25135397/fs/25167284

The power of RDNA2 architecture! I was able to beat my friends overclocked RTX 3080 EVGA FTW3 Ultra on their 450 watt custom bios by 38%! This means in theory I beat a stock RTX 3090 by around 28%. Absolute lunacy guys. These cards on juice got some power!

 

Whats with the BIOS mod ?

I can't get more than 10% more perf from my 6800 with just driver OC .

 

Greeting to the community.
I have been tuning my 6800XT to find more performance out of it.

So installed the new driver 21.3.1, and just raised the values in the MPT a little.
Power Limit GPU from 284W to 289W and TDC limit GFX from 340A to 350A.
The frequency and other settings are still the same as I posted before at TPU:
https://www.techpowerup.com/forums/threads/the-rx-6000-series-owners-club.276164/post-4480167

This is how my score looks now.

I could not believe my eyes.
The Time Spy score has gone up from 19959 to 20401.
The Port Royal went from 10193 to 10336.

The Sapphire 6900XT Toxic that I tested produced 20500 max by pushing 368W core and 384A TDC and by default only 20200.

There is an orange exclamation mark in the score as driver is new and not verified by 3DMark.




Note that my memory before was not stable above 2100MHz. I raised TDC Limit SoC from 55A to 63A and now I can run memory up to 2120MHz without having a score regression.
Higher SoC TDC is helping to raise the memory speed a little.

 

I didn't find anyone reporting that they repasted their card with the search. I have an MSI 6800 XT Gaming X Trio and was wondering if people here think it would be worth a repaste with IC Diamond to try and get the hotspot temperatures down as they are usually a solid 30C higher than the edge temperature. I have a case with good airflow so that should not be the problem. I see in reviews this is normal for the card under heavy loads and I also see the junk paste that MSI puts on their cards (all cards seem to come with junk paste) in the teardowns which got me thinking maybe some IC Diamond would help out a bit. I also know that MSI had no issue with me repasting my 1080 Ti Gaming X Trio with keeping the warranty and it helped a bit with temperatures versus stock paste so I am pretty sure they are pretty relaxed about people repasting their cards. I started to think about this because the MSI 6800 XT has a lower power limit than other AIB cards and therefore less OC potential. I tried using the more power tool and although it worked wonders it also shot up that junction/hotspot way too much for my liking.

I seem to hard crash to a black screen and need to hard reboot my computer when overclocking my VRAM to 2100 in Hunt Showdown with resizable bar off and when resizable bar on it crashes a lot of games at 2100 so I have decided to just keep it at default but would raising the SoC voltage in MPT help with this? It seems that everyone is able to handle 2100 on the RAM; has anyone tested these speeds specifically with Hunt Showdown? Should I also try my hand at undervolting the card via MPT rather than just wattman? I have been reading through this thread and a lot of people seem to be getting good results.

Other than the black screen issues and not being able to play forza gorizon 4 at 4k on my samsung Q80T for some reason (all other games play fine at 4k on the TV) I am absolutely loving this beast of a card and can't wait for AMD's DLSS competitor to come out so we can get some solid raytracing going on with this beast. With fairly luke warm OCs I am able to either meet or beat my friends on Time Spy with their 3080s which is a really nice feeling. I haven't been with AMD since the spectacular HD 7950 (800mhz OC on the core was INSANE) and I am really happy to be back with AMD with this solid top end generation! Any tips and tricks would be much appreciated as I have zero experience with AMD's control panel and really any of the features.

Thanks everyone for your input.

 

"BuT buT RtX aNd dLsS and nVeNcOdEr nViDiA gOt FeAtUrEs MaN aNd AmD dRiVeR sUxXxXx"

... yea, can't wait for them 6800XT's to get in stock, eventually !
I will patiently wait for the mining craze to wind down...

 

First of all - test for stability without any OC and uV. If the card passed your stability tests then move on. If not the culprit is somewhere else.
uV the card slightly in Wattman, and add a few Amps in the TDC Limit of SoC. E.g. I have added 3 to increase stability of VRAM. I would not advise to uV your card in MPT.... Wattman should be enough and juuuuust fine.

I would first try everything else before disassembly. If you do this step prepare well in advance - check for YT disassembly videos, buy also new thermal pads of correct thicknesses and sizes etc. But I would generally not advise this step if you are not into electronics and REALLY sure the hotspot is the root cause.... etc..

 

Noice!

I feel really kind of stupid for canceling my 6800 order for 780 € as "expensive" back in December.

Don't seem so expensive now... well I hope some day prices return to normal.
At least with Radeon 7000 series, lol.

For a fleeting moment I eyed a 6900 XT Asus TUF (in stock) but at 1800+ € just... can't.

 

So I am very confident in my skills so I just went ahead and opened up the card which was super easy. It looked like MSI actually used good thermal paste this time so the IC diamond really only helped like 2-3C on the edge temp and my junction is the same. I retuned all my case fans to run higher at lower temps and that is not helping my junction temp at all either. I have 3 x 140mm arctic cooling case style fans on the intake with full mesh and one of those fans on the back blowing out while I also have 2 x 140mm arctic cooling static pressure fans pulling air over my radiator on the top of the case plus a like 80mm arctic cooling fan just by the gpu at the back of the case to help take the hot air coming out of it right out of the case. My cooling should be almost as good as it gets for an air cooled case yet I can't seem to affect my junction temps... I might take off the fan shroud on the gpu and strap a couple static pressure 120mm fans to it and see how that does. My buddy did it on an old card of his and the results were crazy.

I just want to take down my junction temps so I can push a bit more wattage and high clocks through this thing.

Thanks for the tip on raising the SoC voltage I will try that and see if it stabalises my memory.

So, you put your SoC (A) to 58 and not 60? Why?

 

Because my card was already stable at 2100 MEM & Fast settings in 99.999% cases. I am one of the careful persons rather than going full berserk/rage mode and bumping up everything without a second thought.
My card will thank for it later when some other (heavier OC) cards will eventually give up their lives.

 

Sounds reasonable. My memory seems to be stable at 2100 fast timings in most games but if I hit a very demanding one like Hunt Showdown I will crash unless I put the memory to default. I will see what 58A for the soc does for stability and go from there. Hey, what do you have the fan acoustic limit (RPM) set to? It was 1200 default for me and I put it up to 1800 and have not noticed much difference in terms of sound so far.

Also, undervolting in watmann actually does very little as using afterburner it was able to show the card was ramping up to 1150mv no matter what I did so I am trying limiting the max voltage to 1100 in MPT and so far it seems to be working very well for keeping around 2450mhz.

Have you done any tuning to your fan within wattman and if so what or just changed settings in sppt?

 

You mentioned that you are still getting black screens.
I thought you fixed it with using other version of MSI bios.
Maybe you should have considered to RMA your card before you modify it. It could be that it is defective.

The junction temperatures under normal condition without heavily Ocing should usually be ok. If that is also too high then there seems to issue with your GPU.

On my 6800XT reference, I had junction hitting about 92°C when OCing before with original cooler.
I have now watercooling for 6800XT and my junction temperature never goes much higher than 70°C no matter how high I OC my 6800XT and push power towards total of 380W.

 

Since I have a quite uhm... "BIG" Tower which is also silenced by me using a combination of silencing mats.... I was putting more effort on keeping the card cool rather than quiet... and on the 6900xt the limits are different vs 6800xt stock and/or custom models anyway...
I have actually changed only the zero RPM Start temperature and set it for 60 degrees (was 70) as I want the fan to kick in a little bit earlier. and I changed the curve in Wattman and set it gradually to increase to 100% when reaching ca. 85 degrees.
For me it works flawlessly with stock 6900xt - even with 100% (which is never really reached in normal games) the card is only barely audible...

So in this case sorry but I don't think I can help you more with the fan optimizations as I have gone in a completely different direction.

 

Yes not lot of RDNA2 are stable when VRAM is raised much above 2100MHz.

I am also careful with my daily using setting of 6800XT.
For testing purposes I push the power higher but still don't like to go much higher than what some custom cards do.

I oriented to max values towards what ASUS did with 6800XT Strix LC but my reference card has 10 phases Vcore compared to 14 on customs so I try not to push this as high as custom cards.
Until now I did not use core power higher than 320W total. It already gives decent performance while the card runs cool and stable without issues in Benchmark and games.
Actually in games 6800XT still consumes less than 300W core with the setting.
In Assassin's creed valhalla which I played very often in the recent time my 6800XT consuming around 250W on core.

6900XT needed to be pushed much higher with power to get more performance and there I had real fear that long term the card will be killed.
I was only getting 5% higher on 6900XT while pushing it 384W core power and 384A TDC which was for me too high to be comfortably have it set permanently.
Without that push performance increase was within 2% to my 6800XT.

 

I had crazy black screen issues in th ebeginning but I ended up fixing them by raising my SA and IO voltages to way beyond what they needed to be in order to have a stable system with my RTX 2080 and then I found out that since the 6800 XT seems to stress the system a lot more and is much more sensitive to voltage fluctuations than nVidia cards I was able to further tweak some previous stable voltages to get rid of all my black screens except for one. For some reason I cannot play Forza Horizon 4 on my Samsung Q80T @ 4k as it will just lock up the system completely. I can play it at 1440p on the Q80T and I can play all other games just fine at 4k on the Q80T - very strange. I really didn't want to RMA my card as it seemed like a lot of other people were having similar problems and I was confident I could sort them out to a degree and AMD would be fixing a lot of the issues over the next 6-12 months like they did with the 5xxx series. The last issue was Hunt crashing but it came down to my memory overclock but now that I am getting familiar with MPT I am hoping that slightly increasing the SOC (A) will help stabalize a memory OC and if not I guess it ireally isn't a big deal. This card is a monster and I am very excited to see what AMD has in store for us over the next 6-12 months.

I love to overclock everything I can get my hands on (my current cpu is a delidded 8086k with liquid metal going super strong) and I find that I have more fun with tweaking new hardware than actually using it for quite some time after purchasing it. I am going to figure out the best MPT settings for this card and maybe modify the cooler with some aftermarket fans. So excited!

I guess I need to learn about locking P-States next and more about this DXVK wrapper or whatever you call it. So much new and interesting stuff to go over after being away from AMD for so long. I am also excited to upgrade to Ryzen 6xxx when it comes out and have an all AMD build for the first time since the Athlon X2 and whatever GPU was out then, haha.

 

I have a weird feeling after reading this and I've gotta ask ---- how old is your PSU and how sure are you that the PSU is not the root cause ? If your 2080 was lil unstable then it would already ring alarm bells in my head... See the review on TPU >
https://www.techpowerup.com/review/toughpower-pf1-argb-1200-w/
"The platform is similar to the one used in the Thermaltake Toughpower iRGB Plus 1200, which had some reliability issues—my sample died after a while, and I got other reports that samples of that model died suddenly after a while. According to my sources, CWT fixed it. The main difference between the iRGB Plus and PF1 ARGB units is that the latter doesn't have a USB interface."

Make sure you are connecting the 8pins to separate lines etc. etc and if you can you might verify it with a different PSU.... sounds a little bit weird imho that you had also some other issues... especially under stress...

I personally would have a problem stating 100% stability with such issues you describe.... and would put the whole system through hell (and back) to check the stability --- e.g. running OCCT + Furmark etc etc to maximize the load on everything.

 

So my 6900 XT Asus TUF Gaming is here, installation went smooth, results in 3DMark Time Spy is 17800 points, all factory settings, only 9900k @ 4.9GHz and RAM profile at 4000@CL19, so dunno if this is okay score or not.

Temperature is high too I think, for GPU it's 72 Celcius and even 92 Celsius for hot spot!
Performance is okay, not in OpenGL games/apps sadly (I game in 4k and the performance is really bad with OpenGL and Ray Tracing games), still I'm not convinced if I will stay with this GPU tbh.

Btw, I really prefer AMD drivers control panel, can't believe it but it's the truth lol