The RX 6000 series Owners' Club

[Veseleil]

After a bit of testing, now that I have the cooling headroom for an OC, I may say my card is stable at 2860MHz core and at 2180MHz VRAM Adrenaline settings.
Note that I don't want to push my luck with VRAM, so I returned it to default, and plan to keep it that way. This was a mining card, VRAM is stressed enough already. The performance uplifts of VRAM OC are negligible anyway, and are often cause for all sorts of weird instability issues.



That was captured on a second display while the RDR2 occupied the main one.

All in all, considering nothing was configured outside the Adrenaline, no MPT, stock heatsink and an air cooling at around 25C ambient, I'm happy as a puppy.

 

[KedsDead]

Nice, I have a reference card, so my options do not go near that high. So I tuned for performance and temps. Not that your temps are bad, no where near that. Runs cooler than mine.

I would still oc your memory, just don't push it high. I can slide the bar fully to the right, and it seems stable, have not had a crashing issue, but I just set it to 2112, so it runs at an even 2100.

 

[Lew Zealand]

Nice, I have a reference card, so my options do not go near that high. So I tuned for performance and temps. Not that your temps are bad, no where near that. Runs cooler than mine.

I would still oc your memory, just don't push it high. I can slide the bar fully to the right, and it seems stable, have not had a crashing issue, but I just set it to 2112, so it runs at an even 2100.

I do the same on my 6800 XT. I haven't gotten it to crash at 2132 (2120 real MHz) but I clock it down a bit to 2112 just to ensure stability. However I understand Veseleil's concern for his VRAM with an ex-mining card as I have a used 5600 XT which was likely a mining card as well. It works great and it's rarely hitting 100% GPU usage so keeping it at stock VRAM speed will stress the already well-used VRAM a bit less.

 

[Veseleil]

I understand Veseleil's concern for his VRAM with an ex-mining card as I have a used 5600 XT which was likely a mining card as well. It works great and it's rarely hitting 100% GPU usage so keeping it at stock VRAM speed will stress the already well-used VRAM a bit less.

Couldn't say it better myself.

Nice, I have a reference card, so my options do not go near that high.

OC Formula is an XTXH chip, which in the reference model 6900XT LC is combined with a 18Gbps memory (2250MHz). The AIB 6900XTXH models have a 16Gbps memory (2000MHz) instead, just as any other AIB 6900XT or 6900XTX model. Yes, there are 3 variants. The reference 6900XTX model (never released) is the XTXH with a 18Gbps memory, just like the reference LC one.

Apart from a higher power limit than the non-XTXH (330W vs. 300W), the VRAM slider is also unlocked to 2624MHz. My card is unstable at 2250MHz, but it can run over 2200MHz. Not really stress tested, since I don't want to push my luck, as stated earlier.

The difference between regular (non-binned) 6900XT and 6950XT models are simple: 16Gbps vs. 18Gbps memory, and 300W vs. 335W power limit. Also a bit higher declared boost clocks (base, game, boost).
If you're interested, for starters you can read more about it here.

 

[KedsDead]

Didnt know all that. The only reason I have this 6900 is MicroCenter had an openbox (had no box, only came in a static bag) that they would give me a full warrantee on and priced it $35 cheaper than the 6800XT I was there to buy.

 

[AVATARAT]

The difference between regular (non-binned) 6900XT and 6950XT models are simple: 16Gbps vs. 18Gbps memory, and 300W vs. 335W power limit. Also a bit higher declared boost clocks (base, game, boost).
If you're interested, for starters you can read more about it here.

The difference between 6900 and 6950 is really big, you can read here

 

[Veseleil]

The difference between 6900 and 6950 is really big, you can read here

Yeah, the difference between HD 6950 and RX 6900XT is huge alright.

 

[AVATARAT]

Oh, yeah I mistake the links

 

[Nordic]

A few weeks ago I re-pasted my 6750xt from stock to MX-4 and saw a mild reduction in temperature. I have some PTM7950 coming. Every review I have read says it performs better than the kryosheet, is not conductive, and never pumps out. I will definitely post results when I have it installed.

It took a lot longer than expected for the PTM7950 to arrive. I purchased from ModDIY.
After I installed PTM7950 I was getting worse significantly worse hotspot temps than MX-4.
I disassembled and saw that 1/4 of the die was not making contact. I assumed user error. I must have done something wrong.
I reapplied, this time trying to take extra efforts to make perfect contact. I could visibly see space between the die and the heatsink.
I removed all the PTM7950 and reapplied MX-4. I had to apply a lot of MX-4 to ensure contact was made.
I tested with OCCT to let it increase GPU load in increments of 5% starting from 10%. At 55% it would overheat and shut down the PC.
I disassembled once again and checked that contact was being made.

It is as if something has physically changed somewhere and there is now more space between the GPU die and heatsink. I removed all the thermal pads to see if they might be causing it somehow. I made sure the raised screw holes in the heatsink were aligned and going through the holes in the GPU PCB.

I can't figure out what changed if anything.

 

[kapone32]

It took a lot longer than expected for the PTM7950 to arrive. I purchased from ModDIY.
After I installed PTM7950 I was getting worse significantly worse hotspot temps than MX-4.
I disassembled and saw that 1/4 of the die was not making contact. I assumed user error. I must have done something wrong.
I reapplied, this time trying to take extra efforts to make perfect contact. I could visibly see space between the die and the heatsink.
I removed all the PTM7950 and reapplied MX-4. I had to apply a lot of MX-4 to ensure contact was made.
I tested with OCCT to let it increase GPU load in increments of 5% starting from 10%. At 55% it would overheat and shut down the PC.
I disassembled once again and checked that contact was being made.

It is as if something has physically changed somewhere and there is now more space between the GPU die and heatsink. I removed all the thermal pads to see if they might be causing it somehow. I made sure the raised screw holes in the heatsink were aligned and going through the holes in the GPU PCB.

I can't figure out what changed if anything.

The height of the pad may be compromising the heat pads on the Memory or VRM. They may be too shallow to make contact with the heatsink.

 

[Nordic]

The height of the pad may be compromising the heat pads on the Memory or VRM. They may be too shallow to make contact with the heatsink.

With MX-4 paste on the die and no pads on the VRM or memory, there is approximately a full mm of space or more between the GPU die and heatsink.

 

[RJARRRPCGP]

The performance uplifts of VRAM OC are negligible anyway, and are often cause for all sorts of weird instability issues.

Yeah, that's true with Radeon. Signs of VRAM instability: Jerkiness, can look like packet loss and possibly WHEA for "Cache Hierarchy Error" as reason.

 

[Nordic]

It took a lot longer than expected for the PTM7950 to arrive. I purchased from ModDIY.
After I installed PTM7950 I was getting worse significantly worse hotspot temps than MX-4.
I disassembled and saw that 1/4 of the die was not making contact. I assumed user error. I must have done something wrong.
I reapplied, this time trying to take extra efforts to make perfect contact. I could visibly see space between the die and the heatsink.
I removed all the PTM7950 and reapplied MX-4. I had to apply a lot of MX-4 to ensure contact was made.
I tested with OCCT to let it increase GPU load in increments of 5% starting from 10%. At 55% it would overheat and shut down the PC.
I disassembled once again and checked that contact was being made.

It is as if something has physically changed somewhere and there is now more space between the GPU die and heatsink. I removed all the thermal pads to see if they might be causing it somehow. I made sure the raised screw holes in the heatsink were aligned and going through the holes in the GPU PCB.

I can't figure out what changed if anything.

I forgot to come back and mention that I resolved this issue. It appears that I had somehow bent one of the standoffs. I removed the standoffs and now have good thermal performance again.

 

[itzFizzer]

Hey, new here!

Requesting entry to the club, XFX QICK RX 6750 XT here...Had this thing a couple of months now and absolutely love it!

Smashes 1080p, brilliant at 1440p and totally capable of 4K with a bit of FSR magic!

Running a 1120mV under volt with 2700 overclock and vram maxed at 2312 (fast timings) & power limit at 15%

 

[GamerGuy]

Hey, new here!

Requesting entry to the club, XFX QICK RX 6750 XT here...Had this thing a couple of months now and absolutely love it!

Welcome! The gold plated membership card is at the entrance, help yourself to the caviar and champagne just inside at the foyer.

 

[itzFizzer]

Welcome! The gold plated membership card is at the entrance, help yourself to the caviar and champagne just inside at the foyer.

Thank you!!

I've been a long time lurker of the forum and site in general, so I thought it was about time I joined!

The champagne is bottomless, I take it?

 

[KernelPanixc]

6800XT's memory is cooking itself 3 min of CS2 gameplay. In amd software i set the memory at 2000mhz max and this is changed nothing on memory temp

i run CS2 at 4k virtual resulation

 

[Jism]

30 degree difference in between hotspot vs GPU core temperature screams for a proper repaste.

I think i reached 12 degree difference after repasting with MX-5. So core 50 hotspot 62.

 

[chris189]

Just request from me to mod the SoftPowerPlayTable & Limit the Temperatures across the board to 84C so as not to damage your precious card?

 

[AusWolf]

6800XT's memory is cooking itself 3 min of CS2 gameplay. In amd software i set the memory at 2000mhz max and this is changed nothing on memory temp

i run CS2 at 4k virtual resulation

View attachment 357128View attachment 357095

What do you mean "the memory is cooking itself"? I don't see your memory temperature anywhere.

 

[itzFizzer]

I have noticed that in Adrenalin that sometimes when the settings are stock before I adjust that sometimes the max frequency is 2609 and other times it's 2614. Is there a reason why it's different sometimes?

I have an XFX 6750 XT...Also curious to know what UV/OC settings you guys run with!

 

[AVATARAT]

I have noticed that in Adrenalin that sometimes when the settings are stock before I adjust that sometimes the max frequency is 2609 and other times it's 2614. Is there a reason why it's different sometimes?

I have an XFX 6750 XT...Also curious to know what UV/OC settings you guys run with!

5MHz is just in error margin.
Just push Undervolt button in Performance tab and don't worry.

 

[KernelPanixc]

What do you mean "the memory is cooking itself"? I don't see your memory temperature anywhere.

Sorry i was tought "hot spot" stands for VRAM temperature. This is my brothers GPU its not repasted since 3 years so this is normal for this condition. I'll repaste it.

 

[Nordic]

85c hotspot is beyond fine. You are not cooking anything.

 

[AusWolf]

85c hotspot is beyond fine. You are not cooking anything.

This. Don't bother repasting.