RDNA2 RX6000 Series Owners Thread, Tests, Mods, BIOS & Tweaks !

Yes, by default the 5800X3D only goes up to 5.54 Ghz. The 5800X goes to 5.7

 

You mean 4.5 and 4.7ghz

Anyway that is some nice timespy scores on the 6800xt. I think it beats even 3080 12gb.

 

Yes, but Im not even hitting the 5800x3d clocks. 4.5 is the max. Im barely seeing 4.3

@Horus-Anhur

So -30 gets me 4.45 all core, which is up from 4.1 all core.
4.5 also for single core.
CPUZ is 604.1 and 6373.3 currently like this.

 

See, I told so

 

4.55 single core 4.45 all cores

 

Thank you for all the help! Thats gonna be a huge help when in game. It jumps all over the place.

 

You are correct. It's 4.55 Ghz max on the 5800X3D. I'm past my bedtime
My 5800X3D manages to be both in single and multi thread, on 4.55Ghz

 

On multi you at 4.45

 

This is how mine is doing now.
6800xt is flying now in God of War. Able to keep 100fps avg now, was having drops into the 70s with the 5600x.

 

Yes, 4.45 Ghz. I misread my values.
In fact I always get 4.45Ghz. Single and multi.

 

You need to go to sleep ... you get 4.54 in single

 

You are right. I need to go to sleep.
A few days ago, I noticed I was getting a lower score on CPU-Z single score. And that Aida 64 latency results were 2ns higher.
I was tweaking with the C-State Controls, and didn't even realize I had already fixed the issue.
Sorry everyone for the nonsense

My CPUZ and Aida 64 scores are back to normal.

 

As seen in the first picture in my post, the 6800 XT itself is quite strongly tweaked, with undervolting, raised max clock, raised power limits to 15% and +5% to memory with fast timings.
Just the CPU tweaking by itself doesn't do much.

The difference to stock is quite significant, in the ballpark of 10% higher scores, if not more !

But obviously, undervolting is silicon lottery... go -1% at a time and +1% to max clocks, run tests, see if scores improve or your system crashes.
When it does, go 1% back and that should be stable enough.
Mine instantly crashes at -13% voltage and is is somewhat unstable at -12%, but -11% seems to work fine in both benchmarks and games.

The advantage in games is that it consumes very little power when not needing that much compute power, as it hits the VSync of the monitor and doesn't spew out more frames.

---
I actually leave these settings permanently on, with the exception of power limit, which goes from +15% when benchmarking to -15% when gaming, to save power as much as possible.

With the tweaking set, I'm getting in games 98-101% of stock performance for that 15% reduction in power... which is pretty amazing

---
As for the CPU tweaking, I use the Curve Optimizer which unfortunately doesn't seem to exist on 5800X3D, as well as 3600 CL16 memory with 1:1 sync with the fabric clock ( 1800 ).
Memory timings are also tightened up a tiny bit (not the CL, but the rest), but that doesn't make much difference.

 

this is mine 5800x @ mod hydra 1.2fpro
mem 16 giga patriot rgb
red devil 6700 xt @ 2750/2100

 

p.s.


Make sure Coupled Mode is ON at whatever frequency your Fabric can go, because uncoupled you're severely crippling CPU performance.

 

For me the recording via the Radeon software has been broken for a few months.

 

Yes but it hits (for me) 4.5ghz easily and is faster in CB20/23 at the same speed as my 3800X. PBO2 tuner brings the voltage and temps down a huge amount.

 

What is the default power limit on your card (in watts, not in %%) ?
What's good using %% in tweaking when user card's base values set differently? How we are suppose to compare scores then?

 

It doesn't say, there is no setting in watts, only %

However, in use, it does show how much it supposedly consumes in watts, and it's about 275W for 0%, 315W max for +15% and ~235W for -15%

 

Ok, here is the TS run with PPT 312W = 272 * 1.15 tweaked settings (just found TUF BIOS at techpowerup for you, where it showed 272W default PPT). Actually HWInfo64 sensors does show your GPU limits:


Edit: added SP and FS Ultra score. FS Extreme seems to be cpu-bound