Overclock.net banner

3441 - 3460 of 4332 Posts

·
Registered
Joined
·
330 Posts
Quote:
Originally Posted by gupsterg View Post

@Wimpzilla

No worries on info share
smile.gif
.

I usually increased TDC as a % of what I increased TDP & MPDL, as highlighted in OP section How to edit PowerLimit . On Hawaii the values I found in an updated ROM for my card via Sapphire support service were ample for 1150MHz GPU with DPM 7 VID 1.325V, so VDDC was lower than that when GPU loaded as LLC/slope was stock. I never had need for higher so never tried. Several members have used PT roms TDP/MPDL/TDC (ie 999 for all 3), even if you use those values the "protection" features of IR3567B should kick (ie OCP, unless modified from stock).

VMAX at 66h will limit voltage at 1.48125V, so closer to 1.5V than 1.4V.
So lets say register 33 has data value 74h. Take 74h and convert to binary, 0111 0100 . Count the bits from right starting at 0, so bits [4:0] is 1 0100, convert to dec = 20, 20 x 2A = 40A. AFAIK on reference PCB out of the rear 6 phases only 5 are used by loop 1 for GPU (see image in OP How to edit PowerLimit) so basically 40A x 5 = 300A is OCP limit loop 1. Bits [7:5] are 011, convert to dec = 3, so OVP is +325mV. AFAIK OVP is relative to VID, so for example say GPU is requesting VID of 1.3V then IR3567B will allow a swing up of +325mV = 1.3325V, if I was meddling with LLC/slope I'd probably lower OVP to keep overshoot in control if you get what I mean.

I've lowered OCP in stages on my Fury X, this was handy to gauge roughly what IR3567B is seeing as monitor amps. Stock is 240A on my card, 204A resulted in card cutting out as soon as 3DM FSE loaded, 216A has been ample for all gaming/benching @ upto 1440P and done lot's of runs of [email protected] at 48hrs+.

@Worldwin

I'm pretty sure you've latched on to wrong hex values. Any chance you can attach ROM to post so can check for you?
Id love to but i get this error when i attach the rom:"AJAX response unable to be parsed as valid JSON object. "
 

·
Registered
Joined
·
122 Posts
Quote:
Originally Posted by Adromi View Post

Hi
wave2.gif


I have BIG news for you !!

I was watching some random crazy reverse engineering stuff --> watch?v=6hxEUm-pHUE

And as you can see this guy found the Christmas gift that you were waiting for
rolleyes.gif


A very nice dev put on the internet the ... register's doc of Bonaire GPU !!! In plain text, with commentary 8)
and the mapping appear to get similarities with Hawaii
thumb.gif


I don't have skill for bios editing but I guess you can make something with it
biggrin.gif


Sorry for my English it's not my native tongue and I don't know if these links have their place here, so feel free to edit/delete my post! (or maybe i'm 100% off topic ><)

Have a good night
typer.gif
radeon-tools
smile.gif
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #3,443
Quote:
Originally Posted by Wimpzilla View Post

So for resume, if i get it well, the register 33 have 2 setting on it, the OVP/UVP together and the register 34 is the OCP, wait?
It seems to me that the key is the register 34, because even with ff on VMAX 3D and OVP 33, i didn't see significant vccd modulation changes.
The vccd is a bit lower now +225mv 1.287/1.305v but give roughly the same/more current that stock bios with +300mv 1.3051.320v.
My stock register 33 is 7Ch, 124dec, ‭0111 1100‬bin. 11100‬=28*2A=56A 56A*5=280A OCP with 280A TDC, OVP is 011=3 so +325mv.
My stock register 34 is 79h, 121dec, ‭0111 1001‬bin. So what is it, if the OCP/OVP are merged? The UVP or slow OCP, because i set ff here too!
Yes, register 33 and 34 have 2 settings each. Reg 33 is OVP for both loops AFAIK, OCP is just for loop 1. Reg 34 is UVP for both loops AFAIK, OCP is just for loop 2.

Changing OCP values on Fury X made no difference to how VID/VDDC behaves, I have no Hawaii cards anymore so can not test. For example last night I was doing a run of TimeSpy @ 1175MHz @ 1268mV on My Fury X, the OCP in my everyday OC ROM of 1145/545 is set right for that OC 216A, card shutdown on TS test loading, I had to use stock 240A, VID/VDDC was unaffected in monitoring on either ROM/OCP setting for 1175MHz @ 1268mV.

Basically OCP is a limit value if reached card shuts down it should not at all alter any other aspects as it's not meant to AFAIK. OCP is set to fast mode and slow is disabled, it is set as per phase, as phase balancing is employed each phase should have equal loading.
Quote:
Originally Posted by Worldwin View Post

Id love to but i get this error when i attach the rom:"AJAX response unable to be parsed as valid JSON object. "
Zip ROM and attach
wink.gif
.
 

·
Registered
Joined
·
509 Posts
Quote:
Originally Posted by gupsterg View Post

Yes, register 33 and 34 have 2 settings each. Reg 33 is OVP for both loops AFAIK, OCP is just for loop 1. Reg 34 is UVP for both loops AFAIK, OCP is just for loop 2.

Changing OCP values on Fury X made no difference to how VID/VDDC behaves, I have no Hawaii cards anymore so can not test. For example last night I was doing a run of TimeSpy @ 1175MHz @ 1268mV on My Fury X, the OCP in my everyday OC ROM of 1145/545 is set right for that OC 216A, card shutdown on TS test loading, I had to use stock 240A, VID/VDDC was unaffected in monitoring on either ROM/OCP setting for 1175MHz @ 1268mV.

Basically OCP is a limit value if reached card shuts down it should not at all alter any other aspects as it's not meant to AFAIK. OCP is set to fast mode and slow is disabled, it is set as per phase, as phase balancing is employed each phase should have equal loading.
Zip ROM and attach
wink.gif
.
Thanks you so much, got it right now!
happysmiley.gif


Well the only thing i can say is, in my case removing all these bios protections like the PTX bios do, allow you to get the same VCCD with lower added voltage. I only kept the TDP/PL/TDC values in the bios, like 300/300/280.
Like now i only give +150/175mv instead +300mv for roughly the same VCCD, 1.29v instead 1.32v, the lower VCCD is balanced by an higher current draw and obviously consumption i think.
Now i'm testing the OCN custom 390 bios, i flashed the one with modded memory tables and 390 VID profiles.
My stock VCCD is really low 0.898v idle, 1.00v full and the card hold its 1050/1250mhz. When i push 1250/1450mhz the VCCD is between 1.266v to 1.305v. On average it is 1.289v.

The problem now is the ram do not oc anymore, at 1450mhz already i think it is unstable and then lock my max gpu clock.
It is true i get roughly the same performances now than with the stock bios, ram at 1665/1700mhz.
So don't know if this lower VCCD impacted on lower memory voltage and then oc.
Next steep is so the 390 bios VID with stock memory tables, and without any limits, even the TDP, etc!
biggrin.gif
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #3,445
Cheers for info
smile.gif
.

If you are using the Insan1ty ROM that has tighter RAM timings which will lead to different RAM clock capability.

I only tested 390/X ROM on 290/X few times when had cards, usually the 390/X ROM will set VID in "auto calculate" mode different to 290/X, so set the VID per DPM as you would have on 290/X ROM on 390/X ROM.

Some have gains on 390/X ROM, I didn't when comparing with 290/X ROM with tighter timings + 390/X MC timings, Hawaii bios mod OP has same VRAM_Info tables which you can insert into 290/X ROM to give it 390/X MC Timings, they have stock RAM timings so mod as and if you require. It's also best to use the VoltageObjectInfo from your stock ROM with a 390/X ROM, as then VRM setup, etc will be for your PCB.
 

·
Registered
Joined
·
509 Posts
Quote:
Originally Posted by gupsterg View Post

If you are using the Insan1ty ROM that has tighter RAM timings which will lead to different RAM clock capability.
Exact.
Like i said, i noticed that, it really impact on gpu too, like it strongly bound the gpu on memory performance, behavior.
Quote:
Originally Posted by gupsterg View Post

I only tested 390/X ROM on 290/X few times when had cards, usually the 390/X ROM will set VID in "auto calculate" mode different to 290/X, so set the VID per DPM as you would have on 290/X ROM on 390/X ROM.
Well if i look into vrmtool with the 390 VID, i have into the 4D register already 20 value set, instead of the 60 on 290 stock bios.
Then the stock VCCD is really low given the 4D register value, it is a bit crazy! Anyway it is fine for me, the card is almost 80% asic, so it hold low VID pretty well, happy with that.
Quote:
Originally Posted by gupsterg View Post

Some have gains on 390/X ROM, I didn't when comparing with 290/X ROM with tighter timings + 390/X MC timings, Hawaii bios mod OP has same VRAM_Info tables which you can insert into 290/X ROM to give it 390/X MC Timings, they have stock RAM timings so mod as and if you require. It's also best to use the VoltageObjectInfo from your stock ROM with a 390/X ROM, as then VRM setup, etc will be for your PCB.
Well just for change the max clock of the OCN custom bios that is lock to 1250Mhz was a nightmare. I'm not so comfortable editing registry values, i lost myself.
So even if i understood how to do it, i took me 2/3 times before modify the right string, i couldn't imagine the whole VObjInfo or the ram tables.
biggrin.gif
 

·
New to OCN?
Joined
·
26,919 Posts
what kind of modification needs the BIOS when the clock speed doesnt change when overclocking/underclocking with MSI AB with MLU ROM, the bios have some issue when load with OC the gpu core clock doesnt load at the full frequency and drops 2MHz and then when you try to underclock the clocks arent applied
 

·
Registered
Joined
·
509 Posts
Quote:
Originally Posted by PontiacGTX View Post

what kind of modification needs the BIOS when the clock speed doesnt change when overclocking/underclocking with MSI AB with MLU ROM, the bios have some issue when load with OC the gpu core clock doesnt load at the full frequency and drops 2MHz and then when you try to underclock the clocks arent applied
The MLU bios have low basic TPD/PL/TDC values, the OCP/OVP/UVP are stock, but the VMAX was lowered by The Stilt compared to the stock bios, then it have lowered vrm switching frequency.

Eveything put together make it a good bios if you want run the card stock or with a basic oc, but not so good if you push over the allowed power and voltage configuration allowed by the bios.

MSI AB do not do nothing over changing the clocks, PL and the gpu VCCD/VCCDI offset by I2C/driver modifications.

MLU bios is awesome if you fold or mine with a small oc, keep lower t° and power consumption, but not if you want a strong oc, imo.

So the gpu stay at 300mhz mark even at load? Or reach it oc clock then lower it by 2Mhz?
 

·
New to OCN?
Joined
·
26,919 Posts
Quote:
Originally Posted by Wimpzilla View Post

The MLU bios have low basic TPD/PL/TDC values, the OCP/OVP/UVP are stock, but the VMAX was lowered by The Stilt compared to the stock bios, then it have lowered vrm switching frequency.

Eveything put together make it a good bios if you want run the card stock or with a basic oc, but not so good if you push over the allowed power and voltage configuration allowed by the bios.

MSI AB do not do nothing over changing the clocks, PL and the gpu VCCD/VCCDI offset by I2C/driver modifications.

MLU bios is awesome if you fold or mine with a small oc, keep lower t° and power consumption, but not if you want a strong oc, imo.

So the gpu stay at 300mhz mark even at load? Or reach it oc clock then lower it by 2Mhz?
it has a 2MHz lower core clock , and while underclocking doesnt change anything so you have to use 1000/1250 to have the lowest power draw
 

·
Registered
Joined
·
509 Posts
Quote:
Originally Posted by PontiacGTX View Post

it has a 2MHz lower core clock , and while underclocking doesnt change anything so you have to use 1000/1250 to have the lowest power draw
I don't know sorry, my actual knowledge on hawaii bios modding stop to simple stuff, so i looked for HawaiiBR and VRMTool, the bios hasn't any special features that explain why it can't downclock or it have this 2Mhz less on the core. It just show it is a power efficient oriented bios.

I suppose he modified some other registers other than the basic ones, together with low vrm switching and power efficiency optimizations, i suppose the bios calculate the clocks, using the max clock you set to him minus this 2mhz, for his VCCD and power balance.

Or it is a trick to joke the bios/gpu, max clock values were downclocked, minus 2Mhz, like this the bios/gpu is happy even if he went over his TDP settings.
 

·
Registered
Joined
·
452 Posts
Please haaaaaaaaaaalp! I have been flashing ATI cards for a long damn time and all of a sudden I cannot get atiflash to work from dos! Is this a Windows 10 issue? I have tried the usual usb tool makers with the msdos boot files, as well as freedos. Every time I try to run atiflash from dos (I have tried 3 different versions, and yes it's atiflash not atiwinflash), I get the error "This program cannot be run in dos mode" WHAT THE EFF????
 

·
Registered
Joined
·
509 Posts
Quote:
Originally Posted by madmanmarz View Post

Please haaaaaaaaaaalp! I have been flashing ATI cards for a long damn time and all of a sudden I cannot get atiflash to work from dos! Is this a Windows 10 issue? I have tried the usual usb tool makers with the msdos boot files, as well as freedos. Every time I try to run atiflash from dos (I have tried 3 different versions, and yes it's atiflash not atiwinflash), I get the error "This program cannot be run in dos mode" WHAT THE EFF????
Are you using ATIWinflash or ATIFlash?

Use this : USBBOOT

And this: TPU ATIFLASH

I have win10, i can use dos atiflash on boot with a usb, and flash without problem.
Do not go with UEFI thing just boot on your USB, the normal way.
 

·
Registered
Joined
·
452 Posts
Quote:
Originally Posted by Wimpzilla View Post

Are you using ATIWinflash or ATIFlash?

Use this : USBBOOT

And this: TPU ATIFLASH

I have win10, i can use dos atiflash on boot with a usb, and flash without problem.
Do not go with UEFI thing just boot on your USB, the normal way.
ends up it was the atiflash version(s) I was using.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #3,456
Quote:
Originally Posted by madmanmarz View Post

ends up it was the atiflash version(s) I was using.
If you use atiflash.exe from AtiWinFlash in pure DOS it will not work, as that is windows command line version. So as you figured out you need atiflash pure dos version, aka v4.17 or v4.18.
 

·
Registered
Joined
·
509 Posts
Little update on modding the bios:
I get bit frustrated and i need some help there because i got stuck.
First i noticed that if i remove any limit, so TDP/PL/TDC/OCP/OVP etc, the card oc fine until i give +150mv, then the VCCD stop rising and get stuck to 1.25v.
Since +150mv allow me a 1250mhz clock, then i'm convicted i can go higher but the VCCD refuse to follow.
There is another protection that came out when there is no protection that lock the VCCD to 1.25v??!!

Editing the OCP/OVP/UVP to higher limits instead of ff, brick the card.
I tried 62A max limit OCP with +400mv OVP and default UVP, card refuse to boot.
TDP/PL/TDC 300/300/280.

Another thing quite strange, if i flash a different bios and do not power down the vrm carefully, i mean shut down completely the pc from the power array, i do not get the new vrm bios settings.
Flashed a couple of bios where unstable, flashed the 390 mod was unstable too, generally it is not. Power down the pc, leave it a couple of hours shut down, boot and oc stable again, lol?!
wheee.gif


Last thing, the lower VCCD i need now compared before the HG10 mod, is not due to the bios modding as i though.
As i tried again a stock bios, so the gains come from the HG10 mod itself, loosing almost 5°/10° on gpu is a lot.
I think i fully recover the contact between the IAO and the gpu, happy with that.
smile.gif
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #3,458
@Worldwin

That ROM has no VDDC offset.

@Wimpzilla

IR3567B only gets reprogrammed by ROM when you do full power down and up.

When you get VDDC as 1.25V what is VID of GPU?
 

·
Registered
Joined
·
330 Posts
Quote:
Originally Posted by gupsterg View Post

@Worldwin

That ROM has no VDDC offset.

@Wimpzilla

IR3567B only gets reprogrammed by ROM when you do full power down and up.

When you get VDDC as 1.25V what is VID of GPU?
I thought offset was the change in voltage at a specific state eg. mine at dpm goes from ~1.078mv-1.108mv at 100% load. I wanna squish that range down to the lower half and see if it reduces power consumption.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #3,460
@Worldwin

VDDC Offset reduces or increases voltage to all DPM states and not a specific DPM state.

VID for a DPM reduces or increases voltage to a specific DPM state.

The range you are seeing of VDDC (ie ~1.078mv-1.108mv at 100% load) is normal, I know no suggestion to give regarding "squish that range to the lower half".

It does not matter if ROM uses EVV VID for a DPM, manually set VID for a DPM and any of those combined with positive/zero'd/negative VDDC offset, you will get a range of VDDC even if GPU is not changing DPM states/clocks.
 
3441 - 3460 of 4332 Posts
Top