The RX 6000 series Owners' Club

[AusWolf]

I'm sure it is, as my 5700xt kicked ass with r5 3600 in 1080p with 100+ fps in almost everything.

Tbh, seeing how expensive the higher-end gpus got in the latest generations I have no good expectations. Inflation did its thing emptying our pockets as well. And since the gpu cryptomining is dead, the fine priced oportunities for a highest-end gpus will be gone really soon. I'm tempted to buy another oc formula from the same seller, for less money than the first one. Silly investment in a future we might never live to see anyway.

That's a good point. So are you saying that a 620 GBP 6950 XT might be a good investment for the future (that is, performance-wise, not money-wise)?

I'm also wondering because it's only 30-50% faster than my 6750 XT. Part of me is doubting if it's worth spending hundreds on, instead of just buying another mid-range card a few years later.

 

[Veseleil]

That's a good point. So are you saying that a 620 GBP 6950 XT might be a good investment for the future (that is, performance-wise, not money-wise)?

I'm also wondering because it's only 30-50% faster than my 6750 XT. Part of me is doubting if it's worth spending hundreds on, instead of just buying another mid-range card a few years later.

I think both camps murdered their midrange cards performance in the current generation. Heavily cutting the cores, bandwidth and the rest. There is hope they won't in the upcoming ones, but personally I won't be buying midrange cards anymore, it's not worth it imo. Time has proven that high-end ones last for much longer period, performance wise. Now let's hope they will physically last long as the older ones (1080ti, I'm looking at you, many heavily mined on for years are still playing games for example).
I'm not posting this to pursuade or such, just sharing my thoughts and experience. Decision is yours only after all.

 

[AusWolf]

I think both camps murdered their midrange cards performance in the current generation. Heavily cutting the cores, bandwidth and the rest. There is hope they won't in the upcoming ones, but personally I won't be buying midrange cards anymore, it's not worth it imo. Time has proven that high-end ones last for much longer period, performance wise. Now let's hope they will physically last long as the older ones (1080ti, I'm looking at you, many heavily mined on for years are still playing games for example).
I'm not posting this to pursuade or such, just sharing my thoughts and experience. Decision is yours only after all.

All fair points. Mid-range is basically dead, that's for sure, but the high-end is just a bit too high for my liking. Question is, is a roughly 50% performance bump worth 600 quid? What would you do?

 

[Veseleil]

IDK really, because we kinda lose doing either. It's a nicely planned obsolescence tactic. Buying high end vs. mid range. Less upgrades vs. upgrading almost each gen. Still left to see are the flagship cards in the last few gens built to last, or not. And when we found out, they might become obsolete.
I've already answered that, kind of. Many regrets for going cheap, almost none going enthusiastic, so far.

 

[GamerGuy]

IDK really, because we kinda lose doing either. It's a nicely planned obsolescence tactic. Buying high end vs. mid range. Less upgrades vs. upgrading almost each gen. Still left to see are the flagship cards in the last few gens built to last, or not. And when we found out, they might become obsolete.
I've already answered that, kind of. Many regrets for going cheap, almost none going enthusiastic, so far.

I agree, hence my predilection for higher end cards myself. Bought a Sapphire Nitro+ RX 6900XT back at around launch time, and while it's been replaced by a MERC310 RX 7900 XTX, it's still a kickbutt card, built a 2nd rig (mostly spare parts) around it and it's still quite the rasterize gaming performer at 3440x1440 75Hz Freesync (Acer XR341CK). Played DSR with FSR2 at max ingame and it's been great (other than those $%#@ traversal hitching).

 

[tpa-pr]

Just putting out a call for advice in case anyone has experience: has anyone tried enabling undervolting via Adrenalin on a 6500M? I've tried a couple of different methods to unlock the sliders since they're locked by default:
1) MSI Afterburner - no change.
2) MorePowerTool - causes BSOD no matter what settings are toggled.

I'd like to avoid flashing custom BIOSes for example (and I'm fairly certain that because this is a HP OEM version, it probably can't be done), so if anyone has any other ideas let me know. This isn't a critical thing, it's just for fun.

 

[Shou Miko]

Just putting out a call for advice in case anyone has experience: has anyone tried enabling undervolting via Adrenalin on a 6500M? I've tried a couple of different methods to unlock the sliders since they're locked by default:
1) MSI Afterburner - no change.
2) MorePowerTool - causes BSOD no matter what settings are toggled.

I'd like to avoid flashing custom BIOSes for example (and I'm fairly certain that because this is a HP OEM version, it probably can't be done), so if anyone has any other ideas let me know. This isn't a critical thing, it's just for fun.

View attachment 314360

If the power setting are available in the Adrenalin software by AMD then try to lower the voltage with 0.010V at a time it takes time but then you can see if you have won the silicon lottery if the card refuses to run it might run at the lowest voltage it can.

You never specified which voltage you tried with MorePowerTool

 

[tpa-pr]

If the power setting are available in the Adrenalin software by AMD then try to lower the voltage with 0.010V at a time it takes time but then you can see if you have won the silicon lottery if the card refuses to run it might run at the lowest voltage it can.

You never specified which voltage you tried with MorePowerTool

Oh I was using MPT to try and unlock the voltage options! This is what my Adrenalin looks like at stock:

I was hoping someone might have some ideas as to how to unlock the sliders or if it's possible.

 

[AMF]

Oh I was using MPT to try and unlock the voltage options! This is what my Adrenalin looks like at stock:
View attachment 314476
I was hoping someone might have some ideas as to how to unlock the sliders or if it's possible.

this is minimum install look make sure u have full install of adrenalin

 

[tpa-pr]

this is minimum install look make sure u have full install of adenalin

I definitely have the full install straight from AMD's website. That's the full Adrenalin app, not the one from the Microsoft Store.

 

[chomuco]

Hitting those magical 3GHz in Overlay with a 6600XT. The fastest card to air, although not many people submit scores.
[ADJUNTAR]295039[/ADJUNTAR]

:rockear:
Technology.
https://postimg.cc/xNTzk9d0 WOW!
Unequal

 

[tpa-pr]

Reporting back in from the undervolt mines for my 6950 XT, I am officially throwing in the towel Try as I might I can't get my card to keep a stable undervolt, I always have at least one game crashing no matter the values I set (including the AMD auto-undervolt values).

Oh well, i'll just crank the default fan curve to keep the chip cool, I don't like it running too hot. Seems like I might've lost the silicon lottery.

 

[Lew Zealand]

Reporting back in from the undervolt mines for my 6950 XT, I am officially throwing in the towel Try as I might I can't get my card to keep a stable undervolt, I always have at least one game crashing no matter the values I set (including the AMD auto-undervolt values).

Oh well, i'll just crank the default fan curve to keep the chip cool, I don't like it running too hot. Seems like I might've lost the silicon lottery.

Sorry to hear that. One thing I noticed when doing a bunch of voltage and load testing on my 6800 XT is that the card is stable at significant undervolts at high speeds and powers with DX12 loads but if a lower demand is put on it, it's not stable at those settings. Crank up the voltage about 25 to 50mV and it will be stable in lower demand DX11 games at VSync with far less than 100% GPU use. So sometimes I'll fiddle with the voltage depending on the game I'm playing which I don't mind doing as IMO that's a fun meta-game.

But if I wanted to be stable in all use cases I'd need to be at 1.087V and maybe 1.093V just to be 100% sure. Yet in high-demand games like CP2077, I can run 1.037V at 2550 MHz core clock and it's happy. And with ~300W GPUs, these voltage differences can lead to 5-10% power savings, lower temps, cooler feet near the PC case.

 

[tpa-pr]

Sorry to hear that. One thing I noticed when doing a bunch of voltage and load testing on my 6800 XT is that the card is stable at significant undervolts at high speeds and powers with DX12 loads but if a lower demand is put on it, it's not stable at those settings. Crank up the voltage about 25 to 50mV and it will be stable in lower demand DX11 games at VSync with far less than 100% GPU use. So sometimes I'll fiddle with the voltage depending on the game I'm playing which I don't mind doing as IMO that's a fun meta-game.

But if I wanted to be stable in all use cases I'd need to be at 1.087V and maybe 1.093V just to be 100% sure. Yet in high-demand games like CP2077, I can run 1.037V at 2550 MHz core clock and it's happy. And with ~300W GPUs, these voltage differences can lead to 5-10% power savings, lower temps, cooler feet near the PC case.

Just to double check: when you're referring to your voltage there, are you referring to the slider in Adrenalin or the actual voltages (because I'm to understand that Adrenalin is applying voltages as an offset)?

I mean, when undervolting the only game that crashes is Call of the Wild. If I set a mild undervolt like 1175mV on the slider it works fine in everything else (including temperamental Time Spy). But from my understanding an undervolt isn't stable until everything works without crashing.

 

[Lew Zealand]

Just to double check: when you're referring to your voltage there, are you referring to the slider in Adrenalin or the actual voltages (because I'm to understand that Adrenalin is applying voltages as an offset)?

I mean, when undervolting the only game that crashes is Call of the Wild. If I set a mild undervolt like 1175mV on the slider it works fine in everything else (including temperamental Time Spy). But from my understanding an undervolt isn't stable until everything works without crashing.

The voltage (offset) as set by slider. I realize that it bumps up the target voltage if you increase the core clock sider and vice versa if you reduce clocks.

I get the impression that the curve it's sliding along in the background is a little too aggressively undervolted for when there's a low load and while I do like the ease with which Adrenaline manages this, I wish I could select an alternate curve with a shallower dropoff to keep the voltages higher when load is reduced.

This is an advantage to how Nvidia cards are undervolted in Afterburner as you can set this manually though it's way more tedious/tiresome and frequently not exact when setting this (when the curve shifts as I apply and save, there is much yelling). But you can at least manually set the curve to keep voltages a little higher to avoid low-GPU-demand crashes. I found this the hard way after repeatedly crashing out when undervolting my 1080 and finally realizing that all I needed to do was keep the lower core speed voltages a little higher when overclocking/undervolting than the default curve suggests. Yeah, it uses fractions more power but it's already using less power at low demand, no need to reduce more and cause instability.

 

[tpa-pr]

The voltage (offset) as set by slider. I realize that it bumps up the target voltage if you increase the core clock sider and vice versa if you reduce clocks.

I get the impression that the curve it's sliding along in the background is a little too aggressively undervolted for when there's a low load and while I do like the ease with which Adrenaline manages this, I wish I could select an alternate curve with a shallower dropoff to keep the voltages higher when load is reduced.

This is an advantage to how Nvidia cards are undervolted in Afterburner as you can set this manually though it's way more tedious/tiresome and frequently not exact when setting this (when the curve shifts as I apply and save, there is much yelling). But you can at least manually set the curve to keep voltages a little higher to avoid low-GPU-demand crashes. I found this the hard way after repeatedly crashing out when undervolting my 1080 and finally realizing that all I needed to do was keep the lower core speed voltages a little higher when overclocking/undervolting than the default curve suggests. Yeah, it uses fractions more power but it's already using less power at low demand, no need to reduce more and cause instability.

OK, I can't believe I'm saying this, but I might've solved my instability by sheer luck. The entire time I've been trying to undervolt I've been following the rule of "min clocks should be max clocks -100". Today I got home from work and started with default clocks, then started playing with power levels and voltages. I got my first crash at 1150mV so I shifted it to 1155mV and for the next 2.5 hours of testing, Time Spy did not crash once. The only time I was able to replicate the crashes I had before was tinkering with the min voltage slider! Move it up at all, crash, don't ask me why.

Anyway, pic related is the "how far can I push this thing" score:

And now my current settings:

GPU Tuning: Enabled.
- Advanced Control: Enabled.
- Min Frequency: 500Mhz
- Max Frequency: 2600Mhz
- Voltage: 1155mV
Fan Tuning: Enabled.
- Zero RPM: On
- Advanced Control: On
- Temp: 50 (Speed: 35)
- Temp: 60 (Speed: 40)
- Temp: 70 (Speed: 50)
- Temp: 80 (Speed: 65)
- Temp 90 (Speed 75)
Power Tuning: Disabled.
- Power Limit (%): 0
VRAM Tuning: Enabled.
- Memory Timing: Default
- Advanced Control: Enabled
- Max Frequency: 2400Mhz

End result of this is performance on parity in Cyberpunk 2077 at stock with almost 10-15C drop, maybe more! I'm now testing my white whale game (Call of the Wild) to see how it goes. If it crashes, I might just keep it on stock settings. If it keeps passing Time Spy I consider that "stable". I have no idea why the min voltage slider would cause instability but it does!

Edit: And that's a crash in Call of the Wild. So Call of the Wild can have stock settings, everything else can run the undervolt.

 

[AusWolf]

Sorry to hear that. One thing I noticed when doing a bunch of voltage and load testing on my 6800 XT is that the card is stable at significant undervolts at high speeds and powers with DX12 loads but if a lower demand is put on it, it's not stable at those settings.

At high load, the card tries to stay stable by supplying lower clocks and voltages than in low load. This might do at factory default, or with a modified power limit, but may fail with an OC or UV.

 

[tpa-pr]

OK, I can't believe I'm saying this, but I might've solved my instability by sheer luck. The entire time I've been trying to undervolt I've been following the rule of "min clocks should be max clocks -100". Today I got home from work and started with default clocks, then started playing with power levels and voltages. I got my first crash at 1150mV so I shifted it to 1155mV and for the next 2.5 hours of testing, Time Spy did not crash once. The only time I was able to replicate the crashes I had before was tinkering with the min voltage slider! Move it up at all, crash, don't ask me why.

Anyway, pic related is the "how far can I push this thing" score:
View attachment 315267

And now my current settings:

End result of this is performance on parity in Cyberpunk 2077 at stock with almost 10-15C drop, maybe more! I'm now testing my white whale game (Call of the Wild) to see how it goes. If it crashes, I might just keep it on stock settings. If it keeps passing Time Spy I consider that "stable". I have no idea why the min voltage slider would cause instability but it does!

Edit: And that's a crash in Call of the Wild. So Call of the Wild can have stock settings, everything else can run the undervolt.

Final follow-up on this: I am a fool. Call of the Wild crashes on the problematic map on stock and I never realised. So that's about two months of testing based on faulty data!

The good news is the settings I've posted above are rock solid. The card is so much quieter and cooler now while maintaining stock performance. And if I want I have the wiggle room to push the card harder. Overall I am VERY happy with the 6950 XT.

 

[Lew Zealand]

Final follow-up on this: I am a fool. Call of the Wild crashes on the problematic map on stock and I never realised. So that's about two months of testing based on faulty data!

The good news is the settings I've posted above are rock solid. The card is so much quieter and cooler now while maintaining stock performance. And if I want I have the wiggle room to push the card harder. Overall I am VERY happy with the 6950 XT.

That's great news and I should have realized that earlier. I have many times been led astray thinking I had an unstable undervolt when instead the test was broken and a reinstall fixed it, Unigine Valley and TimeSpy most recently.

Second most recently. Today I installed the CP2077 2.0 update and it worked great for a while and now it crashes at the same point on all GPU settings. I thought I'd undervolted too much and backed off and it still crashes. Backed off to stock and it still crashes. CP2077 reinstall time...

 

[tpa-pr]

I got the 1150mV undervolt to stick I'm learning more as I go on, I think I was pushing my memory a bit too hard and causing instability.

 

[s0meoNE666]

Dear Owners,

I have bought a used Asrock RX 6800 non XT Phantom Gaming 16G OC and i want to chnage all of the thermal pads to GELID GP-Extreme or Ultimate but i dont know the size of them. Is there anyone who has the same card and know the thickness of the thermal pads for sure.
Thank you the help.

BR

 

[Beginner Macro Device]

End result of this is performance on parity in Cyberpunk 2077 at stock with almost 10-15C drop, maybe more!

How much wattage saved?

 

[Veseleil]

Played a bit of RDR2 yesterday. The Radeon settings were set as usual:



After finishing intro part of the game, I've checked temps on HWInfo, And whoops, the hotspot reached over 100C, the rest was around 70. Mind that everything is on Ultra/High, 1440p, uncapped FPS, Native Res/no upscaling. It's around 100 most of the time So I ramped the GPU fan curve up, tested lowering the Freq, as it scales almost perfectly with voltage. Ambient was almost 30C.
So I had almost immediate temp drops as it also scales with the mentioned, it's unreal. 2400MHz -> 900mV -> 80C (from 90+ mind you), 2300MHz -> 800mV -> 70C: I'm talking about hotspot temps, the die sensors numbers are scaling with the rest as well, so ~10C increment drops. FPS also scales with the rest, 3-7FPS loss at each -100MHz, and I'll stop here with the damn scaling, will haunt my dreams. Thing is, the recent 1440p monitor upgrade made this GPU profile kinda obsolete, as this card is pushing FPS like crazy anyway (from 1080p expected to scale accordingly, but no, it thrives even more in 1440p).

It was like 65C GPU, 75C hotspot on that setting in 1080p before, and the fans were quiet (below 1200RPM on moderate load, and full 1800RPM on high FPS, in demanding visuals scenario. P12 PWM's (PST) are not annoyingly loud at max, and I have a good hearing (can't miss any high freq noise like high pitch), on a lower power card they would be more than adequate.

I was lazy to take screenshots, but here I am, thinking about ordering/ installing 3x Phanteks T30's on max, to test it near 3GHz, and ultimately watercooling it down to ambient (and hopefully below with the outdoor rad), unlocking that sweet Navi 21 XTXH die potential, that ASRock provided with OC Formula. I will probably kill this card eventually, it can pull over 450W without a single sweat but still, it's rescued from the mines of Moria, she's enjoying her extreme sports pension when not sleeping for weeks. Just my 13,37 cents.
Seems at first glance that the only limit of this GPU is cooling, as it is with all the OC oriented beasts of the past decade and beyond.
But I missed the opportunity to rescue another identical card, for 30% less money than the one I own, from the same guy's ex mining equipment big sale. He had maybe dozen of these, smh... I would buy all for the science purposes, but I lack whale like funds. It was (and still is) their crypto game, we're indeed appear like zombies to their white crystal eyes. Remember these were $1300+ at launch:

 

[Pouhon]

After MX-4 pumping out quickly and temps becoming worse than the stock paste (90C core 110C hotspot with 2500 rpm fan instantly) just in 2 weeks, i bought PTM7950 to solve the issue at once. (Also it wasn't too expensive) I applied it in last October, and since then temps are virtually the same, no sign of pump out. Ignore HWINFO reporting it as Merc 319, it's an inferior Swft 319, but i think these readings are very respectable for this model. Maybe i could've got away with a paste more suitable for direct die application (like MX-6, SYY-157 etc.) I'm sure MX-4 is not a fake because it's on my CPU for nearly 3 years and no problems.

System 4 = Intake (3x), PUMP = Exhaust (2x) (more info on my previous post on last page)

Stock fan curve, its very relaxed and even then fans doesn't end up spinning too fast

Changed the curve to be more reactive, higher rpm but not too much

 

[Kodehawa]

After MX-4 pumping out quickly and temps becoming worse than the stock paste (90C core 110C hotspot with 2500 rpm fan instantly) just in 2 weeks, i bought PTM7950 to solve the issue at once. (Also it wasn't too expensive) I applied it in last October, and since then temps are virtually the same, no sign of pump out. Ignore HWINFO reporting it as Merc 319, it's an inferior Swft 319, but i think these readings are very respectable for this model. Maybe i could've got away with a paste more suitable for direct die application (like MX-6, SYY-157 etc.) I'm sure MX-4 is not a fake because it's on my CPU for nearly 3 years and no problems.

System 4 = Intake (3x), PUMP = Exhaust (2x) (more info on my previous post on last page)
View attachment 332169

Stock fan curve, its very relaxed and even then fans doesn't end up spinning too fast

View attachment 332170

Changed the curve to be more reactive, higher rpm but not too much

Been rocking 7950 on my 6700 for a good while and temps have been even better than in the first application here.
Pretty impressive, specially considering I got my hotspot down 20C.