AMD RED BIOS EDITOR und MorePowerTool - BIOS-Einträge anpassen, optimieren und noch stabiler übertakten | Navi unlimited

[PJVol]

@Veii and @everyone​

It's funny, but I wouldn't be surprised if CPOs are indeed part of the system since Navi21. Or, is there something so obvious that I missed while trying to figure out what makes the VF curve stop in the first picture ?
(power/tdc uncapped, temps окay)


Upd: nvm, it was power, but until a certain point, where clocks start stretching a lot (> 25mhz), seemingly due to lack of voltage.

 

[ShirKhan]

CPO=?

 

[Veii]

@Veii and @everyone

Wanna see something funny:

The overboost issue is back with DS wakeup slingshots
I soo hope that SVI3.0 implementation will finally resolve it.
2 Years, maybe 3 already - nothing, still the same issue even if fully obvious

It's such an instability issue, but AMD doesn't try to care

 

[PJVol]

CPO=?

Ah... sorry, this was the specific way of talking with Veii since OCN ))
Critical Path Oscillator (CPO) circuits is what we expect to see in the AVFS modules for the rdna 3. They basically functioning for the purpose of run-time Fmax tracking and collection data for the parts (tiles) of a compute or other units (which may utilize different type of cells by design, i.e. Vth, process corners, etc) to fight timing issues and gain additional power savings.
Not an easy topic, but you can start from this short article for the basics:
(direct link is not available since it's behind the registration wall, but not hard to get it some other way )

https://www.researchgate.net/publication/301610844_Adaptive_Voltage_Frequency_Scaling_Using_Critical_Path_Accumulator_Implemented_in_28nm_CPU

 

[Veii]

@PJVol I certainly see constant clock gating even at highs
But it's sub 50ms , often faster. It's not trackable. FID & AC clock gating

Adaptive?~PCC topic really baffles' me.
If it is what you tell it is, then why is it disable by default except for 6800XT - but works on all cards when enabled
Why can't it break like early PPT1 FUSE (enable only on supported bios) // nothing works when enabled, till supported

Same goes for aging margins
They certainly can work, but it's not implemented. I'm sure it needs some toggle to make it run

it's only running BTC per soft're-boot - unless you do big FMAX changes
then it crash/reboots itself and retrains // else without FMAX change, FID/VID changes instant train on soft-reboot
But it predicts min-max curve , their steps. Either driver or SMU itself. Soo APPC topic really questions me things
I see no point that would say it falls back to predefined voltage table (like for RDNA3). Only see realtime adjustments by fast training

 

[Veii]

Ah from your perspective @PJVol
Would DC AVFS adapt itself by higher delta margins to AC
Or there is no adaptiveness and it is by design as close as it gets // soo when fooling with curve it will messup balance

Asking,
When fooling around with Input and at the same time adjusting AFVS on last stage
Would it make any sense to try and figure out if it missmatches balance between AC/DC - or you think module is already automatically adapting ?

 

[Veii]

Also any idea on APT (adaptive power target?) and PPM (powerplay margin? )
// do they say anything to you ?

 

[PJVol]

Would DC AVFS adapt itself by higher delta margins to AC

They are not really good in naming things, look at the Ac/Dc stuff.

I think now I see what @hellm meant, lol
Pls, someone call Sam Naffziger to join the chat

// do they say anything to you ?

No, should it?

 

[Veii]

No, should it?

Probably, as it's the odd 6800XT that uses them
While nothing else really does

Its the same question why 6800XT has 33.33inf~ % higher L$ bandwidth
Vs their 6900 XTX "clone"
No other card i can see, runs it this high

 

[RX480]

may be caused by the difference between the nonXT and the XT in the number of ROP´s
that seems to be an advantage in some cases like TS with additional high async compute load

 

[Veii]

It's an in-bios thing that's capped
6800XT with 6950XT spoofed bios , with same ROPs runs 1200GB/s cache
with native 6800XT bios it's 1600GB/s

But "why"

 

[RX480]

We don´t know at which time the first driver and bios are testet and so its probably a fault by copying.
They made at the same time a non XT and a XT, may be therefore used different speeds for some testings/reasons.
The 6900 is a bit later with the bios+driver and may be only with the requiered speed/bandwith.

Also may be they did some things together with miners.(the miners let the nonXT run@soc 616MHz)

whats about the deep sleep states@6800xt, seems with low controlerload some vrams sleep and the soc runs with
lower watts, for example a effizient tweak by the luxx user RaceOwl (this is achieved with feature memscaling=OFF, ... only DPM0-1 active)

 

[Shaav]

Hi there,

I want to flash my RX 6600 from ASRock.

I saved the BIOS, loaded it into MorePowerTool and saved the MTP-file.
Then I openen Red BIOS Editor but could not load the MPT-file.
Do you have any idea what is causing the problem?

 

[RedF]

Hi there,

I want to flash my RX 6600 from ASRock.

I saved the BIOS, loaded it into MorePowerTool and saved the MTP-file.
Then I openen Red BIOS Editor but could not load the MPT-file.
Do you have any idea what is causing the problem?

This is not yet possible with RDNA2.

 

[ApolloX]

"Not yet" →also forget it for the future, I don't expect any flashing for rdna2.

 

[ShirKhan]

except

except? accept? expect?

 

[PJVol]

Can someone make it clear what all this fuss around flashing the VBIOS is about? Doesn't pp tables cover most of avg. OC-er's needs?

 

[ApolloX]

Yes, yes, and, yes.
One very small group of people is dreaming for editing VRAM timings and one way might be flashing.
And a large group just has no clue and assumes that RDNA2 is the same like RDNA1, which is not true.

 

[Veii]

"Not yet" →also forget it for the future, I don't expect any flashing for rdna2.

I wish you would supply a bit more positivity.
You speak confident about a denial while not even sharing any experiences of your attempts.
It has a bit of a "fear-of-denial" taste.
I'm sorry, but that's how it comes over.

What you do well,
Is fuel motivation of people to not even attempt looking into it.
This hinders progress and is as far away of "German" as it could get, in terms of innovation.
It is possible btw. But the targets are not met.

The targets are not memory timings, which are already usable by spoofing on-boot check, loading different ID-BIOSes with different utilisation/driving of memcontroller.
Yes, "different driving" as write-only supplied patches into SMU/PSP.

 

[RX480]

@Veii
Can you provide a 6800xt-Bios with 6950 Timings to redF. (and a 6900xt to other one)
He is a very positive german TESTER.
and
likes to give his experience to all others in this tread
may be it works only till Vram2150MHz@WM but with not so sharp timings usefull@24/7
or in best case FT2 works

whats about spoofing, flash @linux, or are other tricks needed?
edit: OK, i see your explanations in the LUXX-Thread:
https://www.hardwareluxx.de/community/threads/offizieller-amd-rx6700-rx-6700xt-x6800-6800xt-6900xt-overclocking-und-modding-thread-wakü-lukü-ln2.1282285/page-854#post-29463360

"different driving" only by load to MPT needs some work, to set some values back to 6800xt
would be nice to have a modded bios without that work for max frequencies and voltages

may be, with easier use, there will be more posts/expierences
edit: but @Veii say´s that only 1/4 of the 6800xt can use properly the mod-Bios
(we will just have to wait for moore)