Overclock.net banner
7041 - 7060 of 7233 Posts

·
Registered
Joined
·
119 Posts
Yup, that hitch - happens to me, too...I checked the Superposition 4K results at Unigine's site a while back, and the hitch either happens at scene 10, or 13/14 (my 6900Xt 'likes' the former, while the 3090 'prefers' the latter :p ).
Okay, so it's normal. Thanks!
 

·
Registered
Joined
·
49 Posts
NP. I won the shi*t lottery with this card. It doesn't OC well and won't even take relatively tame MPT values. And the card doesn't even have the decency to die properly so I could RMA it. It works without issues on default.
Tried the tdp and the soc your having issues with and no shut downs or temp issues. not sure what power limit you had though.
 

Attachments

·
Invalid Media
Joined
·
6,438 Posts
Tried the tdp and the soc your having issues with and no shut downs or temp issues. not sure what power limit you had though.
NP. I won the shi*t lottery with this card. It doesn't OC well and won't even take relatively tame MPT values. And the card doesn't even have the decency to die properly so I could RMA it. It works without issues on default.
...It's a tough call to make re. warranty rules in your jurisdiction, but sometimes, cards aren't screwed together perfectly to begin with though function well for a while. I would arm up with good, thick thermal paste and decent correctly-sized thermal pads (or thermal putty for that matter) and take the card apart for a repaste job.

What might have happened (pure speculation on my part) is that during earlier MPT-PL runs, paste might have dried out or an incorrectly seated thermal pad exposed a component to more heat (or indeed a deteriorating component on the PCB, such as an input filter). These cards are full of various sensors for the boost algorithms, and it obviously is getting a signal from somewhere that beyond a certain point, there are issues...PL is expressed in Watts which is a measure of heat energy...
 

·
Registered
Joined
·
134 Posts
...I told my little XTX that if it does not hit at least 19.1K in Superposition 4K, I'll just get the Sapphire Nitro+ Pure 6950XT instead...seems to have worked 😁
MPT 450W/400A, down-clocking and down-volting enabled, DF C-State enabled.
BTW, the preview driver gives a significant boost to Superposition, I went from ~18500 before to about 19300 on the preview driver.
 

·
Invalid Media
Joined
·
6,438 Posts
BTW, the preview driver gives a significant boost to Superposition, I went from ~18500 before to about 19300 on the preview driver.
Definitely a good driver. As posted before, I was already at ~19k with the old driver, but this one scored higher with 70 MHz lower clock and full down-clock / down-volt. Hopefully, the final WHQL version of this driver is similar.
 

·
Registered
Joined
·
251 Posts
...It's a tough call to make re. warranty rules in your jurisdiction, but sometimes, cards aren't screwed together perfectly to begin with though function well for a while. I would arm up with good, thick thermal paste and decent correctly-sized thermal pads (or thermal putty for that matter) and take the card apart for a repaste job.
This is out of question. Warranty would be completely void.

What might have happened (pure speculation on my part) is that during earlier MPT-PL runs, paste might have dried out or an incorrectly seated thermal pad exposed a component to more heat (or indeed a deteriorating component on the PCB, such as an input filter). These cards are full of various sensors for the boost algorithms, and it obviously is getting a signal from somewhere that beyond a certain point, there are issues...PL is expressed in Watts which is a measure of heat energy...
I think you might be right. I will still try an RMA. I have proof of this card shutting down on 4 different PSUs, I might be able to force them to accept RMA based on that.
 

·
Registered
Joined
·
59 Posts
Hi guys, I'm following this thread for a while but I didn't find the answers I'm searching, anyway u are nice community here, so I have a reference 6900XT (XTX) no MPT and I score around 21500 in time spy with 2000/2500 MHz core - 2150Mhz FT memory and + 15% PL with 1100mV in wattman, if I want to use MPT, with the stock cooler, how much can I improve? Wich value I have to modify? Anyway no custom loop for me until September/October so I have to wait.
 

·
Registered
Joined
·
119 Posts
Hi guys, I'm following this thread for a while but I didn't find the answers I'm searching, anyway u are nice community here, so I have a reference 6900XT (XTX) no MPT and I score around 21500 in time spy with 2000/2500 MHz core - 2150Mhz FT memory and + 15% PL with 1100mV in wattman, if I want to use MPT, with the stock cooler, how much can I improve? Wich value I have to modify? Anyway no custom loop for me until September/October so I have to wait.
I'm not sure about the MPT settings but I have a couple suggestions for reference 6900 XTs.
Try lowering your memory to 2100 MHz when using fast timings to see if you get better scores. I foumd that mine ran better (higher benchmark scores) at 2100MHz + FT than at 2150MHz + FT. It was also unstable in some benchmarks at 2150MHz + FT.

I would also do the easy pcb back thermal pad mod - [Official] AMD Radeon RX 6900 XT Owner's Club
 

·
Registered
Joined
·
59 Posts
I'm not sure about the MPT settings but I have a couple suggestions for reference 6900 XTs.
Try lowering your memory to 2100 MHz when using fast timings to see if you get better scores. I foumd that mine ran better (higher benchmark scores) at 2100MHz + FT than at 2150MHz + FT. It was also unstable in some benchmarks at 2150MHz + FT.

I would also do the easy pcb back thermal pad mod - [Official] AMD Radeon RX 6900 XT Owner's Club
I'm using 2150Mhz FT because I didn't see any degradation in scores in time spy and a little boost in superposition around 50/100 points, about the thermal pad mods my VRAM are around 85/95 ºC during long heavy game sessions, anyway the gelids are in my Amazon list 😏
 

·
Registered
Joined
·
251 Posts
Open question to all, before I submit my card for RMA , I want to try one more thing. The shutdown temps for TOXIC EE and LC OEM are 103 C. It could be that this temp is being reached for small instance of time which the sensors are not able to pick up. Do any of you know of any XTXH bios which has shutdown temps higher than 103. I remember the shutdown temps for refence card being 118 C.
 

·
Registered
Joined
·
251 Posts
Open question to all, before I submit my card for RMA , I want to try one more thing. The shutdown temps for TOXIC EE and LC OEM are 103 C. It could be that this temp is being reached for small instance of time which the sensors are not able to pick up. Do any of you know of any XTXH bios which has shutdown temps higher than 103. I remember the shutdown temps for refence card being 118 C.
Holy [email protected], I think we know our culprit exactly now. Just flashed a PowerColor bios with 118 C shutdown temps. @Scorpion667 was right all along. Just need to do a little more testing to remove any remaining doubts.

Colorfulness Font Screenshot Line Software
 

·
Registered
Joined
·
3,324 Posts
Holy [email protected], I think we know our culprit exactly now. Just flashed a PowerColor bios with 118 C shutdown temps. @Scorpion667 was right all along. Just need to do a little more testing to remove any remaining doubts.

View attachment 2560386
I wasn't aware the PowerColor XTXH vBIOS works on this card, great find. I though since the PowerColor has a different PCIe power cable setup (3x8pin) that it would brick a toxic EE. You have the Toxic Extreme edition right?

Sorry to hear regarding the issue but yeah something is up with this model and HS temps. You're the 4th person I've seen experience this with toxic EE, myself included. Probably a mix of high thermal density, bad factory thermal paste and bad mounting. My other theory is that both the die and cold plate are concave thus exacerbating TIM pump out
 

·
Registered
Joined
·
251 Posts
I wasn't aware the PowerColor XTXH vBIOS works on this card, great find. I though since the PowerColor has a different PCIe power cable setup (3x8pin) that it would brick a toxic EE.
I took a chance since the card has dual bios.
You have the Toxic Extreme edition right?
Yes
Sorry to hear regarding the issue but yeah something is up with this model and HS temps. You're the 4th person I've seen experience this with toxic EE, myself included. Probably a mix of high thermal density, bad factory thermal paste and bad mounting. My other theory is that both the die and cold plate are concave thus exacerbating TIM pump out
Could be. Now I don't know whether to RMA it or not. Someone from Sapphire's local service center told me that it could be 2 and half months before I get a replacement, or they could send me back the same card again.
The card won't even do 2800Mhz even with 1.287 volts. Truly a shi*ty situation.
 

·
Registered
Joined
·
35 Posts
So I flashed my 6900XT Toxic EE with 6950XT Toxic LE and then I had to load my original stock bios into MPT and pretty much everything works as before but now I can overclock memory to almost 2500 MHz. However, benchmarks did not improve by much. I guess 2100 with tight latency still scores almost the same.


Cloud Light Product Sky Lighting
 

·
Registered
Joined
·
49 Posts
Just out of curiosity what pl are you air cooled guys using in MPT. I know it depends on how well your particular card cools and silicon lottery but just wondering what's a the typical limit you guys are using and what is your max core clocks in something like timespy
 

·
Registered
Joined
·
251 Posts
So I flashed my 6900XT Toxic EE with 6950XT Toxic LE and then I had to load my original stock bios into MPT and pretty much everything works as before but now I can overclock memory to almost 2500 MHz. However, benchmarks did not improve by much. I guess 2100 with tight latency still scores almost the same.


View attachment 2560400
Can you please provide a link to the 6950LE bios?
 
7041 - 7060 of 7233 Posts
Top