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

[GerryB]

LD4 only 2.615MHz

btw.
if TDV is activated in FeatureControl, when yes, then try more TDV
881mV should match with LD2 and minClock 1475MHz (MCT)

 

[cra3khead]

LD4 only 2.615MHz

LD4 at 2.615 and LD5 at 2.860? I don't use TDV, I just copied the settings you sent a while back; should I use it?

LD4 = 2.625 passed this time; maybe the ambient temperature was too hot the last time I tried it.

 

[GerryB]

LD4 = 2.625 passed this time; maybe the ambient temperature was too hot the last time I tried it.

I think that´s enough, stay with this values, ... hope, it works also in BO6 with some watt less

try more UVen together with minClock 1475MHz and TDV
MCT 987mV -->975mV like Fas7

btw.
after that runs stable
as additional question 1
What happens, if you go lower with MCT- GPU-Power 112%-->110% ?
You are so much tempthrottel related, may be nearly the same score.

additional 2
may be TDC needs less Ampere because of the Power 112...110% (MCT)

 

[cra3khead]

try more UVen together with minClock 1475MHz and TDV
MCT 987mV -->975mV like Fas7

Anhang anzeigen 1623081

I'm getting higher clocks in games now, but the voltage jumps up to 1.084V. Previously, when I maxed out at 2633MHz, the voltage would stay around 1.071V or 1.072V.

 

[GerryB]

try 5MHz more maxClock (MCT)

 

[cra3khead]

This was at 2685Mhz, but in games I can't go past 2660 without inevitably crashing.

 

[GerryB]

This was at 2685Mhz, but in games I can't go past 2660 without inevitably crashing.

some games are different
use less UV in MCT

What is your MCT-Setting in this TS run?
and, ... MPT = 1112mV?

 

[cra3khead]

What is your MCT-Setting in this TS run?
and, ... MPT = 1112mV?

I tried going down to 968mv in MCT, but it hard crashed. I think it's completely tapped out. Thanks for all the help

 

[GerryB]

just for fun, try some final finetuning
I would use 2 settings, first only in TS and a safer second in 24/7 gaming.

 

[cra3khead]

just for fun, try some final finetuning

I don't know why it dropped my score around 100 points.

 

[GerryB]

GPU-Utilization in GT2 not 99% constant?
if so, then may be too much Windoof in the background
or your special HAGS works only sometimes

or
try TDC 250-->255A and SOC 30-->32A

 

[cra3khead]

GPU-Utilization in GT2 not 99% constant?

How do I check that? I'm using release preview channel windows. Should I switch to Windows 11 by Ghost Spectre? I stopped using the HAGS driver; I only use it if something like Oblivion Remastered comes out like for unoptimized heavy Unreal Engine 5 games at launch, just to squeeze out as much fps as possible.

 

[GerryB]

How do I check that?

UL shows you all Infos: I have this problem with my little old CPU often.

 

[cra3khead]

@GerryB Damn, it crashed in Expedition 33 after an hour. Should I switch back to a pre-TDV setting or just put the voltage in MCT back to 987mv?

 

[GerryB]

try a little bit more Volt in MCT
sometimes the same UVing works not in every game
or
How is the eff.Clock in 33 ?
may be too high, then 5...10MHz less in MCT
or
both

btw.
crash after a hour can also mean, CPU+Ram are not stable enough/too hot, produce some failure
set CPU+Ram to defaults and try a second hour

 

[cra3khead]

try a little bit more Volt in MCT

For some reason E33 needs 1012mv in MCT; but, if I try that in TimeSpy it crashes. TimeSpy passes with 975mv. That doesn't make any sense to me.

I'm still using these settings

 

[GerryB]

Is the same behaviour like RDNA4.
Synthetics are not Gaming

or
How is the eff.Clock in 33 ?
may be too high, then 5...10MHz less in MCT

in PR is my highest clock@Synthetics, but Rome-Demo=UE5@AFMF boosts higher
(in Dune Awakening it´s required for me, to set the maxClock 10...20MHz lower, may be in 33 the same)

this is moore Boost then usually@stock:

so it´s no surprise, that not every game can run stable with this boost-overshot

edit:
der 25.6.2 läuft auch ganz gut

 

[therealcola]

Ups Falscher Thread

 

[GerryB]

Ups Falscher Thread

Könnt ihr mal laufen lassen, basierend auf der News
https://wccftech.com/rx-9070-xt-sam...-worse-than-sk-hynix-memory-based-rx-9070-xt/
Das Tool gibt es hier
https://github.com/clamchowder/Micr...ownload/08bbe54/MicrobenchmarkGui-08bbe54.zip

... ein interessantes Tool, müsste man mal für FT1 vs. FT2 testen
2066-FT1 läuft bei mir so, siehe Anhang
(fclk nur 1433, falls das auch ne Rolle spielt)

 

[cra3khead]

2108FT2 Fclk at 1900