The RX 6000 series Owners' Club

[Felix123BU]

New poster, and proud owner of a new 6800XT! Just up picked a Power Color Reference model last week. Had to pay £100 over the MSRP to secure it but well worth it in the end once I got it home, put it under water and started running some benchmarks.

I bought myself an EK Quantum-Vector Water Block and installed it a few days ago as well, and put it on a custom loop with a couple of EK 360mm Radiators (1 x PE and 1 x XE) to cool both the CPU and GPU.

On idle, I'm around 30 degrees and on full load during a Port Royale/Time Spy stress test, I'm around 54-55 degrees with a Junction Temp. of around 70 after an hour of looping. During gaming, I'm usually around 50/66-67 respectively.

During a TimeSpy bench, I was hitting a graphics score of 20329 just by tinkering with the Radeon software and under volting. Setting my core clock slider between 2600-2700mhz, 960mv, and 2110mhz Fast Timings on the memory. It would stay at a consistent 2620-2620mhz on benches. When setting the "auto overclock" feature on the software, it would set it as 2508mhz automatically, so I knew immediately I had a pretty good card on my hands.

I think tinkered around a bit more with MPT (More Power Tool) and increased the total wattage output on the card from 300w to 340w, and managed to get it benchmarkable on the slider between 2690-2790mhz and outputting around 2740mhz whilst increasing the voltage to 1085mv. I did manage to then score 20900+ on TimeSpy (screenshot attached) but for some reason I had an issue with System Mark so couldn't validate the score. It wasn't stable though.

Then, put it back down to 2655-2755 on the slider and scored a 20650. This was the most stable I've managed to get it since.

All in all, I must have lucked out with the silicon lottery to pick up a card that decent on OC'ing, so I'm quite happy to have paid the extra £100 for it in hindsight.

Also have the reference form Power color, and mine is also a vey good overclocker, even with the stock cooler, can do 2.6 on stock air, and will see how much more it can on water. Temps become an issue after 2.6, hence the coming water loop.

I am using EK-DDC 3.2 PWM pump. EK branded Laing with extra EK heat sink mounted at the bottom to help with cooling of the motor and prolong the life of it.

I have the PWM from DDC connected to my ASUS Crosshair VIII hero CPU fan port.
I have in AI Suite a profile to control the speed of cooling fans and pump depending on CPU temperatures.
When doing normal office stuff like surfing, the loop is quiet and ramp up higher when gaming.
I allow it ramp up quite high while gaming, producing more noise but optimum temperatures.
I game in any case with headset so it does not disturb much.

One thing that I noticed after I went water cooling with 6800XT reference card is that there is significant coil whine.
With original cooler it was hidden behind the noise of cooler but the Vardar fans from cooling loop don't make that high noise and I hear it always if I run some 3D task like Time Spy whie not wearing headset.

coil whine, oh yeah, the whine is strong in this one, but I also game with headsets, so noise while gaming is not an issue, what I want is quiet while working. Seems all reference 6800 XT have coil whine. Was thinking of doing some coil modding, but seems overkill right now.
Regarding pump, might have not made the ideal choice, but as long as it moves water and the temps are at least decent, I can live with it until I can get a stronger pump, if needed. Right now I have a Corsair H90 AIO on the 6800XT, thats only a 140mm rad, has absolutely no issues with cooling the gpu chip, but the Vram chips need better cooling.

 

[turbogear]

I think tinkered around a bit more with MPT (More Power Tool) and increased the total wattage output on the card from 300w to 340w, and managed to get it benchmarkable on the slider between 2690-2790mhz and outputting around 2740mhz whilst increasing the voltage to 1085mv. I did manage to then score 20900+ on TimeSpy (screenshot attached) but for some reason I had an issue with System Mark so couldn't validate the score. It wasn't stable though.

Then, put it back down to 2655-2755 on the slider and scored a 20650. This was the most stable I've managed to get it since.

All in all, I must have lucked out with the silicon lottery to pick up a card that decent on OC'ing, so I'm quite happy to have paid the extra £100 for it in hindsight.

Welcome to owners Club.
You are lucky guy and got a very good performing card on your hands.
Mine can most stably reach 2580-2612MHz with 19372 Time Spy score.

 

[xenosys]

Welcome to owners Club.
You are lucky guy and got a very good performing card on your hands.
Mine can most stably reach 2580-2612MHz with 19372 Time Spy score.

Thanks! Yeah, definitely lucked out.

I was getting a stable 2610-20mhz without using MPT, so a similar clock to yours.

However, I was only getting 18000-19000 when I first started tinkering around with the settings. Most people would be tempted just to max out all the sliders immediately, see what they score, and leave it there if it's stable. This card seems to benefit a lot from "undervolting" in Wattman. As soon as I started adjusting the core clock along with the voltage, the score just kept going up and up, until I hit 20k.

 

[turbogear]

Thanks! Yeah, definitely lucked out.

I was getting a stable 2610-20mhz without using MPT, so a similar clock to yours.

However, I was only getting 18000-19000 when I first started tinkering around with the settings. Most people would be tempted just to max out all the sliders immediately, see what they score, and leave it there if it's stable. This card seems to benefit a lot from "undervolting" in Wattman. As soon as I started adjusting the core clock along with the voltage, the score just kept going up and up, until I hit 20k.

Yes, it takes a lot of time to find the max performance.
Like you, I played with a lot of parameters and MPT.
In the beginning I was not getting that high and after many trials found best for me.
2650MHz@970mV with 275W for GPU and 305 TDC in MPT plus 15% power limit in Radeon Software gave me best.

Above 2650MHz setting in radeon nothing is very stable for me on long term.
I am able to set 2750MHz@1025mV (290W and 350TDC in MPT Plus 15% power limit) which gets me ~19750 in Time Spy but not really stable in games. With that settings I often got crashes in Assassin's creed valhalla, but 2650MHz@970mV is rock stable.

 

[xenosys]

Also have the reference form Power color, and mine is also a vey good overclocker, even with the stock cooler, can do 2.6 on stock air, and will see how much more it can on water. Temps become an issue after 2.6, hence the coming water loop.

Sounds like you might have a nice overclocker on your hands as well!

When you use the Radeon software presets to auto-overlock, what core clock number does it give you?

 

[Felix123BU]

Sounds like you might have a nice overclocker on your hands as well!

When you use the Radeon software presets to auto-overlock, what core clock number does it give you?

I never used the auto-oc presets, played around with it manually,, and max I can get game stable is set to 2750 max and 2650 min, stays around 2680 to 2710 in game 100% stable, but I don't use it on a day to day basis because on the stock cooler it gets to hot for my liking even with an aggressive fan curve. Undervolting is a must for these frequencies in my opinion, you can get it running with stock voltage, but the clocks wont stay that high without undervolting. Until the water loop components arrive, I keep it at 2550 max and 2450 min, which give me 2.5 in games.

 

[turbogear]

Sounds like you might have a nice overclocker on your hands as well!

When you use the Radeon software presets to auto-overlock, what core clock number does it give you?

I did not try the auto-overclock before.
I tried it right now and I get 2509MHz which is similar to yours.
I guess I need to investigate further.
Maybe there is a possibility to go higher on the overclock than I achieved so far.
I have to say that I did not try that high voltage when I was using 2750MHz.
I went to around 1030mV and I see that you go up to 1085mV.

 

[xenosys]

I did not try the auto-overclock before.
I tried it right now and I get 2509MHz which is similar to yours.
I guess I need to investigate further.
Maybe there is a possibility to go higher on the overclock than I achieved so far.
I have to say that I did not try that high voltage when I was using 2750MHz.
I went to around 1030mV and I see that you go up to 1085mV.

I'd definitely do it. Usually a 2500mhz+ on the auto-overclock recommendation is a sign of a good OC card. Keep in mind that I did need to use MPT to adjust the wattage of the card to 340w in order to support that clock speed.

 

[turbogear]

I'd definitely do it. Usually a 2500mhz+ on the auto-overclock recommendation is a sign of a good OC card. Keep in mind that I did need to use MPT to adjust the wattage of the card to 340w in order to support that clock speed.

I tried tuning it further after our discussion.
I was able to achieve 20121 score.

The settings are the similar to what I tried before and that got me before ~19750 score.
The difference is that I fiddled with VRAM now and found strangely that 2050MHz was causing issues before but it accepted 2100MHz without issues.Higher than 2100MHz caused benchmark crash.
Increasing the VRAM clock brought the extra performance which was missing to reach 20K score when I last tried similar setting.

I am not too confident though that this setting is fully stable. I did had one time benchmark crash today when I increased clock setting to 2755MHz from 2750MHz.
I was not able to stabilize it last time I tried similar settings.

I will try tomorrow to run stability tests. It is getting too late in Germany now.

 

[GamerGuy]

Keep 'em comin', guys! This thread is growin'!!!

 

[ratirt]

I will need to try OC but i will need some time which I don't have at the moment. I tried a bit with the voltage but it would seem, going lower than 1volt is out of the picture. Also I would need to purchase new PSU and there is absolutely nothing available. :/ Still waiting for some stuff to get to my place. I need to admit, that the Red Devil I got, stays relatively cool but I know there's room for improvement.

 

[xenosys]

I tried tuning it further after our discussion.
I was able to achieve 20121 score.

The settings are the similar to what I tried before and that got me before ~19750 score.
The difference is that I fiddled with VRAM now and found strangely that 2050MHz was causing issues before but it accepted 2100MHz without issues.Higher than 2100MHz caused benchmark crash.
Increasing the VRAM clock brought the extra performance which was missing to reach 20K score when I last tried similar setting.

I am not too confident though that this setting is fully stable. I did had one time benchmark crash today when I increased clock setting to 2755MHz from 2750MHz.
I was not able to stabilize it last time I tried similar settings.

I will try tomorrow to run stability tests. It is getting too late in Germany now.

Nicely done my friend. Welcome to the 20k club!

I was in a similar position with my memory settings. For some reason my card just doesn't like the memory at 2150mhz, whereas I've seen a lot of others maxing out that particular setting. My stable setting is anywhere between 2100-2120Mhz. Anything higher that 2130mhz will definitely produce a green screen of death. For every 10mhz overclock on memory, TimeSpy gives you an extra 20-30 points approx.

I'd run a Timespy GFX2 test on loop for about an hour to test stability.

I will need to try OC but i will need some time which I don't have at the moment. I tried a bit with the voltage but it would seem, going lower than 1volt is out of the picture. Also I would need to purchase new PSU and there is absolutely nothing available. :/ Still waiting for some stuff to get to my place. I need to admit, that the Red Devil I got, stays relatively cool but I know there's room for improvement.

With a Red Devil, it should be capable of doing a good overclock, especially if you keep the temps low. The Red Devil are some of the better binned cards.

 

[ratirt]

I tried tuning it further after our discussion.
I was able to achieve 20121 score.

The settings are the similar to what I tried before and that got me before ~19750 score.
The difference is that I fiddled with VRAM now and found strangely that 2050MHz was causing issues before but it accepted 2100MHz without issues.Higher than 2100MHz caused benchmark crash.
Increasing the VRAM clock brought the extra performance which was missing to reach 20K score when I last tried similar setting.

I am not too confident though that this setting is fully stable. I did had one time benchmark crash today when I increased clock setting to 2755MHz from 2750MHz.
I was not able to stabilize it last time I tried similar settings.

I will try tomorrow to run stability tests. It is getting too late in Germany now.

Damn. The 5800x CPU does the trick for your 6800XT.

 

[Kapone33]

I was having an issue over the weekend where after a long session my PC would just shut down. I thought that it was my Air cooler. Well I had the PSU (XPG Core Reactor 750) installed backwards so the PSU was getting no fresh air. I undervolted the GPU in Adreniline to 10% less. I saw my clocks running at a stable 2400 MHZ. My Timespy score was 18454 with a 5600x. For some reason 3dMark does not register my 4.7 GHZ OC.

 

[xenosys]

Here's my 6800XT validated result on TimeSpy that I completed a few moments ago.

Upped the wattage on MPT to 365w, upped the memory to 2130mhz, and adjusted the slider to 2685 - 2785mhz, then crossed my fingers and toes. It stuck around 2740mhz for most of the test. Ended up with 20930+ again. There's no way that will be stable in games though. Did a check on 3D Mark and it's the 12th highest GPU score across the board for a 6800XT. I'll take that!

I tried maxing everything out to try and get at least one 21000+ result but kept crashing on GFX Test 2. No surprises there!

I've also included my stable clock as well.

 

[turbogear]

Here's my 6800XT validated result on TimeSpy that I completed a few moments ago.

Upped the wattage on MPT to 365w, upped the memory to 2130mhz, and adjusted the slider to 2685 - 2785mhz, then crossed my fingers and toes. It stuck around 2740mhz for most of the test. Ended up with 20930+ again. There's no way that will be stable in games though. Did a check on 3D Mark and it's the 12th highest GPU score across the board for a 6800XT. I'll take that!

I tried maxing everything out to try and get at least one 21000+ result but kept crashing on GFX Test 2. No surprises there!

I've also included my stable clock as well.

Wow that's a great result.

With 365W on GPU core you are torturing your GPU VRM my friend.
According to my estimations you need to add around 45W for other components on GPU which means you were drawing around 410W total GPU power which is ~35W above 375W spec of 2x8pin and PCIe slot.

 

[xenosys]

Wow that's a great result.

With 365W on GPU core you are torturing your GPU VRM my friend.
According to my estimations you need to add around 45W for other components on GPU which means you were drawing around 410W total GPU power which is ~35W above 375W spec of 2x8pin and PCIe slot.

Oh yeah, I've since taken it right back down again, it was just for benchmarking purposes. I'll take it down to around 350-360w total draw tonight. Just out of curiosity, what does the other 45w consist of on the card?

 

[turbogear]

Oh yeah, I've since taken it right back down again, it was just for benchmarking purposes. I'll take it down to around 350-360w total draw tonight.

Yes I had 275W on core before plus 15% power limit which is around 361W total power that gave me before ~19300 score with 2650MHz@970mV setting.
That is still my target daily gaming setting.
As I now noticed that 2100MHz VRAM was stable.
I will add that to my above setting and see how high my score goes above 19300

I applied the VRAM 2100MHz overclock and score jumped from 19300 to 19670.


I tuned further the settings to squeeze out more performs in the total GPU power consumption range of 360W.
The max I could achieve with total GPU power of 361W at 310TDC was 19828 on Time Spy and 10158 on Port Royal.
This is with 2685MHz@980mV.

I got close to the 20K from before without going to the 290W GPU Core (total GPU power of 378W) like I did with previous experiment where score was higher than 20K.
This would be my new daily gaming settings as I don't like to drive the power constantly in 375W range.

I ran Fire Strike Extreme stability test with a score of 99.5% which can be increased to 99.8% by increasing the voltage to 990mV from 980mV but with that the Time Spy score goes down to 19789 instead of 19828.
I think I will leave it at 980mV and test games to see if this is stable everywhere.

 

[xenosys]

Yes I had 275W on core before plus 15% power limit which is around 361W total power that gave me before ~19300 score with 2650MHz@970mV setting.
That is still my target daily gaming setting.
As I now noticed that 2100MHz VRAM was stable.
I will add that to my above setting and see how high my score goes above 19300

I applied the VRAM 2100MHz overclock and score jumped from 19300 to 19670.
View attachment 188506

I tuned further the settings to squeeze out more performs in the total GPU power consumption range of 360W.
The max I could achieve with total GPU power of 361W at 310TDC was 19828 on Time Spy and 10158 on Port Royal.
This is with 2685MHz@980mV.

I got close to the 20K from before without going to the 290W GPU Core (total GPU power of 378W) like I did with previous experiment where score was higher than 20K.
This would be my new daily gaming settings as I don't like to drive the power constantly in 375W range.

I ran Fire Strike Extreme stability test with a score of 99.5% which can be increased to 99.8% by increasing the voltage to 990mV from 980mV but with that the Time Spy score goes down to 19789 instead of 19828.
I think I will leave it at 980mV and test games to see if this is stable everywhere.


View attachment 188508View attachment 188510
View attachment 188507View attachment 188509

Wow, nicely done!

Anything above 10k in TSE is very rare for a 6800XT, but you've topped it with a 158 points to spare. In fact it puts you 4th on the all time TSE list, according to 3D Mark.

I've noticed that when playing games, it doesn't draw nearly as many watts from the GPU itself that a Firestrike or Timespy stress test does. Sometimes it would hit 330w for me, but in games, it never usually goes above 300w but still maintains the same core clock rate.

 

[turbogear]

Wow, nicely done!

Anything above 10k in TSE is very rare for a 6800XT, but you've topped it with a 158 points to spare. In fact it puts you 4th on the all time TSE list, according to 3D Mark.

I've noticed that when playing games, it doesn't draw nearly as many watts from the GPU itself that a Firestrike or Timespy stress test does. Sometimes it would hit 330w for me, but in games, it never usually goes above 300w but still maintains the same core clock rate.

Thanks.
I will run it again today in the evening to check if with same settings the 10158 score is reproducible or one time surprise only.

Yes I also noticed that in game power consumption was lower.
In Assassin's creed valhalla the GPU Core consumption for me is usual in range from 220W to 250W. This was tested before on 2650MHz@970mV settings.
I will check with the new 2685@980mV in the next days and log with HWiNFO to get plots for averages and peaks.
I need to run some tests with Cyberpunk as well. I haven't played that since many weeks now.

 

[ratirt]

In Assassin's creed valhalla the GPU Core consumption for me is usual in range from 220W to 250W. This was tested before on 2650MHz@970mV settings.
I will check with the new 2685@980mV in the next days and log with HWiNFO to get plots for averages and peaks.

Dude. Your scores and OC is awesome. how did you manage to get 2685Mhz with only 980mv? Did you move the slider for power (15% up) for maximum as well?
I understand the mem is maxed out to 2150Mhz. Have you validated your OC and you have no regression in FPS? I know that people claim a lot when OC but the cards perform worse than expected in games.

 

[turbogear]

Dude. Your scores and OC is awesome. how did you manage to get 2685Mhz with only 980mv? Did you move the slider for power (15% up) for maximum as well?
I understand the mem is maxed out to 2150Mhz. Have you validated your OC and you have no regression in FPS? I know that people claim a lot when OC but the cards perform worse than expected in games.

I have 15% power slider plus in MPT GPU power set to 275W and TDC to 310A. So GPU is allowed to consume a total power of 361W compared to default power of 300W.
I am supplying 60W extra to GPU core and 10A more to TDC to go that high in frequency.

I don't have VRAM at 2150MHz. I can only go to 2100MHz. Higher is unstable.

As I explained last time, 980mV is a setting that is used in the Radeon software but that is not the actual voltage that GPU uses when boosting.
You can't reach 2600MHz with only 980mV real GPX voltage. 980mV is a setting in Radeon that does not correspond directly to voltage used to boost.
The 980mV is only adjusting the boost curve so that towards 1150mV real voltage the GPU can boost much higher than default.

Looking at HWINFO, the actual voltage is going towards 1150mV when GPU boosts towards 2620-2630MHz.
You can't influence the actual voltage that GPU uses to boost through Radeon Setting. It is going higher as far as GPU needs it but limit is 1150mV.
The only way to limit real maximum voltage allowed during boosting is by using MPT. If I put hard limit in MPT to lower than 1130mV the GPU crashes when boosting in direction of 2600MHz.

For 2650MHz@970mV, I did gaming check and FPS was higher.
For this setting, in games GPU held 2580-2612MHz boost range.
The performance was higher.

For 2685MHz@980mV as I mentioned before, I will do validations in the next day by logging data and comparison to default setting performance. This new setting is work in progress.

I am thinking of installing NVidia FrameView utility to analyse the FPS in more scientific way than looking at the counter during games.
My only concern is if it will play nicely with AMD hardware.

I use Rivatuner statistics server to see frames while gaming. HWINFO can log this but for some reason it does not log all time. There are many empty lines left.

If anybody here knows which utility I can best use to log frames please let me know. I like to make diagrams like the reviewers are doing.

 

[Robert Bourgoin]

I got an AMD XFX Radeon X6800, strictly stock. Love it. Was part of my new build in my signature.
I paid too much for it on Amazon but I wanted it for my build. Already a collectors item. LOL
Part of my nicest build. This system has been flawless from day one.

 

[turbogear]

Dude. Your scores and OC is awesome. how did you manage to get 2685Mhz with only 980mv? Did you move the slider for power (15% up) for maximum as well?
I understand the mem is maxed out to 2150Mhz. Have you validated your OC and you have no regression in FPS? I know that people claim a lot when OC but the cards perform worse than expected in games.

I did some data analysis with the 2685MHz@985mV overclock.
You can see the performance is uplifted by OC.
In the Assassin's Creed Valhalla and Borderland benchmarks the increase is around 7% in performance.

Here are the plots comparing Superposition run at default and OC. I have also done Borderland 3 built in Benchmark and compared the default to my OC.
The plots compare the GPU frequency and the FPS data from Riva Tuner Server.

I discovered a small tool called GenericLogViewer that can help to compare HWINFO log files.
LogViewer for HWINFO is available ! | HWiNFO Forum

Borderland 3 benchmark:
The gain is around 7% on average FPS.




Superposition benchmark:


Results from Assassin's Creed Valhalla.
The gain is around 7% on average FPS.

With the 2685MHz@985mV:


With default GPU settings:

 

[Felix123BU]

OK, so the 6800 XT, and whole system is under water
As expected, made some rookie water build mistakes, as in did not think of how the tubes would go, hence had to do some getto routing, but all good for a start.
The gpu cooling is insane with this setup, even after 1 hour furmark with the hottest setting I could figure out, it never went above 60 and 80 on the hotspot. General gaming is like max 50 and 70 on hotspot, the more demanding games.

Tried a 2.75ghz OC, it was running, but I am 99.9% sure it had not enough power for that clock. It needs moooar powaah
Any of you have experience with the More Power Tool, as in, can you increase available power, and if yes, does it actually allow the card to draw more?
I had a nice experience long time ago with a rx 480, modded it and let is use 300w, basically double what it should have, I can only dream of what the 6800 XT could do with lets say a 450w limit