X1900gt Unlock Bios Found!!!

[tofu]

Open your bios up with RaBiT 2.1. Under clocks tab you'll be able to change voltages. Raise it by 10%, save it, then flash it. If you put a fan on the vregulators like I did with my x1800, running 1.55V is possible.

RaBiT doesn't let you change the voltage by more than 10% at a time so you'll need to save the bios, open and apply an extra 10% if you need more juice.

BTW, where did you get your card?

Yes I know the voltage trick, but I never tried it.

Best Buy.

 

[tofu]

my card does 749/945 on 1,525 vgpu, will try 1,55 with a 120mm on that volt reg

I hope you can hit 800 core with 1.55!

 

[sleepeeg3]

Oh thats cool.. Hey why doesn't this work: Take an X1900XTX bios and open it with RaBiT2.1, change memory size to 256 and change the memory timings to the same on your gt bios...

I would if I could. The RaBiT author can not figure out what hex code needs to be modified to change the memory size on the XT/XTX from 512->256, because there is no point of reference.

then the probability of finding one that unlocks now is slim to none...

Based on...? Pessimism doesn't help this thread at all.

Well I finally plugged in the GT and noticed we can't pull the fuse data, so so much for that. There are a few other things I am going to try.

tofu, did you ever post your beta bios? Can you post it again?

 

[tofu]

I would if I could. The RaBiT author can not figure out what hex code needs to be modified to change the memory size on the XT/XTX from 512->256, because there is no point of reference.

Based on...? Pessimism doesn't help this thread at all.

Well I finally plugged in the GT and noticed we can't pull the fuse data, so so much for that. There are a few other things I am going to try.

tofu, did you ever post your beta bios? Can you post it again?

Below is the BIOS from my card, USE AT OWN RISK.


EDIT: I noticed that my "beta BIOS" is a tad smaller than the other BIOSes other X1900GT owners were posting. Hmmm...

 

[laydback]

Tofu, so was I right about the hex code?

Below is the BIOS from my card, USE AT OWN RISK.


EDIT: I noticed that my "beta BIOS" is a tad smaller than the other BIOSes other X1900GT owners were posting. Hmmm...

 

[OUTSIDE]

who is the first to try the tofu's bios in his x1900gt that can't unlocked?

ByE!

 

[Migou]

who is the first to try the tofu's bios in his x1900gt that can't unlocked?

ByE!

I have already flashed this bios 1 week ago on my HIS :

"After the reboot i get a black screen (No signal) this append only when the ATI driver load.

To get back my original bios I have to restart in safe mode and remove the ATI driver,
restart normaly , flash back and reinstall the ATI driver. "

http://forums.techpowerup.com/showthread.php?t=13191&page=2

 

[hot_fifty]

my card can max 775/945 on max 1,575 need more vgpu on watercooling with xtreme air cooling on volt reg

 

[hot_fifty]

load temp is 45 on vgpu and 54 on volt reg xtreme air

 

[tboy2k5]

Any news on GB_PIPE_SELECT ( 0XAD4E ) ? Is it still pending ?


Here is my bios ...

 

[hot_fifty]

nope, I have the same

 

[tboy2k5]

nope, I have the same

What do you mean ? It is still pending or not possible to unlock ?

Thanks hot_fifty ....

 

[tofu]

Tofu, so was I right about the hex code?

I donno yet, this is the one where I change bit 19 into 18.

 

[sleepeeg3]

Guys I want to go over what I was looking at to show you where the most likely areas probably are that a bit or bits need to be changed to do anything.

1. X1900GT
2. X1900GT Beta (Tofu's)
3. X1900XT


The image below reverses the bit order and shows you where the DeviceID and all that is being pulled from RaBiT and ATiTool:


Some things I have tried:
1. Created a new bios for each bit difference between the GT and the XT up to the first byte on the third line and tried to flash.
2. Changed the DevID and the 1890/1990 byte to match the beta bios.
3. Changed the header of the GT bios to match the header of the beta bios
4. Changed the header of the GT bios to match the header of the XT bios
5. Changed the DevID via RaBiT (does not match correctly 7249 or 724A -> 7248, but fuses are changed that still allow it to boot)
6. Looked at the locations where the fuse data was changed in RaBiT and matched the fuses from the beta bios to the GT.
7. Multiple in between combinations

RaBiT allows you to modify the mem size of the X1900GT to 512, so this may be where the mem size is on all of the cards.

I did not bother modifying that bit, but RaBiT did not get the DevID right when modded, so I would not completely trust it.

So far nothing has worked, but there are still a number of bits I have not tried modifying. All the mem latency changes are down at the end, so there are alot of areas down there where a pipe mod could be hiding...

RAMBIOS is fairly reliable in checking whether a bios will work before you flash, so that saved me alot of time. When the flash went bad, ATIFlash version 3.14 + a PCI card was a 100% successful recovery method.

Every single bios with 16 pipes has a DevID ending in "7XX9" or "7XXB" and an 1890 byte. Any time I tried to change either or both, it would not post. RaBiT allows you to set a 7248 DevID/1890 byte, but it also changes the fuse information to compensate and I guess 7248 is not a valid ATi DevID because it does not show up. I think what it will take is finding out what it needs to set a 7249 or 724A device code with the correct fuses.

 

[tofu]

Guys I want to go over what I was looking at to show you where the most likely areas probably are that a bit or bits need to be changed to do anything.

1. X1900GT
2. X1900GT Beta (Tofu's)
3. X1900XT


The image below reverses the bit order and shows you where the DeviceID and all that is being pulled from RaBiT and ATiTool:


Some things I have tried:
1. Created a new bios for each bit difference between the GT and the XT up to the first byte on the third line and tried to flash.
2. Changed the DevID and the 1890/1990 byte to match the beta bios.
3. Changed the header of the GT bios to match the header of the beta bios
4. Changed the header of the GT bios to match the header of the XT bios
5. Changed the DevID via RaBiT (does not match correctly 7249 or 724A -> 7248, but fuses are changed that still allow it to boot)
6. Looked at the locations where the fuse data was changed in RaBiT and matched the fuses from the beta bios to the GT.
7. Multiple in between combinations

RaBiT allows you to modify the mem size of the X1900GT to 512, so this may be where the mem size is on all of the cards.

I did not bother modifying that bit, but RaBiT did not get the DevID right when modded, so I would not completely trust it.

So far nothing has worked, but there are still a number of bits I have not tried modifying. All the mem latency changes are down at the end, so there are alot of areas down there where a pipe mod could be hiding...

RAMBIOS is fairly reliable in checking whether a bios will work before you flash, so that saved me alot of time. When the flash went bad, ATIFlash version 3.14 + a PCI card was a 100% successful recovery method.

Every single bios with 16 pipes has a DevID ending in "7XX9" or "7XXB" and an 1890 byte. Any time I tried to change either or both, it would not post. RaBiT allows you to set a 7248 DevID/1890 byte, but it also changes the fuse information to compensate and I guess 7248 is not a valid ATi DevID because it does not show up. I think what it will take is finding out what it needs to set a 7249 or 724A device code with the correct fuses.

I think the deviceID you meant is NOT 7248, in fact it is 724B

 

[hot_fifty]

htkml is still testing... damn, what is he doing?

 

[dolf]

He is shooting at random .

 

[hot_fifty]

random?

 

[chron]

Is it possible that when ATITool says you have 16 pipes enabled, you really only have 12 working?

 

[sleepeeg3]

I think the deviceID you meant is NOT 7248, in fact it is 724B

No, I didn't. In the actual bios hex code it is listed as 7248. I finally got it to post with 724A/1890 in the code, but ATITool/RaBiT still lists it as 724B.

Is it possible that when ATITool says you have 16 pipes enabled, you really only have 12 working?

It's always been reliable in the past... I dunno. I think however Wiz is reading it, it is what is actually posting in Windows. I was wondering that myself, but you'd have to ask him.

 

[tofu]

No, I didn't.

It's always been reliable in the past... I dunno. I think however Wiz is reading it, it is what is actually posting in Windows.

My bad.

 

[chron]

what do the stars mean under our names? how handsome we are?

 

[W1zzard]

No, I didn't. In the actual bios hex code it is listed as 7248. I finally got it to post with 724A/1890 in the code, but ATITool/RaBiT still lists it as 724B.

It's always been reliable in the past... I dunno. I think however Wiz is reading it, it is what is actually posting in Windows. I was wondering that myself, but you'd have to ask him.

the number of pipelines = rasterizers (not shaders) is read directly from the gpu. you can easily verify pipeline changes by running the 3dmark fillrate test

 

[dolf]

The dispute here remembers me my enthusiasm when I saw the RivaTuner core-register values which later on I realized are the same with the FUSE values in ATITool (W1zzard told me that it were "secret" up to that moment and reveals the combination for obtaining them - THANKS). These simple values (which was hard to understand in the begining are very reliable tool to estimate whether some core can be unlocked to 8, 12 or 16 pipes). Nowadays we know much more about the meaning of these values and knowing more I would like to help to all of you which are trying to unlock pipes on X1900GT. But you should know that most of the cores on these cards are with locked quads (on hardware level) and no BIOS flash can get them back. Otherwise these cores would be used for 16 pipe cards! Of course there will be unlockable cores in X1900GT family but they will be with limited number.

Everybody who want's to unlock pipes on X1900GT with BIOS modding MUST have unlockable core. Only then it is worth to search the BIOS for pipe number information (if we accept that isn't situated like in X1800 BIOS - 0x7B). But the unlocking of one X1900GT just confirms that. The only way to estimate the core unlocking potential now is the GB_PIPE_SELECT and we have to learn what the values mean.

I would join the efforts for unlocking of X1900GT if there is people which understand the above. All of you which will receive cards with unlockable cores GB_PIPE_SELECT=0x3FE4 should try the pipe mod (It should be successfull in 99%). If you can't unlock with these GB_PIPE_SELECT=0x3FE4 it means that the pipe number 1/2 bit is not situated in 0x7B and we have to look for other possibility.

Trying various BIOS modifications on locked (on hardware level) cores is waste of time

 

[htkm1]

Going back on topic...

htkml and Changis, do your cards unlock?

Hi tofu,
Sorry for the late reply. been busy at work. As for flashing the mod bios, I think I'll give it a pass. since mine is a R580XL(LE) core and I can't afford to get a new one if I killed it.

Most likely I'll run it will watercooling and see how far it can be clocked using the ATITool 25.beta GT. This X1900GT runs really hot IMO. So if watercooling if not enough I'll wack a 170watt peltier on a GPU water block and run it. Voltage wise increasing up to 1.7~1.8v should be OK if running water or peliter or the core. I'll post my results in the forum.

Regards,
htkm1