Overclock.net banner
1961 - 1980 of 2526 Posts

·
Registered
Joined
·
60 Posts
...tx. That CPU fluctuation is normal for 16-core AM4 (my 5950X does that also). It's when they're left on auto and the 3950X will boost past 4750 until it gets reigned in by the GHz police. Below is the same default GPU setting as before but with the 3950X on all-core 4400; that's all this little AIO can take....nice and smooth :) ...once CPU and GPU are properly water-cooled, it's time to push this setup more.

...on the 6900XT w/ MPT 330W/350A, definitely voltage readout and may be voltage control are behaving a bit weirdly - is that normal after MPT changes ? I'm wondering if it relates to the Giga 3x8 pin arrangement and a VRM that's different from the reference card :unsure:






...at the end of the day, the clock differences will not be noticeable in every-day apps. That said, what gives me pause is the huge delta between overall GPU temp and Hotspot temp you're seeing. On my card, the delta is rarely more than 20 C, often less.

...I've had four new IDENTICAL GPUs before purchased at the same time (incl. for workstations) and when taking them apart for w-cooling, some of them had some serious TIM imprint issues, others had a slightly crooked mount of the cooler, and/or some select loose screws. I realize that taking apart a brand-new (and hard-to-replace) 6900XT is not high on your agenda re. warranty etc., but have you considered emailing the EU HQ of your vendor and ask if you can re-paste / re-pad, given the issues you're seeing ? I've done that before and as long as you're straight-up with them ("no, I have no idea how the flux and shunt mods got on there") they tend to be helpful more often than not.

With Covid-19 impacts, IMO, I see more quality issues in various assembled computer components :mad:
This is my run. I'm not seeing those CPU fluctuations at all.

2514080
 

·
Registered
Joined
·
92 Posts
That's really nice.
I don't know too much about MPT yet either, but what I've read multiple times is that you really only want to touch those two values you already changed.



True. Better have it close to default, just like everything else 😅
I guess this brings me back to my initial question and so far enforcing my belief of having a below average binned 6900 XT. Maybe it's because I have a reference card. Or maybe just bad luck 🤷‍♂️
I did only pay MSRP, so I guess I should be happy for even getting one and running default/stock settings, but I'm not. I want more lol
I'd rather have a below average one at MSRP than the $2300 Red Devil Ultimate that I got. Mine constantly thermal throttles. So I just spent another $300 for a custom water loop for this card.
 

·
Registered
Joined
·
25 Posts
Usually reference cards perform very well. I just think you've been unlucky in the silicon draw :(

One good thing about reference cards is there's plenty of options for water cooling blocks :)
Yeah, just not sure if it's worth it with this specific one. I don't know anymore lol
...at the end of the day, the clock differences will not be noticeable in every-day apps. That said, what gives me pause is the huge delta between overall GPU temp and Hotspot temp you're seeing. On my card, the delta is rarely more than 20 C, often less.

...I've had four new IDENTICAL GPUs before purchased at the same time (incl. for workstations) and when taking them apart for w-cooling, some of them had some serious TIM imprint issues, others had a slightly crooked mount of the cooler, and/or some select loose screws. I realize that taking apart a brand-new (and hard-to-replace) 6900XT is not high on your agenda re. warranty etc., but have you considered emailing the EU HQ of your vendor and ask if you can re-paste / re-pad, given the issues you're seeing ? I've done that before and as long as you're straight-up with them ("no, I have no idea how the flux and shunt mods got on there") they tend to be helpful more often than not.

With Covid-19 impacts, IMO, I see more quality issues in various assembled computer components :mad:
Yeah, true. My settings that give me 1000points higher score in Time Spy lead about 3-5% FPS increase in the few games I tested. Undervolting though drops the temps so much. In Assassins Creed Origins I can do 1110mV which drops the junction temp from 100°C to about 92°C at 1800rpm. Yeah, contacting them seems like a good idea.

If you want a non H chip that does well you need to pick cards that have strong factory overclocks because that's a pretty good indicative of the quality of the silicon.
For instance, my Toxic comes with a boost clock of 2660 Mhz and 2100 Mhz memory OC (16.8 Gbps). This is guaranteed so I already knew the card was higher than average but still, there are better cards here that aren't H chips but that's rare.

In fact, I've seen only 1 bench here in Time Spy higher than my Toxic non H chip (22.200 Graphic Score).

Being on water helps a ton too because hotspot temp is nowhere near throttle limit.

Though, I have seen a case like yours, where the card barely beat 3080's on synthetics. All in all I wouldn't stress too much about synthethics because the difference between 2500 and 2700 is like 1 to 3 frames most of the time??
Yeah, I'm probably being crazy considering returning the card to get an aftermarket one. It really irks me that I'm just limited in every regard. But it's true, the FPS difference is not that big in actual games, so I'm just being silly.

I'd rather have a below average one at MSRP than the $2300 Red Devil Ultimate that I got. Mine constantly thermal throttles. So I just spent another $300 for a custom water loop for this card.
Isn't that a really highly binned version? If it actually throttles at stock, I'd think that's a good enough reason to RMA, not?
 

·
Invalid Media
Joined
·
4,396 Posts
This is my run. I'm not seeing those CPU fluctuations at all.

View attachment 2514080
The AMD 16-cores with dual physical chiplets are binned differently and also have extra boost steps (> heat of 16c/32t). Below is a quick run I did with the 5950X/3090 on the other system, it too squiggles, though not as much as the 3950X does, because the former has a custom loop for the CPU...as shown before, if I lock all-core, it becomes smooth...the 3950X/6900XT just needs to get the full custom loop-treatment installed (all parts are here)

2514095


I can't even get a water block for my card :(

Have you checked Bykski ? They seem to have a lot of different 6090 XT blocks for both reference and even rarer custom PCBs...below is the one for my Gigab. Gaming OC, waiting for my attention...

2514096
 

·
Registered
Joined
·
25 Posts
The AMD 16-cores with dual physical chiplets are binned differently and also have extra boost steps (> heat of 16c/32t). Below is a quick run I did with the 5950X/3090 on the other system, it too squiggles, though not as much as the 3950X does, because the former has a custom loop for the CPU...as shown before, if I lock all-core, it becomes smooth...the 3950X/6900XT just needs to get the full custom loop-treatment installed (all parts are here)

View attachment 2514095




Have you checked Bykski ? They seem to have a lot of different 6090 XT blocks for both reference and even rarer custom PCBs...below is the one for my Gigab. Gaming OC, waiting for my attention...

View attachment 2514096
Btw, mine looks similar to your 5950x. Fluctuating a bit more

2514097
 

·
Invalid Media
Joined
·
4,396 Posts
Btw, mine looks similar to your 5950x. Fluctuating a bit more

View attachment 2514097
...yeah, typical dual chiplet stuff - I bet if you lock your CPU at all-core, it will smooth out.

BTW, the temp graph for your GPU is 'telling'...I don't think you actually have a bad sample, just a cooling issue (possibly related to mounting, TIM etc)
 

·
Registered
Joined
·
25 Posts
...yeah, typical dual chiplet stuff - I bet if you lock your CPU at all-core, it will smooth out.

BTW, the temp graph for your GPU is 'telling'...I don't think you actually have a bad sample, just a cooling issue (possibly related to mounting, TIM etc)
Where do you see the temp graph :unsure:

I just redid a run with 100% fans, 2460 max clk (effective clock was actually between 2200 and 2350 though), 1150mV and +10%PL. Max hot spot was 110°C. Mostly hovering right under it.

2514100


Let me know if you want me to try any specific settings
 

·
Invalid Media
Joined
·
4,396 Posts
Where do you see the temp graph :unsure:

I just redid a run with 100% fans, 2460 max clk (effective clock was actually between 2200 and 2350 though), 1150mV and +10%PL. Max hot spot was 110°C. Mostly hovering right under it.

View attachment 2514100

Let me know if you want me to try any specific settings
...sorry, I misread one of the colours at 4K - however I had checked back w/ your table that had GPU score at 19155, gpu temp at 72 and hot spot at 113. You clocks were ok there, but the delta of 41 C between GPU temp and hot spot looked odd. My first TS run on this last night had GPU max temp at 60 and hot spot at 81, later runs w/ MBT 71 and 93 respectively
 

·
Registered
Joined
·
25 Posts
...sorry, I misread one of the colours at 4K - however I had checked back w/ your table that had GPU score at 19155, gpu temp at 72 and hot spot at 113. You clocks were ok there, but the delta of 41 C between GPU temp and hot spot looked odd. My first TS run on this last night had GPU max temp at 60 and hot spot at 81, later runs w/ MBT 71 and 93 respectively
Yeah, my deltas seem off. 60 and 81... impressive.
 

·
Old to Overclock.net
Joined
·
2,303 Posts
Where do you see the temp graph :unsure:

I just redid a run with 100% fans, 2460 max clk (effective clock was actually between 2200 and 2350 though), 1150mV and +10%PL. Max hot spot was 110°C. Mostly hovering right under it.

View attachment 2514100

Let me know if you want me to try any specific settings
I have been reading your past few post. I agree with others not to really compare scores, however your card seems to be running hot and as a result under performing. Maybe the graphite pad over the core is misaligned. Your card is performing worse than my 6800XT.

Below is what I get in timespy.

2514103


AMD Radeon RX 6800 XT video card benchmark result - Intel Core i9-9980XE Processor,ASRock X299 Taichi CLX (3dmark.com)

Might want to check to see if the heatsink is making appropriate contact with the card.
 
  • Rep+
Reactions: Kir

·
Invalid Media
Joined
·
4,396 Posts
Yeah, my deltas seem off. 60 and 81... impressive.
...while you try to figure out what to do about the temp issue, have a look at the ATI Tool > here ...it is 'ancient', when ATI was not part of AMD yet, but it is part of a suite of softwares that allows you to evaluate your clocks (GPU, VRAM) via 'check for artifacts while only putting a very light load on and thus not overpower everything with high temps. The ATI Tool is not so useful for anything else anymore but if you want to find out the quality of your Chip + VRAM w/o cooling getting in the way, it should help
 

·
Registered
Joined
·
25 Posts
I have been reading your past few post. I agree with others not to really compare scores, however your card seems to be running hot and as a result under performing. Maybe the graphite pad over the core is misaligned. Your card is performing worse than my 6800XT.

Below is what I get in timespy.

View attachment 2514103

AMD Radeon RX 6800 XT video card benchmark result - Intel Core i9-9980XE Processor,ASRock X299 Taichi CLX (3dmark.com)

Might want to check to see if the heatsink is making appropriate contact with the card.
Thanks for the input. Yeah, that GPU score is currently unreachable for me. I will contact the support to see if it's okay for me to open it up. I'm guessing this is not reason enough for them to replace it.

...while you try to figure out what to do about the temp issue, have a look at the ATI Tool > here ...it is 'ancient', when ATI was not part of AMD yet, but it is part of a suite of softwares that allows you to evaluate your clocks (GPU, VRAM) via 'check for artifacts while only putting a very light load on and thus not overpower everything with high temps. The ATI Tool is not so useful for anything else anymore but if you want to find out the quality of your Chip + VRAM w/o cooling getting in the way, it should help
Nice, I'll check it out.
 

·
Professional Lurker
Joined
·
205 Posts
Isn't that a really highly binned version? If it actually throttles at stock, I'd think that's a good enough reason to RMA, not?
Bin quality is NEVER a reason to RMA, ultimate edition or no. That is abusing the system, and drives costs up for all of us if they do approve the RMA.

I'd rather have a below average one at MSRP than the $2300 Red Devil Ultimate that I got. Mine constantly thermal throttles. So I just spent another $300 for a custom water loop for this card.
This is what I've been telling people that see the AIB RX 6900 XT's sitting on the shelves at Micro Center. AIB pricing is out of control right now, and let the AIB RDNA2 cards rot on the shelves for all I care, the markup they're asking is NOT worth it. :(
 

·
Registered
Joined
·
25 Posts
Bin quality is NEVER a reason to RMA, ultimate edition or no. That is abusing the system, and drives costs up for all of us if they do approve the RMA.
The RMA was in reference to his card throttling at stock, not binning. I was just surprised that such a premium card isn't perfect.
 

·
Registered
Joined
·
25 Posts
...while you try to figure out what to do about the temp issue, have a look at the ATI Tool > here ...it is 'ancient', when ATI was not part of AMD yet, but it is part of a suite of softwares that allows you to evaluate your clocks (GPU, VRAM) via 'check for artifacts while only putting a very light load on and thus not overpower everything with high temps. The ATI Tool is not so useful for anything else anymore but if you want to find out the quality of your Chip + VRAM w/o cooling getting in the way, it should help
Hmm, I can set the clock all the way up to around min 2630 - max 2730 (need to set min clock, otherwise it stays low), but the effective clock is always just 1000 MHz. Does that mean anything? I have the impression that since the effective clock is still 1000MHz, it might not really mean much? Idk
 

·
Registered
Joined
·
416 Posts
Guys I have a new 6900xt reference and alphacool waterblock. Does it stand a chance when head to head with a 3090 or 3080ti? If not I may sell it and get other card.
 

·
Old to Overclock.net
Joined
·
2,303 Posts
Guys I have a new 6900xt reference and alphacool waterblock. Does it stand a chance when head to head with a 3090 or 3080ti? If not I may sell it and get other card.
In raw power the 6800XT/6900XT are impressive cards. Nvidia excels in ray tracing and DLLS. I don't own a 3080Ti or 3090 but my 6800XT is on average 20% faster than my 3070.
 

·
Registered
Joined
·
37 Posts
Sure, thanks for the help. At 100% I get about 2930rpm. Yeah, I hit "Write SPPT" and restarted and it worked as far as I can tell, since the reported GPU PPT Limit in HWInfo was now something like 390W (with +12% from Wattman), which it also hits for a short moment before throttling because of high temps. All the MPT does is allow higher Wattage though, right? So considering no MPT and only Wattman +15% brings me to temps beyond 110°C already, giving it more power won't really do anything, right?

What CPU Score are you referring to? From the links in my spreadsheet or the 13660 in the picture last page? I ran all early tests with one CCD deactivated, because I was still waiting on my cooler. But I think 13660 is normal for a stock 5900x? Yup, XMP is active and I also have tried high performance.
Like others have pointed out, It could be a case of just bad Thermals due to improper heatsink setup, etc, or just a bad binn. You could as a final test try it in another system (if you have any spare)?
 
1961 - 1980 of 2526 Posts
Top