The RX 6000 series Owners' Club

[kak4rot]

others have fixed the stutter by setting the Min OC and max OC. your just setting the max higher but leaving min at 500. i set my min to 2100, and max 2550, and have 0 issues. but stutter could be from your cpu or anything overheating, no idea. i have 0 issues at stock or oc'ing.

I had no these microstutters in the same pc with a 1080 ti, I have tried the solution you have provided me too. Actually my settings are 2539 Mhz Max Core Frequency and 2439 Mhz Min Core Frequency. I have tried too with power saver options, fresh install of Windows, Clear CMOS, Bios stock settings, etc. All the advices I have read at the whole internet.

 

[Space Lynx]

I had no these microstutters in the same pc with a 1080 ti, I have tried the solution you have provided me too. Actually my settings are 2539 Mhz Max Core Frequency and 2439 Mhz Min Core Frequency. I have tried too with power saver options, fresh install of Windows, Clear CMOS, Bios stock settings, etc. All the advices I have read at the whole internet.

your min is too high.

 

[kak4rot]

your min is too high.

can it be harmful? Any recommendation?

Yes I think you should RMA it if it stutters all time, but I am not sure if your warranty is still valid as you have opened it.

The stutter is very curious issue. My reference 6800XT also started stuttering but for me it happened after ESD issue.
My 6800XT card has been in repair over a month now.
I would be lucky if they repair it as I had it also under water block.
I took off warranty stickers in good state when I opened it. Still it would be possible to see that cooler was taken off if they look closely.

So let's see if at some point I get a replacement or they will send it back without repair.

In anyways I think it was good that I bought a replacement otherwise I would not have been able to use my computer for very long time.
I still don't have any definite time frame until when I get it back.

I didn't know it was taking amd so long to send replacement cards

 

[Felix123BU]

can it be harmful? Any recommendation?

What is the age and wattage of your PSU? I know for most its a rhetorical question, but did you connect it with two separate power cables? What CPU and RAM do you have? Answer to those questions might give an insight into why you might get stuttering.

 

[aRcane305]

That looks good. Only thing is that check if 2150MHz on VRAM gives you higher performance or something less.
On many 6800XT and 6900XT cards actually score degrades if memory is pushed to the limit.
Many times less than 2130MHz gives actaully higher performance.

I see. I did some more testing, and managed to get it to 2817Mhz core and 2100Mhz vram stable in Heaven and Timespy. any reason i shouldnt keep it as it is? I know for a fact it hurts mining hash rates, but would daily use/gaming be fine? i’ll do more testing to compare performance. 6900XT Toxic fyi, people said 2.7 was barely possible without LN2, but i managed to hit it with afterburner

 

[kak4rot]

What is the age and wattage of your PSU? I know for most its a rhetorical question, but did you connect it with two separate power cables? What CPU and RAM do you have? Answer to those questions might give an insight into why you might get stuttering.

My psu is a Cooler Master V1000 (1000w), as far as I know is bulit by Seasonic. It must be four or five years old. I use two separate pci cables to feed the monster.

 

[Felix123BU]

My psu is a Cooler Master V1000 (1000w), as far as I know is bulit by Seasonic. It must be four or five years old. I use two separate pci cables to feed the monster.

1000w is more than enough, but 4-5 years old could be an issue, power supplies rarely fail outright but over time they can deliver less power and more important less stable. I am not saying this is your case, might be, might not be, but if I where in your situation, I would either try to test it with another power supply, or test it in someone else's pc (which would be problematic since you put a waterblock on it), just to make sure its not the card. Any other things that might cause stutters are unstable overclocks on CPU and or RAM, which do not show up on day to day usage. One other thing could be a crappy display cable, had once that issue, all kinds of weirdness, weird screen stutter between them, new cable fixed it, but that is sort of rare.

As Turbogear said, its also possible that the card is physically defective, but for such a rare commodity as a GPU these days, I still would try to eliminate other potential problems not card related first.
If lets say you test it with another PSU, and it displays the same characteristics, RMA would be the way to go.

 

[turbogear]

can it be harmful? Any recommendation?


I didn't know it was taking amd so long to send replacement cards

In my case it is not AMD directly but rather XFX.
The reference card was XFX branded and the online shop that I bought it from sent it back to them.

I see. I did some more testing, and managed to get it to 2817Mhz core and 2100Mhz vram stable in Heaven and Timespy. any reason i shouldnt keep it as it is? I know for a fact it hurts mining hash rates, but would daily use/gaming be fine? i’ll do more testing to compare performance. 6900XT Toxic fyi, people said 2.7 was barely possible without LN2, but i managed to hit it with afterburner

WOW 2817MHz is really great.

How high did your score in Time Spy go?
Can you paste screen shot of your settings from Radeon Software as well as your MPT?

If the settings are really stable and the power consumption at this frequency is okay for you then you can leave it.
I usually run some stress test like Fire Strike Extreme Stress test to check the stability.
Or even Port Royal Stress test.
If these tests pass then the setting is stable.

Sometimes with Time Spy or Fire Strike Benchmarks, you can pass a run but on long term setting are not stable.
The stress tests helps to check the stability of the settings.

I know some guys with 6900XT Ultimate cards went to 500W Core Power and 384A TDC to get really high score and stability at that high frequencies like yours.
I am personally not comfortable of setting my 6900XTU to pull 500W on core.
Maybe only for benchmark but not for daily use.

I am keeping mine at the moment at 2750MHz@1175mV for daily gaming which is really stable and power consumption target is 382W and 368A TDC (default for my card) but does not consume that much in games.
To be honest, with my current 2K 165Hz monitor , I have more than enough FPS in the games that I play.

But for benchmarking I may go higher for short time.
Still fighting with myself if I should bite the bullet and put LM on my card to keep temperatures low when pushing it higher than 2750MHz.
The only reason not to move to LM is that after a few months the Nickel plated surface of the water block will start getting dull as well as the surface of the GPU Die gets some dullness which could cause issues if you RMA the card.

At least I found out that warranty stickers for my card can be bought on Aliexpress.

 

[aRcane305]

In my case it is not AMD directly but rather XFX.
The reference card was XFX branded and the online shop that I bought it from sent it back to them.


WOW 2817MHz is really great.

How high did your score in Time Spy go?
Can you paste screen shot of your settings from Radeon Software as well as your MPT?

If the settings are really stable and the power consumption at this frequency is okay for you then you can leave it.
I usually run some stress test like Fire Strike Extreme Stress test to check the stability.
Or even Port Royal Stress test.
If these tests pass then the setting is stable.

Sometimes with Time Spy or Fire Strike Benchmarks, you can pass a run but on long term setting are not stable.
The stress tests helps to check the stability of the settings.

I know some guys with 6900XT Ultimate cards went to 500W Core Power and 384A TDC to get really high score and stability at that high frequencies like yours.
I am personally not comfortable of setting my 6900XTU to pull 500W on core.
Maybe only for benchmark but not for daily use.

I am keeping mine at the moment at 2750MHz@1175mV for daily gaming which is really stable and power consumption target is 382W and 368A TDC (default for my card) but does not consume that much in games.
To be honest, with my current 2K 165Hz monitor , I have more than enough FPS in the games that I play.

But for benchmarking I may go higher for short time.
Still fighting with myself if I should bite the bullet and put LM on my card to keep temperatures low when pushing it higher than 2750MHz.
The only reason not to move to LM is that after a few months the Nickel plated surface of the water block will start getting dull as well as the surface of the GPU Die gets some dullness which could cause issues if you RMA the card.

At least I found out that warranty stickers for my card can be bought on Aliexpress.

Alright I will do more testing. 2735Mhz and 2150Mhz got me a 21609 gpu score in Timespy for a total of 17 850 (with a 5600x).
Actual sythetic clocks seem to be -50Mhz, giving a 2701Mhz in timespy.
I will do more test rounds to test, and i'll keep bumping up the core clocks.

Power consumption still hasn't passed 400W.

My settings are default bios from Techpowerup, use MPT to raise Power Limit (GPU) to 525W, TDC GFX to 525A and TDC SOC to 80A.
Then Wattman, custom settings, only thing i did was raise the power limit to 15%.
Then in Afterburner drag the vram and core clocks up.

Have yet to test difference between 2100, 2130, and 2150Mhz for vram. 2150 is the maximum the slider allows.

For the core clocks in afterburner, I can slide it up to 2860Mhz, for a real synthetic performance of 2817Mhz in Heaven and Timespy.

Anything over 2860Mhz in Afterburner leads to Heaven crashing.

I did these all yesterday late at night, have yet to properly test/benchmark.


Do you reckon i should mess with more settings in MPT?


I'll send screenshots after i get home and do more testing.

 

[turbogear]

Alright I will do more testing. 2735Mhz and 2150Mhz got me a 21609 gpu score in Timespy for a total of 17 850 (with a 5600x).
Actual sythetic clocks seem to be -50Mhz, giving a 2701Mhz in timespy.
I will do more test rounds to test, and i'll keep bumping up the core clocks.

Power consumption still hasn't passed 400W.

My settings are default bios from Techpowerup, use MPT to raise Power Limit (GPU) to 525W, TDC GFX to 525A and TDC SOC to 80A.
Then Wattman, custom settings, only thing i did was raise the power limit to 15%.
Then in Afterburner drag the vram and core clocks up.

Have yet to test difference between 2100, 2130, and 2150Mhz for vram. 2150 is the maximum the slider allows.

For the core clocks in afterburner, I can slide it up to 2860Mhz, for a real synthetic performance of 2817Mhz in Heaven and Timespy.

Anything over 2860Mhz in Afterburner leads to Heaven crashing.

I did these all yesterday late at night, have yet to properly test/benchmark.


Do you reckon i should mess with more settings in MPT?


I'll send screenshots after i get home and do more testing.


Side note here are some pics of my build, an upvote would help me get a feature too! https://au.pcpartpicker.com/b/tqzNnQ

Well your power settings are way too high.
As far as I know Saphire Toxic has 2x8 pin and 1x6pin. So total specified PCIe power limit would be only 450W but not 525W.
Higher than that you are running the card at your own risk.

Your MPT setting: 525W and 525A TDC.
You added 15% to that with wattmann. This means your limits are set to: 603.75W GPU core and 603.75A TDC.

603.75W is not total GPU power in this case but only Core GPU power. You need to add to that about another 45W for other GPU components.
This means 648.75W is the total possible power that your card is allowed to draw from power supply over PCIe cables.

So I am not sure if these settings are really safe or you could damage your card although you say you have max 400W used power.

Regarding frequency, yes it is like that for RDNA2 cards. You get max -50MHz to the setting you apply. So if you set it to 2817MHz than you will see max real frequency of 2767MHz.

I personally don't use Afterburner but do all OCing through Wattman.
Some users here on forum had some compatibility issues with Afterburner.

 

[kak4rot]

1000w is more than enough, but 4-5 years old could be an issue, power supplies rarely fail outright but over time they can deliver less power and more important less stable. I am not saying this is your case, might be, might not be, but if I where in your situation, I would either try to test it with another power supply, or test it in someone else's pc (which would be problematic since you put a waterblock on it), just to make sure its not the card. Any other things that might cause stutters are unstable overclocks on CPU and or RAM, which do not show up on day to day usage. One other thing could be a crappy display cable, had once that issue, all kinds of weirdness, weird screen stutter between them, new cable fixed it, but that is sort of rare.

As Turbogear said, its also possible that the card is physically defective, but for such a rare commodity as a GPU these days, I still would try to eliminate other potential problems not card related first.
If lets say you test it with another PSU, and it displays the same characteristics, RMA would be the way to go.

When I notice stutters I can see a fps drop, for example: from 144 fps to 120fps, from 200 fps to 140 fps. I haven't these micro stutters with my oced 1080 ti in the same system.

I have tried with stock bios settings, clearing CMOS and a fresh Window installation.

 

[aRcane305]

Well your power settings are way too high.
As far as I know Saphire Toxic has 2x8 pin and 1x6pin. So total specified PCIe power limit would be only 450W but not 525W.
Higher than that you are running the card at your own risk.

Your MPT setting: 525W and 525A TDC.
You added 15% to that with wattmann. This means your limits are set to: 603.75W GPU core and 603.75A TDC.

603.75W is not total GPU power in this case but only Core GPU power. You need to add to that about another 45W for other GPU components.
This means 648.75W is the total possible power that your card is allowed to draw from power supply over PCIe cables.

So I am not sure if these settings are really safe or you could damage your card although you say you have max 400W used power.

Regarding frequency, yes it is like that for RDNA2 cards. You get max -50MHz to the setting you apply. So if you set it to 2817MHz than you will see max real frequency of 2767MHz.

I personally don't use Afterburner but do all OCing through Wattman.
Some users here on forum had some compatibility issues with Afterburner.

Regarding what you just said, i still have yet to touch 360W, let alone 400W so I reckon its alright for now.

So if i where to change the MPT values, would i change power limit to 425W?
or to 370, because 370 + 15% = 425W.

also do i change anything else in MPT besides Power Limit GPU, TDC Limit GFX, and TDC Limit SOC?

 

[turbogear]

Regarding what you just said, i still have yet to touch 360W, let alone 400W so I reckon its alright for now.

So if i where to change the MPT values, would i change power limit to 425W?
or to 370, because 370 + 15% = 425W.

also do i change anything else in MPT besides Power Limit GPU, TDC Limit GFX, and TDC Limit SOC?

I did not change anything else in MPT.
Many of the other values there are not working anyways for RDNA2.
For example if one increases max voltage or max frequency then the card goes into safe mode and gets locked at 500MHz.

For power settings, you can do it either way. Both will work.
If you use HWINFO then you can also see what are the actual TDC and max Core power limits that are applied.
It can also shows things like current drawn by GPU core.

One thing to note is that if you just for example set in MPT 370W and 350ATDC and then use 15% slider in Wattman, you can see in HWINFO that the actual setting would be 425.5W and 402.5A TDC. So whatever values you put in MPT, these will go 15% higher for each individual value of Core Power and TDC.

 

[aRcane305]

I did not change anything else in MPT.
Many of the other values there are not working anyways for RDNA2.
For example if one increases max voltage or max frequency then the card goes into safe mode and gets locked at 500MHz.

For power settings, you can do it either way. Both will work.
If you use HWINFO then you can also see what are the actual TDC and max Core power limits that are applied.
It can also shows things like current drawn by GPU core.

One thing to note is that if you just for example set in MPT 370W and 350ATDC and then use 15% slider in Wattman, you can see in HWINFO that the actual setting would be 425.5W and 402.5A TDC. So whatever values you put in MPT, these will go 15% higher for each individual value of Core Power and TDC.

Ok, so unfortunately after a system reboot, my previous 2817Mhz is no longer Heaven stable, let alone timespy.

After a 2 hour run, 2800Mhz and 2130Mhz mem clock finally yielded a 1hr stable Heaven run. I will now run Timespy and Port Royal to test.

So "real" synthetic performance is 2750Mhz ish.


I have one question, if the overclock is not stable in benchmarks at 99% utilisation, is it still safe to use the overclock in games if its stable in games?

I did not change anything else in MPT.
Many of the other values there are not working anyways for RDNA2.
For example if one increases max voltage or max frequency then the card goes into safe mode and gets locked at 500MHz.

For power settings, you can do it either way. Both will work.
If you use HWINFO then you can also see what are the actual TDC and max Core power limits that are applied.
It can also shows things like current drawn by GPU core.

One thing to note is that if you just for example set in MPT 370W and 350ATDC and then use 15% slider in Wattman, you can see in HWINFO that the actual setting would be 425.5W and 402.5A TDC. So whatever values you put in MPT, these will go 15% higher for each individual value of Core Power and TDC.

OK suprise suprise, currently running Timespy, and interestingly the overclocks applied in Wattman glitch out, while the overclocks applied in Afterburner were stable. The overclocks in the post above were all done in Wattman.

This is wierd, i can somehow reach higher stable overclocks in Afterburner, compared to Wattman.

 

[sepheronx]

I had no these microstutters in the same pc with a 1080 ti, I have tried the solution you have provided me too. Actually my settings are 2539 Mhz Max Core Frequency and 2439 Mhz Min Core Frequency. I have tried too with power saver options, fresh install of Windows, Clear CMOS, Bios stock settings, etc. All the advices I have read at the whole internet.

So I fixed my stuttering issue of my 6800xt as I noticed it would stutter at random as it would drop to 500mhz and then shoot up. The part where it drops and shoots up in terms of frequency, is where the microstutters come from. The min freq is low and because the card may not be stressed enough in the game, it auto adjusts and it just sucks. I raised my min to 2100mhz and my max is about 2300mhz (I did raise it more but not wanting to ruin life of gpu) and after doing that, my performance has been a lot better. Min still goes low, like 1700mhz at randomly, but the microstutters were reduced considerably.

Its the crappy power management system.

 

[turbogear]

Ok, so unfortunately after a system reboot, my previous 2817Mhz is no longer Heaven stable, let alone timespy.

After a 2 hour run, 2800Mhz and 2130Mhz mem clock finally yielded a 1hr stable Heaven run. I will now run Timespy and Port Royal to test.

So "real" synthetic performance is 2750Mhz ish.


I have one question, if the overclock is not stable in benchmarks at 99% utilisation, is it still safe to use the overclock in games if its stable in games?

View attachment 200625View attachment 200626


OK suprise suprise, currently running Timespy, and interestingly the overclocks applied in Wattman glitch out, while the overclocks applied in Afterburner were stable. The overclocks in the post above were all done in Wattman.

This is wierd, i can somehow reach higher stable overclocks in Afterburner, compared to Wattman.

If the stress test is not 99% stable, it can still work in games, but it could happen that you have sometimes random game crashes depending on the games and if they can drive the GPU to limit.

For example I had issues before with Control and my 6800XT at 2750MHz@1020mV. This game crashed instantly although setting was 99.8% stable in stress tests. For Control I had to drop back to 2675MHz@985mV to get it stable.

You can try it and if the games you play are working fine then you can keep it.

Regarding what you said about now not able to achieve higher frequency after reboot, I had similar experiences in the past with not 100% stable settings.
Sometime they worked perfectly and all benchmarks would go through and next day this setting doesn't work.

OK suprise suprise, currently running Timespy, and interestingly the overclocks applied in Wattman glitch out, while the overclocks applied in Afterburner were stable. The overclocks in the post above were all done in Wattman.

This is wierd, i can somehow reach higher stable overclocks in Afterburner, compared to Wattman.

Okay, that is an interesting discovery.
Seems like Afterburner is less bugy in this situation than Radeon Software.
Maybe I should try Afterburner.
I never used that before for Ocing but only for monitoring. The only time I tried it I was unable to unlock OC feature.

You don't have TRIXX installed or?
I had 6900XT Toxic for a short time, and TRIXX was interfering with Wattmann. I returned that card at that time as my sample was not able to do more than 2620MHz real clock no matter how high I pushed it with power.

 

[aRcane305]

If the stress test is not 99% stable, it can still work in games, but it could happen that you have sometimes random game crashes depending on the games and if they can drive the GPU to limit.

For example I had issues before with Control and my 6800XT at 2750MHz@1020mV. This game crashed instantly although setting was 99.8% stable in stress tests. For Control I had to drop back to 2675MHz@985mV to get it stable.

You can try it and if the games you play are working fine then you can keep it.

Regarding what you said about now not able to achieve higher frequency after reboot, I had similar experiences in the past with not 100% stable settings.
Sometime they worked perfectly and all benchmarks would go through and next day this setting doesn't work.



Okay, that is an interesting discovery.
Seems like Afterburner is less bugy in this situation than Radeon Software.
Maybe I should try Afterburner.
I never used that before for Ocing but only for monitoring. The only time I tried it I was unable to unlock OC feature.

You don't have TRIXX installed or?
I had 6900XT Toxic for a short time, and TRIXX was interfering with Wattmann. I returned that card at that time as my sample was not able to do more than 2620MHz real clock no matter how high I pushed it with power.

For now I'm back to using the Toxic Boost in TRIXX as my daily driver. Overclocking only improves fps and benchmark scores.

Yes, I have TRIXX. Interesting thing today, even 2750Mhz is not stable for me in Timespy. Really wierd.
I'll have a break today and do more testing/overclocking tomorrow.

Here are some pics of my build if your interested. I'm hoping to get featured on PcPartpicker, if you could upvote that be awesome.

https://au.pcpartpicker.com/b/tqzNnQ#cx3954658

 

[Raendor]

Ok, I’m not the club member anymore as I got rid of my 6800xt and traded it for 3060ti and a couple hundred cash until I get my hands on 3080. After month of struggle with amd drivers I just gave up. A bit sad, considering the raw power is great and stock model itself is outstanding. But success to all you guys and hope your experience is miles better than mine turned out to be!

 

[Felix123BU]

When I notice stutters I can see a fps drop, for example: from 144 fps to 120fps, from 200 fps to 140 fps. I haven't these micro stutters with my oced 1080 ti in the same system.

I have tried with stock bios settings, clearing CMOS and a fresh Window installation.

Wait, looking at those fps ranges you mentioned, is your display Freesync? One thing that can cause stutters is if your FPS goes over/under the Freesync range
What would happen if lets say in said game you enable Vsync which would lock the FPS at your monitors refresh rate?

 

[turbogear]

For now I'm back to using the Toxic Boost in TRIXX as my daily driver. Overclocking only improves fps and benchmark scores.

Yes, I have TRIXX. Interesting thing today, even 2750Mhz is not stable for me in Timespy. Really wierd.
I'll have a break today and do more testing/overclocking tomorrow.

Here are some pics of my build if your interested. I'm hoping to get featured on PcPartpicker, if you could upvote that be awesome.

https://au.pcpartpicker.com/b/tqzNnQ#cx3954658

That's really a nice and clean build. It looks really beautiful.

That somehow sounds similar to what I experienced with Toxic.
One time I was able to go to higher than 2750MHz at around 1050mV, but afterwards on next day I tried everything but I could not get it higher than 2675MHz@1090mV.
Never understood the reason but returned the card as at that time I had a super 6800XT sample that went 2750MHz@1020mV and was only 5% slower than my Toxic while consuming much less power.
I kept 6800XT and returned Toxic.

Unfortunately 6800XT died due to my fault of not being careful with ESD and now I own 6900XTU.

 

[kak4rot]

Wait, looking at those fps ranges you mentioned, is your display Freesync? One thing that can cause stutters is if your FPS goes over/under the Freesync range
What would happen if lets say in said game you enable Vsync which would lock the FPS at your monitors refresh rate?

I disabled Freesync on both, wattman and monitor. And microstutters are present with vsync on and off

 

[Felix123BU]

I disabled Freesync on both, wattman and monitor. And microstutters are present with vsync on and off

Other than testing with another power supply, not a lot I can suggest since you did all the other potential issue checking, especially the BIOS reset which you said you tried which would have alleviated any RAM OC instability, so its either PSU related, or the display cable, or the GPU has a physical problem, in which case RMA is the way to go.

 

[kak4rot]

Other than testing with another power supply, not a lot I can suggest since you did all the other potential issue checking, especially the BIOS reset which you said you tried which would have alleviated any RAM OC instability, so its either PSU related, or the display cable, or the GPU has a physical problem, in which case RMA is the way to go.

I don't know why, but I think my problem is related to the drivers. With the settings that you have recommended to me here and in other forums, the microstutter has decreased a lot and now they are related to isolated transitions from low or medium to high load in games, because in benchmarks it always goes full.

But his thing would be to try it on another PC, for this I would have to drain the circuit and reassemble the original heatsink... what laziness

 

[Felix123BU]

I don't know why, but I think my problem is related to the drivers. With the settings that you have recommended to me here and in other forums, the microstutter has decreased a lot and now they are related to isolated transitions from low or medium to high load in games, because in benchmarks it always goes full.

But his thing would be to try it on another PC, for this I would have to drain the circuit and reassemble the original heatsink... what laziness

I soo know the feeling , water makes things lots cooler, but when it comes to even taking out anything..... drain the loop it is oh well, life is not perfect

Easier though if you would have an extra psu, but who keeps a brand new or newer psu not in his case

Why i am suggesting some things, because my experience with my 6800XT has been almost flawless, so i guess or hope for you it's either something that can be fixed, the alternative would be some hardware issue and a a RMA.

 

[Godhand007]

So have people tried any Ray Tracing benchmarks on 6900Xt/6800XT cards? It seems that both these cards are placed between 3070 and 3080 for RT performance (reasonable settings).