Sammelthread Big Navi RX 6700XT/6800(XT)/6900XT Overclocking/Undervolting

[cra3khead]

@GerryB is the voltage curve in MPT worth messing with?

 

[GerryB]

With your very high Watt probably not!

with 230W may be the 718mV and the aging value can make a difference
but care about MPT-Volt to MCT-Volt, may be then not 100mV difference but only 97... 94mV stable

btw.
I looked at the HoF, to find a other User with my 24/7-Score, to compare FT2 vs FT1: 2148MHz vs 2150MHz
see attachment, ... he needs100MHz more avg. gpu clock, to achieve small 57points more= not efficient

 

[cra3khead]

What does a crash in GT1 usually mean?

 

[Evgasüchtiger]

That it's not stable 😁

 

[GerryB]

with 230W may be ...

@cra3khead
probably a other way is also possible, ... set a little change with 25MHz in LD2+3+4 ... 2500-->2525MHz
and then try to optimize the Sweetspot around 2523MHz, ... works in this way, 2200-->2223MHz@my 6800nonXT
but
the other change in the tolerance as second step, may be not soo stable

With your high temps I would take only such a lower maxClock!

 

[cra3khead]

set a little change with 25MHz in LD2+3+4 ... 2500-->2525MHz
and then try to optimize the Sweetspot around 2523MHz

Would you be able to help me do this for 2565Mhz? I know 1050mv at 2565Mhz is damn near 100% stable on my card at the moment.

 

[GerryB]

Would you be able to help me do this for 2565Mhz? I know 1050mv at 2565Mhz is damn near 100% stable on my card at the moment.

It´s not important to stay at 1050mV in MPT because 1056mV is also lower then 1050mV effectiv.
try 1056 with 2563MHz, ... may be 1056 is enough to reach more maxClock@MCT = 2568...2573MHz
(with 265W? only at some scenes of the TS the eff. maxClock >2500MHz, ...
with maxClock 2573 may be useful to change LD4 2500-->2525MHz as described #6.645)

() that would be a nice/cool goal, also 24/7 useful

I think for example, that the 6950 use the higher per part curves only if the effectiv Volt >1050.
at eff. <1050mV there are lower curves (more effizient?)

edit:
In the same manner I changed my MPT 881mV to 893mV to have fun in Nomad Light@DX12.
thats perfect with low Watts to have a great Boost, ... see attachment1
Nomad also likes a little bit FT2, compared with a other User@FT1, ... see attachment2

in real games I use only maxClock 2248

 

[Evgasüchtiger]

hier mal 45 min GTA5

 

[cra3khead]

with maxClock 2573 may be useful to change LD4 2500-->2525MHz as described #6.645)

Would that also be applicable if the max clock is 2568Mhz? I'm not sure if I can reach that high of a clock of 2573Mhz with 1056mv, as when I tried 2580Mhz at 1062mv; it wasn't stable. 1050mv at 2565Mhz I know for sure that it is stable at those settings. I still want to try those optimized curve settings though. Is it universally applicable to any settings or simply to the effective clock being <2500Mhz?

 

[GerryB]

Would that also be applicable if the max clock is 2568Mhz? I'm not sure if I can reach that high of a clock of 2573Mhz with 1056mv, as when I tried 2580Mhz at 1062mv; it wasn't stable. 1050mv at 2565Mhz I know for sure that it is stable at those settings.

1050-->1056=6mV more should be enough to try 10MHz +, ... 2563+10=2573
if not stable
2568 is enough with 265 watt

I still want to try those optimized curve settings though. Is it universally applicable to any settings or simply to the effective clock being <2500Mhz?

every chip is a little bit LOTTERIE
my chip allows@LD4= 2,200GHz-->2,225GHz

may be your chip can do also this 25MHz (2,500-->2525)
(the reference settings should be with a little safety buffer)

 

[cra3khead]

may be your chip can do also this 25MHz (2,500-->2525)
(the reference settings should be with a little safety buffer)

Unfortunately, my card doesn't respond well to the optimized curve settings at all; it just keeps crashing no matter what voltage-fmax settings I use

 

[GerryB]

Unfortunately, my card doesn't respond well to ...

may be, LD3 must stay safe at default 2,300GHz

sometimes it´s helpfully to know at which Clock the crash happens
not every crash is at the highest boostclock

 

[cra3khead]

may be, LD3 must stay safe at default 2,300GHz

Yeah I think that was the issue.

Although my score is 30 points lower than it was previously; but I don't know if it has to do with the weather since it's over 100 degrees today.

 

[GerryB]

looks good, the clock dip is not soo sharp

just for fun
try a little bit less max clock:
2563 and 2558 with MCT 956...957mV
sometimes the avg. clock is then smoother
and may be <eff.2500MHz a little bit better in the efficiency

and
buy a cheap full version
without demo the temps are lower

 

[cra3khead]

just for fun
try a little bit less max clock:
2563 and 2558 with MCT 956...957mV
sometimes the avg. clock is then smoother
and may be <eff.2500MHz a little bit better in the efficiency

What clock speed would LD5 control? I prefer it the way you set it up initially with 2573 because in game I can clock anywhere between 2517 and 2530MHz, whereas previously, without the optimized curve, it was a static 2500. I wonder if a curve optimization would be possible to have my 1093mv–2625MHz profile, boost from 2560 to 2580MHz.

 

[GerryB]

don´t change LD5
may be required@default 3100MHz if the driver crashs

btw.
If maxClock 2553 runs with MPT 1043mV MCT 944mV, that´s would be nicer.
I think, to get 22,3k Score you need more efficiency, means higher avg.clock.

 

[cra3khead]

don´t change LD5

So does the LD4 trick only work for a 2525MHz sweet spot, so for example, if I set 2625MHz (2560MHz effective), it would be pointless because it's running above the LD4? In that case, what voltage would I need for 2585MHz, or should I just keep it at 2573? Sometimes in the game, I can see it hitting 2540MHz for a split second with the setting you provided. I personally am going to keep it that way if there's no other optimization to boost clocks further.

 

[GerryB]

edit:
Try 0,1062500V with 2,550 or 2,575GHz. If not stable, then 0,101250V or 0,101875V.
I don´t know if so much change above default 2,500GHz works.

 

[cra3khead]

with maxClock 2573 may be useful to change LD4 2500-->2525MHz as described #6.645)

With Fmax 2573, wouldn't it only hit 2513 as the effective clock? Wouldn't 2585 be the optimal Fmax to hit 2525, or am I misunderstanding what the LD4 optimization does?

 

[GerryB]

With Fmax 2573, wouldn't it only hit 2513 as the effective clock? Wouldn't 2585 be the optimal Fmax to hit 2525, or am I misunderstanding what the LD4 optimization does?

the old optimization@2,525Ghz should be fine till around eff. 2525MHz and some MHz above
but you ask also, what is with higher eff.MHz.(x)

btw.
There was an 10-error in post#6.658.(see edit)

(x) it is only tought to define more exact the Volt around this clock