The RX 6000 series Owners' Club

[Felix123BU]

Had some hard crashes with the game Destiny 2, it send my mashine to hard crashes at the latest so hard that it took some 10 boot cycles to get it back on again. Tried the new driver and the previous - no matter what I do the machine crashes after 5-10 minutes at best. It seems that the latest season pack has some issues with AMD gpu and cpus. Any one else haveing these issues?

Sorry to hear that, must really suck. I don't have the game, but reading around there is a double issue, the developer and the possibly the driver, not sure if Bungie or AMD are to blame, seems Nvidia also had a ton of issues. But the fact that you are getting hard resets probably points to some underlying general issue that's not necessarily game related. I have had 0 issues with a combo of 6800XT and Ryzen 5800X in any game.

 

[watzupken]

I still wonder, since all here ho tried the new driver noticed a significant performance gain, why is there 0 reference about performance gains in the driver release notes? Been last 10 years on AMD Gpu's, previous 5 with Nvidia, this is the biggest performance leap I have noticed with one driver update

May be a direct effect of this, "Among the issues fixed with Radeon Software Adrenalin 21.3.1 include high CPU utilization for Radeon Software, even when idling", just guessing. You reduce CPU utilization by the software, which may result in a slight bump in performance.

 

[deadman3000]

I have a 6700 XT reference model. Pondering playing with morepowertool to see if I can push a little more performance from it. I read that I should only touch the bottom three settings. Anyone have any 'safe' suggestions?

 

[Felix123BU]

I have a 6700 XT reference model. Pondering playing with morepowertool to see if I can push a little more performance from it. I read that I should only touch the bottom three settings. Anyone have any 'safe' suggestions? View attachment 193356

Congrats! Nobody really knows what "safe" settings are, but for me a maximum of plus 15% to those values would be it, and I would start incrementally, 5% extra, then 10% and see if there is a change. But keep in mind, although increasing by a bit should in theory not damage the card in any way, its a risk, small, but a risk nonetheless. I would also try to find out how much the board can maximally take, no piece of hardware is run at its absolute limits at stock, so there is always some margins to be gained potentially. Generally AMD reference boards where historically over-engineered, so they could take a lot more, but do not take this as granted for the current gen or your card, its still an open question even around here.

My 6800XT is a reference model, and the only limiting factor is power allowance, overclocking helped, raising power via MPT helped also helped. I run it with 15% over the stock power limits, but I think I will lower it to plus 10%.

 

[turbogear]

Congrats! Nobody really knows what "safe" settings are, but for me a maximum of plus 15% to those values would be it, and I would start incrementally, 5% extra, then 10% and see if there is a change. But keep in mind, although increasing by a bit should in theory not damage the card in any way, its a risk, small, but a risk nonetheless. I would also try to find out how much the board can maximally take, no piece of hardware is run at its absolute limits at stock, so there is always some margins to be gained potentially. Generally AMD reference boards where historically over-engineered, so they could take a lot more, but do not take this as granted for the current gen or your card, its still an open question even around here.

My 6800XT is a reference model, and the only limiting factor is power allowance, overclocking helped, raising power via MPT helped also helped. I run it with 15% over the stock power limits, but I think I will lower it to plus 10%.

I have mine set at 30% of stock 255W to 332W.
But as I showed in my previous post, my 6800XT with undervolt to 1020mV at 2745MHz it does not consume more than 290W on average on the core. At least this is the case in Assassin's creed valhalla, Cyberpunk and Borderlands 3.
I have also own Control which I have not yet tested.
I will test more games and if in all of them the core consumption remains below 300W then I will leave the high settings. My GPU only uses 332W in stress tests and benchmarks.

@deadman3000
I would also suggest similar strategy like @Felix123BU suggested to find the good balance for your card.
At certain points the power consumption is much higher than actual gained performance and after that point it is not worth to increase it further. Keep also an eye on the thermals.

 

[Butanding1987]

I think turbogear’s 30% already includes the 15% extra power provided by Wattman (Radeon software). That means he uses a GPU power limit of 289 watts in More Power Tool. This is my card’s default setting ( without MPT). This means I get ~333w in total power. I haven’t decided yet what is optimal for my card. I was thinking maybe I should add 9w, which is roughly equivalent to the 3 stock fans that I lost since I installed a water block. Not sure yet.

 

[turbogear]

I think turbogear’s 30% already includes the 15% extra power provided by Wattman (Radeon software). That means he uses a GPU power limit of 289 watts in More Power Tool. This is my card’s default setting ( without MPT). This means I get ~333w in total power. I haven’t decided yet what is optimal for my card. I was thinking maybe I should add 9w, which is roughly equivalent to the 3 stock fans that I lost since I installed a water bkock. Not sure yet.

Yes 30% includes the 15% extra provided by Wattmann.
Exactly as you said, I have my reference card set to 289W in MPT.
I set it one time for testing to 295W to get higher Time Spy Score but I don't need it for gaming as most of games does not consume that high anyways.
Many of custom cards has 289W by default, but these cards like yours have 14-phase Vcore VRM and mine only 10-phase. 10-phase should also be good enough with decent coolig with the waterblock.

By the way the only game that I found that consumes power like crazy is Control on maximum setting without Raytracing.
I tested it for first time yesterday and that is the only game where my 2745MHz@1020mV OC was not good and led to game crashes. I believe my power target of 330W for core was not enough to run this game that high on clock.
In this game I had to back it up to 2675MHz@985mV. Still at the lower frequency the game was pushing 330W on the GPU core.

 

[Butanding1987]

Yes 30% includes the 15% extra provided by Wattmann.
Exactly as you said, I have my reference card set to 289W in MPT.
I set it one time for testing to 295W to get higher Time Spy Score but I don't need it for gaming as most of games does not consume that high anyways.
Many of custom cards has 289W by default, but these cards like yours have 14-phase Vcore VRM and mine only 10-phase. 10-phase should also be good enough with decent coolig with the waterblock.

By the way the only game that I found that consumes power like crazy is Control on maximum setting without Raytracing.
I tested it for first time yesterday and that is the only game where my 2745MHz@1020mV OC was not good and led to game crashes. I believe my power target of 330W for core was not enough to run this game that high on clock.
In this game I had to back it up to 2675MHz@985mV. Still at the lower frequency the game was pushing 330W on the GPU core.

I finished that game on my 2080 Super, which I have since sold.

 

[deadman3000]

With the stock VBIOS on the reference 6700 XT overclocked to 2.8 Ghz 2130 Mhz memory @1100 Mv and 70% fan curve (15% power limit) on a 3700X (PBO 300.230,230), DDR 4 3600 Mhz XMP1 on a MSI B450M Mortar (latest BIOS and SAM enabled) I got 12116 in Time Spy. Using morepowertool I loaded up the XFX.RX6700XT.12272.210128.rom which raises the power limit from 186 watts to 211 watts and applied. After a reboot I got a score of 12328. I was also able to boost the score to 12355 by bumping the memory to 2150 Mhz but I have not tested for stability. I am now top of the leader board for people who benched on a 3700X + 6700 XT. Now I want a 5800X

I shall wait until some others try raising their power limit higher than 211 watts before I chance doing so myself.

 

[kapone32]

I am downloading the latest driver right now. I look forward to see what this does.

 

[Felix123BU]

I am downloading the latest driver right now. I look forward to see what this does.

Would be cool if you can do a before/after comparison of performance in 1-2 games, I am really curious if you also are in awe like me from the latest driver

 

[kapone32]

Would be cool if you can do a before/after comparison of performance in 1-2 games, I am really curious if you also are in awe like me from the latest driver

Well I just finished running 3D mark. The crazy thing is it showed a boost clock of 2687 MHZ!!!!! I am not even sure what Game to try first but I have to pick up my daughter from Daycare so I think I will run a TWWH2 benchmark.

 

[turbogear]

So I was experimenting and fine tuning my GPU further.

One of the concerns in the past was the binding of the VRAM and VRM to the water block because the standard pads that EK provides with the GPU are 3.5W/m·K.
I have ordered EC360 Silver which are 12 W/m·K (1mm pads).

See the before and after photos from the GPU below.
I also replaced on the backplate EK pads with EC360 (1.5mm und 2mm).

For GPU chip I used as usual Thermal Grizzly Kryonaut.
The hotspot remains as expected after rebuild in the same range.
Max hotspot in Time Spy was always in range of 69°C-70°C.

Now I can push with more peace of mind the Core power draw towards 340W total.

I don't have unfortunately a Thermal Camera to compare the temperature difference on the PCB.

I did see that now HWiNFO has in the recent versions added Vcore Temperatur measurement (GPU VR VDDC Temperature).
This is hotspot temperature of the Vcore. There is 10-phase VCore on the reference 6800XT card.
These power stages are from International Rectifier TDA21472 DrMOS and are rated to -40°C to 125°C.
Max temperature reading is 50°C for me under Benchmarks.

What I also noticed is that for some reason coil whine is now better after I changed the pads.






Here are my actual Benchmark results for the 6800XT with driver 21.3.2.
Some increase with new driver.


With new result, I hit the third place in Leadership board for Time Spy und Port Royal sorted for GPU score for 6800XT with 5800x processor.
I could increase setting from 2745MHz@1020mV to 2750MHz@1020mV without crash. It used to crash before.

Port Royal: Search (3dmark.com)
Time Spy: Search (3dmark.com)

Result with 21.3.1 are here:

The RX 6000 series Owners' Club

I'm not quite sure what to make of stability because TS and PR seem to require different settings for optimum performance. Time Spy likes high clock rates, while Port Royal hates it, at least in my case. Anyway, here are my results.

www.techpowerup.com

 

[Butanding1987]

Any reason why you're not using 63A for TDC limit SOC?

 

[turbogear]

Any reason why you're not using 63A for TDC limit SOC?

Well after my recent testing I saw that for some reason my card can now still achieve same 2110MHz on VRAM even without increasing SoC TDC.
Maybe my better cooling pads affeced that?
I have no idea but yesterday during test of 2110MHz on VRAM without TDC incease, it did not cause my score to regress.

After changing the pads I kind of started with default MPT and went up with the parameters separately.

I will test it again anyways in the next days and see if I can go higher with VRAM frequency or still 2114MHz is my absolute limit.

I am even tempting to test if higher than 340W GPU core will give more performance.
With 340W on core, I am already out of official 375W limit of 2x8pins card as I have approximately 385W draw in total, but with VRM only hitting 50°C during benchmarks, I am tempting to test with another 10-15W.

 

[basco]

interesting that there is so much diff in the time spy score but not port royal with nearly the same clocks.
congrats on the scores even if i think you are overdoing it when ya mention 5mhz difference in ram- but more power to you-du kitzelst wirklich alles raus aus der karte

 

[turbogear]

interesting that there is so much diff in the time spy score but not port royal with nearly the same clocks.
congrats on the scores even if i think you are overdoing it when ya mention 5mhz difference in ram- but more power to you-du kitzelst wirklich alles raus aus der karte

The 5MHz higher for GPU clock does not bring much. It only for rounding it at 2750MHz instead of 2745MHz looks nicer.

The 72 higher Time Spy point from 20608 to 20680 are for same power target for Core and TDC. So these 72 points are coming most probably from new driver 21.3.2 vs 21.3.1 and also better cooling could have helped.
The Port Royal Score is also about 67 higher.

Ja ich glaube das ist limit von meine chip.

In games with these settings, the power draw is not that high.
Here is test from Cyberpunk.

The RX 6000 series Owners' Club

I'm not quite sure what to make of stability because TS and PR seem to require different settings for optimum performance. Time Spy likes high clock rates, while Port Royal hates it, at least in my case. Anyway, here are my results.

www.techpowerup.com

Ones the new patch with Raytracing support is available, I will check how higher power consumption will go.

 

[basco]

and now go rock the charts at hwbot !

3DMark - Time Spy overclocking records @ HWBOT

Overclocking records

hwbot.org

 

[Felix123BU]

So I was experimenting and fine tuning my GPU further.

One of the concerns in the past was the binding of the VRAM and VRM to the water block because the standard pads that EK provides with the GPU are 3.5W/m·K.
I have ordered EC360 Silver which are 12 W/m·K (1mm pads).

See the before and after photos from the GPU below.
I also replaced on the backplate EK pads with EC360 (1.5mm und 2mm).

For GPU chip I used as usual Thermal Grizzly Kryonaut.
The hotspot remains as expected after rebuild in the same range.
Max hotspot in Time Spy was always in range of 69°C-70°C.

Now I can push with more peace of mind the Core power draw towards 340W total.

I don't have unfortunately a Thermal Camera to compare the temperature difference on the PCB.

I did see that now HWiNFO has in the recent versions added Vcore Temperatur measurement (GPU VR VDDC Temperature).
This is hotspot temperature of the Vcore. There is 10-phase VCore on the reference 6800XT card.
These power stages are from International Rectifier TDA21472 DrMOS and are rated to -40°C to 125°C.
Max temperature reading is 50°C for me under Benchmarks.

What I also noticed is that for some reason coil whine is now better after I changed the pads.


View attachment 194580View attachment 194581View attachment 194582View attachment 194583



Here are my actual Benchmark results for the 6800XT with driver 21.3.2.
Some increase with new driver.


With new result, I hit the third place in Leadership board for Time Spy und Port Royal sorted for GPU score for 6800XT with 5800x processor.
I could increase setting from 2745MHz@1020mV to 2750MHz@1020mV without crash. It used to crash before.

Port Royal: Search (3dmark.com)
Time Spy: Search (3dmark.com)

Result with 21.3.1 are here:

The RX 6000 series Owners' Club

I'm not quite sure what to make of stability because TS and PR seem to require different settings for optimum performance. Time Spy likes high clock rates, while Port Royal hates it, at least in my case. Anyway, here are my results.

www.techpowerup.com

View attachment 194600

View attachment 194601


View attachment 194598View attachment 194599

You said you noticed some increased performance, curious, how much and in what games/apps?

 

[turbogear]

and now go rock the charts at hwbot !

3DMark - Time Spy overclocking records @ HWBOT

Overclocking records

hwbot.org

It is tempting.
I am not a member there though.

You said you noticed some increased performance, curious, how much and in what games/apps?

As mentioned in the above post the increase was in Time Spy with 72 point from 20608 to 20680 and the Port Royal Score is also about 67 higher.
I have not test the games with the new drivers yet.

 

[Butanding1987]

It is tempting.
I am not a member there though.


As mentioned in the above post the increase was in Time Spy with 72 point from 20608 to 20680 and the Port Royal Score is also about 67 higher.
I have not test the games with the new drivers yet.

Maybe you should test it further. Sometimes I get 70 points more in Port Royal depending on the time of day.

 

[Athlonite]

Maybe you should test it further. Sometimes I get 70 points more in Port Royal depending on the time of day.

I also find it helps to stand on one leg while hoping up and down and sticking out my tongue

 

[Felix123BU]

I also find it helps to stand on one leg while hoping up and down and sticking out my tongue

You guys are not up to speed, a Lisa Su enchantment will make it glow in the dark with no power needed

 

[Felix123BU]

I was wondering if anybody else got their hand on a 6000 series Gpu, looking at the Steam hardware survey one would think nobody has any 6000 Gpu's, which is weird since I know at least 6 people who got their hand on one of the 6000 models, not including the venerable members of this thread

 

[Space Lynx]

I was wondering if anybody else got their hand on a 6000 series Gpu, looking at the Steam hardware survey one would think nobody has any 6000 Gpu's, which is weird since I know at least 6 people who got their hand on one of the 6000 models, not including the venerable members of this thread

the survey is flawed imo, they should just send it to all steam users. it literally takes 5 seconds. its w.e i dont care anymore