(omg)vflash | Fully Patched nvflash from X to Ada Lovelace [v5.780]

[Deleted member 218758]

after swapping the UEFI GOP module:

GOP changes have a checksum
Lost_N_Bios figured this out.
if you used a tool and automated ~ yet still fails, then its the GopUpdater

If you did it by hand, then its a user error.

 

[yoloSager]

GOP changes have a checksum
Lost_N_Bios figured this out.
if you used a tool and automated ~ yet still fails, then its the GopUpdater

If you did it by hand, then its a user error.

by tool

 

[Deleted member 218758]

by tool

Can you attach a
-6 gopmod.rom -L log.txt
of it ?
Curious if it breaks due to SIG errors (file denied by falcon)
or it breaks due to both tools being "dumb"

 

[PanosX]

same problem with my 3090
more people have the same problem with ampere architecture

same error 3080

Can you attach a
-6 gopmod.rom -L log.txt
of it ?
Curious if it breaks due to SIG errors (file denied by falcon)
or it breaks due to both tools being "dumb"

 

[Deleted member 218758]

same error 3080

EEPROM unsupported - rather failed to read out, because SFDP deviceID is unsupported
created a chain that causes wrong falcon version to be loaded, creates a chain that falcon can not do anything
~failure~

Soo the issue is the GopUpdater, how ever they control at target GOP package , falcon ID
Idk how.
But i've never seen this.

Or ....
It can be those small EEPROMs that 3090 FE's had - which where completely unsupported.
In any case, GOP loaded unknown falcon version, and GPU didnt know what to do ~ soo like a pyramid, everything failed past that.
No falcon, no flashing.

Very bizarre.
Can you try to find a bios on techpowerup with the same subsystem ID , pretty much a copy but with an earlier vbios version ?
Then try to normally flash that ?

I can not differentiate "strange foreign eeprom" vs "strange GOP that was not made for this device"
It can be that eeprom is pulled high and so is undetected till X. It's not uncommon.
omgvflash tried to detect it then as erased eeprom and still flash, but falcon could not be initialized as model detected for it simply was the wrong
Soo flash was halted as Falcon was halted and your flash command result in ~nothing~
Super bizarre, but i don't think it was nvflash's fault here.
I think it was the file reporting and requesting something strange.

---

We could also lean extremely far out of the window, put tinfoil hat's on
And say that GOP was expecting falcon v7, which didn't exist on current nvflash & nothing can be done

Any case, target falcon fw version was not found
Soo everything past that failed like a pyramid scheme ~ as it loaded falcon v4 for target SKU, but could not do anything because it targeted a strange new version.
Super bizarre.

No idea what's exactly going on. I would need to speak with the Dev, how he updated it and what he pulled.
But i blame the GOP for 75-80% , and strange eeprom as rather 20% chance ~ to be what's going on here.
I need more information about the card, to make up my mind
It wouldn't be unlikely that Nvidia™ proactively tries to prevent all options past their breach-accident and rewrite everything.
Although we don't work against them, it wouldn't honestly wonder me. Depends on GOP date and everything around it.

---

Update:

So indeed it can be the file maker that is to blame
and what happens is it roundrobin increasing version from ucode0-5 ~ still fails detection as interface is on another hub
Then falling back on SKU identification which is ucode4 , thinking eeprom is just erased hence failed detection ~ moving on but still failing.

Somebody needs to ping or pull GopUpdater dev
To change register in ROM to SPI1, from 0
or from 2 to 1

Very likely its that
Rom not recognizing EEPROM location, due to bug? useredit or due to strange too new (unfitting) GOP blob.
Nevertheless try to flash something else on your GPU - that is as close as possible , like a VBIOS downgrade or upgrade @PanosX
Just to factcheck. Of course with -L log.txt
https://lore.kernel.org/lkml/abf7dc0e-df13-25ae-83b7-204eee0189d4@linaro.org/T/ SFDP is an interface for SPI lookup. // Serial Flash Discoverable Parameter

 

[PanosX]

Sorry my english is little bit bad!!!!!you need something about my gpu?

EEPROM unsupported - rather failed to read out, because SFDP deviceID is unsupported
created a chain that causes wrong falcon version to be loaded, creates a chain that falcon can not do anything
~failure~

Soo the issue is the GopUpdater, how ever they control at target GOP package , falcon ID
Idk how.
But i've never seen this.

Or ....
It can be those small EEPROMs that 3090 FE's had - which where completely unsupported.
In any case, GOP loaded unknown falcon version, and GPU didnt know what to do ~ soo like a pyramid, everything failed past that.
No falcon, no flashing.

Very bizarre.
Can you try to find a bios on techpowerup with the same subsystem ID , pretty much a copy but with an earlier vbios version ?
Then try to normally flash that ?

I can not differentiate "strange foreign eeprom" vs "strange GOP that was not made for this device"
It can be that eeprom is pulled high and so is undetected till X. It's not uncommon.
omgvflash tried to detect it then as erased eeprom and still flash, but falcon could not be initialized as model detected for it simply was the wrong
Soo flash was halted as Falcon was halted and your flash command result in ~nothing~
Super bizarre, but i don't think it was nvflash's fault here.
I think it was the file reporting and requesting something strange.

---

We could also lean extremely far out of the window, put tinfoil hat's on
And say that GOP was expecting falcon v7, which didn't exist on current nvflash & nothing can be done

Any case, target falcon fw version was not found
Soo everything past that failed like a pyramid scheme ~ as it loaded falcon v4 for target SKU, but could not do anything because it targeted a strange new version.
Super bizarre.

No idea what's exactly going on. I would need to speak with the Dev, how he updated it and what he pulled.
But i blame the GOP for 75-80% , and strange eeprom as rather 20% chance ~ to be what's going on here.
I need more information about the card, to make up my mind
It wouldn't be unlikely that Nvidia™ proactively tries to prevent all options past their breach-accident and rewrite everything.
Although we don't work against them, it wouldn't honestly wonder me. Depends on GOP date and everything around it.

---

Update:
View attachment 311058
So indeed it can be the file maker that is to blame
and what happens is it roundrobin increasing version from ucode0-5 ~ still fails detection as interface is on another hub
Then falling back on SKU identification which is ucode4 , thinking eeprom is just erased hence failed detection ~ moving on but still failing.

Somebody needs to ping or pull GopUpdater dev
To change register in ROM to SPI1, from 0
or from 2 to 1

Very likely its that
Rom not recognizing EEPROM location, due to bug? useredit or due to strange too new (unfitting) GOP blob.
Nevertheless try to flash something else on your GPU - that is as close as possible , like a VBIOS downgrade or upgrade @PanosX
Just to factcheck. Of course with -L log.txt
https://lore.kernel.org/lkml/abf7dc0e-df13-25ae-83b7-204eee0189d4@linaro.org/T/ SFDP is an interface for SPI lookup. // Serial Flash Discoverable Parameter

 

[Deleted member 218758]

Sorry my english is little bit bad!!!!!you need something about my gpu?

Tell dev to come over - issue is 90% his tool
Try to flash some similar official bios on your card - still give a new log.
GopUpdate bug is very very likely issue of this, but we need more data about your GPU (exact model)
Maybe give GPU-Z screenshot of official bios the card came with.

 

[yoloSager]

Tell dev to come over - issue is 90% his tool
Try to flash some similar official bios on your card - still give a new log.
GopUpdate bug is very very likely issue of this, but we need more data about your GPU (exact model)
Maybe give GPU-Z screenshot of official bios the card came with.

here is my log seems different.


also test with oldest bios for my gpu with no diff between my first log expect line 141
witch the in second log has this mismatch

Mismatch Type: Computed: 0x00030108
Mismatch Type: Adjusted: 0x00000000

and the bios ver in the file.

my bios VGA Bios Collection: Inno3D RTX 3090 24 GB | TechPowerUp

 

[Deleted member 218758]

here is my log seems different.

nvflashK support in his thread, not here.
but error says - vbios certification verification failed.
User edit ?

Bios file attempted to flash ~ has broken cert
Cert bypass mod is on nvflash to give full update permissions, will never provide cert bypass for falcon.
// at least 99% chance will never ~ unless we have no other option and Nvidia goes rogue to customer.
// not worth the loss + penalties to consumers and to me

 

[PanosX]

3080 MSI SUPRIM X 10GB NO LHR

 

[yoloSager]

nvflashK support in his thread, not here.
but error says - vbios certification verification failed.
User edit ?

Bios file attempted to flash ~ has broken cert
Cert bypass mod is on nvflash to give full update permissions, will never provide cert bypass for falcon.
// at least 99% chance will never ~ unless we have no other option and Nvidia goes rogue to customer.
// not worth the loss + penalties to consumers and to me

sorry about that this is the correct log

 

[zxTheWolfxz]

nobody should be using any downloads provided by @Wotwow untill they can be vetted by a moderator

the files are signed by some skechy russian cheat site if that doesn't raise a red flag for you
well ye been warned

I seen that and only test on PC that has nothing on it but Folding at Home. lol
If they want to can put points to team 13285 I don't care.

 

[Deleted member 218758]

sorry about that this is the correct log

OK, soo part of the issue is my tool - not finding eeprom
endresult is the same sadly ~ bios cert flash error // file is unsigned by the GOPTool Dev

Can you try to reflash your exported bios again - with my tool and provide log ?

.\omgvflash.exe --save backup_orig.rom
.\omgvflash.exe --protectoff
.\omgvflash.exe -6 backup_orig.rom -L reflash_log.txt

---

Unfortunate that i lack cards - else things would speed up a bit
But honestly, i'm also busy with other projects at the same time ~ soo having 3000/4000 series currently is not really helpful
Pascal foundation needs to be flawless, then Turing foundation needs to start work with certs.
3000/4000 series is far far away. Higher chance to bother boardpartners for signed XOC bios, vs count on me.
I don't know if till december 3000+ series will work. Normal flash i hope to bugfix with next update - advanced rebrands or biosmods, unlikely.
Currently support is too little, and higher priority projects exist. Also income is zero. This is an open-research project after all.

 

[yoloSager]

OK, soo part of the issue is my tool - not finding eeprom
endresult is the same sadly ~ bios cert flash error // file is unsigned by the GOPTool Dev

Can you try to reflash your exported bios again - with my tool and provide log ?

---

Unfortunate that i lack cards - else things would speed up a bit
But honestly, i'm also busy with other projects at the same time ~ soo having 3000/4000 series currently is not really helpful
Pascal foundation needs to be flawless, then Turing foundation needs to stark work with certs.
3000/4000 series is far far away. Higher chance to bother boardpartners for signed XOC bios, vs count on me.
I don't know if till december 3000+ series will work.
Currently support is too little, and higher priority projects exist. Also income is zero. This is an open-research project after all.

fist 2 that i tested failed. you can see the logs

 

[Deleted member 218758]

fist 2 that i tested failed. you can see the logs

mm mm , issue is my eeprom unlock and falcon v4 "unlock" from 2000 series.
i'll fix this
Thank you @yoloSager

EDIT:
Idk when, as i don't have 2000/3000 series cards
But i'll try to fix it
At worst we'll lose downgrade ability temporary in the next version with the exchange of bug fixes and more perm
(hopefully full access to signed bios CID override ... i have stupid-me bugs to fix)

EDIT2:
Hopefully by that time, also some public-alpha version of a viewer, would be finished.
I can't code, soo it's not all on me.

 

[Wotwow]

I don't understand why people flash cards for high power. The graphics card board is designed to not break the card even if the card was flashed with a more powerful BIOS. I understand that you need to slightly increase the boost and increase the power by more than 6 percent. Increasing the power of the card, in which the VRM becomes unstable and overheats a lot, is a bad result. In video cards, boost is 2.0, 3.0, which vice versa gives less gain from higher watts if you don't get GPU binning gold.

For example, my card is 250W. I would gladly increase the Watts of a standard video card by 6% to avoid overheating and breakage.
250 Watts + 6% = 265 Watts

 

[rusTORK]

I don't understand why people flash cards for high power.

Newbie here? Open any 3DMark\Superposition leaderboard.

It's too long to explain, but in short - overcoming limitations (all kinds: by manufacturer, by AIBs, e.t.c.).

Hardware can't be damaged if properly cooled.

 

[Wotwow]

Newbie here? Open any 3DMark\Superposition leaderboard.

It's too long to explain, but in short - overcoming limitations (all kinds: by manufacturer, by AIBs, e.t.c.).

Hardware can't be damaged if properly cooled.

You bought nonsense and want to be in the top for overclocking. You're a newbie. Before buying a card for overclocking, people look at the VRM. VRM capacity +6/+12 % and then your overclocking will worsen. These 6-12% are the difference in the power regulators of your graphics card. I used to be an overclocker and took first place when nvidia didn't disable it and forced people to buy new VRMS evga-epower-V.

 

[Mirbg]

nvflashK support in his thread, not here.
but error says - vbios certification verification failed.
User edit ?

Bios file attempted to flash ~ has broken cert
Cert bypass mod is on nvflash to give full update permissions, will never provide cert bypass for falcon.
// at least 99% chance will never ~ unless we have no other option and Nvidia goes rogue to customer.
// not worth the loss + penalties to consumers and to me

Hey @Veii , thanks for your great job!
Can you tell me if I can flash a modded vbios onto my Clevo RTX 2080? ( I’d like to rise the power from 150W to 200W with Pascal TDP tweaker)

 

[heavyarms2112]

I don't understand why people flash cards for high power. The graphics card board is designed to not break the card even if the card was flashed with a more powerful BIOS. I understand that you need to slightly increase the boost and increase the power by more than 6 percent. Increasing the power of the card, in which the VRM becomes unstable and overheats a lot, is a bad result. In video cards, boost is 2.0, 3.0, which vice versa gives less gain from higher watts if you don't get GPU binning gold.

For example, my card is 250W. I would gladly increase the Watts of a standard video card by 6% to avoid overheating and breakage.
250 Watts + 6% = 265 Watts

Some people think everything is sw limited . Better bios magically make the gpu faster.
Many are gonna end up with blown up VRMs or impact longevity of the components.

Anyone who flashes a higher end GPU bios should verify the components on PCB (on existing GPU) to match or at least is in same league of that higher end gpu.

Newbie here? Open any 3DMark\Superposition leaderboard.

It's too long to explain, but in short - overcoming limitations (all kinds: by manufacturer, by AIBs, e.t.c.).

Hardware can't be damaged if properly cooled.

This is incorrect. Not everything that needs cooling is monitored with sensors. Not all mfg provide sufficient cooling on all heated components.

Eg. Not all mfg provide active cooling contact on VRMs. VRMs can operate at high temps. Some even at 100-110C but not without impacting their efficiency. However, they're still in spec so low end cards skimp on these. If you overdrive them (increasing power limit for example) they're gonna run even hotter & might just die sooner.

Heck mfgs even tend to forget applying thermal pads on components that need active cooling.

 

[Wotwow]

Eg. Not all mfg provide active cooling contact on VRMs. VRMs can operate at high temps. Some even at 100-110C but not without impacting their efficiency. However, they're still in spec so low end cards skimp on these. If you overdrive them (increasing power limit for example) they're gonna run even hotter & might just die sooner.

They just reduce their graphics card boost when the VRM overheats.

 

[OneMoar]

stop with the non-sense most premium cards have lots of vrm overhead
try it on a cheap model well you get what you get ... but +50W isn't likely to cause any real harm with a proper fan profile

 

[Wotwow]

Update:
So indeed it can be the file maker that is to blame
and what happens is it roundrobin increasing version from ucode0-5 ~ still fails detection as interface is on another hub
Then falling back on SKU identification which is ucode4 , thinking eeprom is just erased hence failed detection ~ moving on but still failing.
Somebody needs to ping or pull GopUpdater dev
To change register in ROM to SPI1, from 0
or from 2 to 1

How programmers divided GPIO

 

[Mussels]

It can use that one weird 125% power limit vBIOS, and it does work there, but that caps the fan speed pretty low, so it throttles.
Also if this is such a big issue, where are the shunt mod failure videos?

Because the BIOS was made for an entirely different card with different fans.

You can unlock a 150W card to 1000W, and it'll just overheat and throttle - people doing these shunt mods bypass more than a BIOS limiter, and they tend to do it on cards that are known to handle it instead of random GPUs.

 

[zxTheWolfxz]

Please stay on topic, this thread is about OMGVflash and giving info to help get it working for as many as he can, thanks.