G92-270-A2 to G92-400-A2

************************************************** ********************************

WARNING, WHILE TRYING TO UNLOCK MY VIDEO CARD, I HAVE BROKEN IT AND CANNOT FIX IT.

*****Disclaimer*****

Do any of the following at your own risk.

I will not be held responsible for any more broken video cards, etc. etc.


************************************************** ********************************


Hi,


I have been looking into the possibility of "Unlocking" the G92-270-A2 into the G92-400-A2.

Here are my results so far :-

I got an Xpertvision 8800GT 512mb from http://www.ebuyer.com/product/139632 , now sadly sold out.

Made a bootdisk with drdflash.exe from www.bootdisk.com and copied over all the files for nvflash 5.57 to work (nvflash.exe and cwsdpmi.exe).

I made a backup of my bios file with this command :-

nvflash -b backup.rom

Then in WinXP with NiBiTor v3.7, I changed the Device ID from 0611 to 0600 the same as G92-400-A2 (8800GTS).

I made a G92-270-A2_id_0600.rom and flashed to the video card as follows:-

nvflash -4 -5 -6 G92-270-A2_id_0600.rom and said "Y" to all error messages. Nvflash said that everything was verified OK so I rebooted. The Device ID had stayed the same 0611 ???

OK, now I flashed a genuine "Palit 8800GTS G92-400-A2 512mb" bios file to my card (from a well known tech vga bios collection), same settings

nvflash -4 -5 -6 G92-400-A2.rom and reboot.

Now I have device ID 0610 and WinXP does not recognise the video card, so I hack the drivers and call new card "8800 GT Ultra" and reboot. Video card gets recognised and installs drivers. I reboot. Computer starts OK, loads WinXP OK, but at Desktop; video corruption, flashing pixels, then boarders, then blocks, then system freezes. I Power Down.

Somehow when I flash the bios the Device ID does not change properly ??

My thinking is that either the bios is not "Fully" flashing or the Device ID is set somewhere else on the video card ?

Any thoughts anyone.

Gone for beer after this.

Hal2710

Hi , :-(

"Curiosity Killed the Cat" or "The Devil will make work for idle hands to do !"


Please see Disclaimer at the top of 1st Post.


I will try to explain :-

Quote:

---

Originally Posted by hal2710

Now I have device ID 0610 and WinXP does not recognise the video card, so I hack the drivers and call new card "8800 GT Ultra" and reboot. Video card gets recognised and installs drivers. I reboot. Computer starts OK, loads WinXP OK, but at Desktop; video corruption, flashing pixels, then boarders, then blocks, then system freezes. I Power Down.

Somehow when I flash the bios the Device ID does not change properly ??

My thinking is that either the bios is not "Fully" flashing or the Device ID is set somewhere else on the video card ?

---

nvflash /? give you a list of options and extra switch settings.

I played with one of them --guid, "A 16 digit hex value" ?? . For me I thought this must be the Device ID values so I tried

nvflash --guid=10DE060010DE0502

nvflash said that it set the value ok, so I reboot, but still device id 0610.

nvflsah -v

I can't do a full report, but I think this lists everything nvflash knows about the video card. I'm not sure if this is from the video card or from nvflash, if you see what I mean. However as I had flashed my 8800GT(G92-270-A2) with an 8800GTS (G92-400-A2) bios file, I was pleased to see that "nvflash -v" said the GPU was infact G92-400-A2. To confirm this I flashed my backup.rom and then "nvflash -v" said G92-270-A2.

OK I must be getting somewhere, I can flash my 8800GT with the bios of an 8800GTS, but the video card will only output Device ID 0610 not 0600 for full 8800GTS recognition.

nvflash -a

The output of this with my original 8800GT bios is :-

G92-270 (10DE,0611,0000,0000) H:NRM B:01,PCI,D:00,F:00

After bios flash of 8800GTS file I get:-

Gxx (0610h) (10DE,0610,10DE,0502) H:NRM B:01,PCI,D:00,F:00

So, I am thinking "It must be the video card is outputting the device, It is not being read from the bios !"

Where could the device id be kept on the video card ??

--guidsource "Option must be either main or dedicated"

Bingo I though, there could be a dedicated something on the video card that is outputting the Device ID.
Further reading said that option -1 is the same as --guidsource. Here goes......

nvflash -1 main -4 -5 -6 8800GTS.rom

Yes, Yes ,Yes I reply to nvflash, blaa blaa, Bios flashed OK. So I reboot and system hangs at POST, fans spin up, but not VDU, 20 seconds 1 single shortish beep, 10 seconds same beep, etc for 5 Min's. Power down.

I install a PCI videocard (S3 Virge), Clear CMOS and reboot. Thank you for Default settings as first video card is set to PCI, System reboots and I reflashed my backup.rom, reboot and my 8800GT is saved.

Stops for Dinner. Dead video card explanation to follow.

Hal2710

That's sad man :(

Rbs.
:wave:

The cores are laser cut so you can't unlock them I think

The "Laser cut" theory is bull:banana::banana::banana::banana::banana:. It'll take way too much time to do that for every chip. There's probably a few NAND flash cells built into the chip, preprogrammed from the factory that cannot be access through regular means which control which parts are disabled.

All you've done is make the drivers and VGA bios think that the chip has extra shaders which are still disabled so when it tries to initialize or access them, it screws up. The artifacting on desktop is more likely a memory issue.

Hi All,

Thanks for reply's.

Been trying to fix dead card but it is very dead.

Going to bed now, See you tomorrow with Episode Three...!

"The Case of the Mask Avenger"


Hal2710

****Please read Disclaimer on 1st Post ********


PS..... Remember :-

A. It's only Money.

2. You can't take it with you.

D. £180/£2.35= Only 76.59 pints of Stella.

Good Night All

Quote:

---

Originally Posted by HKPolice

The "Laser cut" theory is bull:banana::banana::banana::banana::banana:. It'll take way too much time to do that for every chip. There's probably a few NAND flash cells built into the chip, preprogrammed from the factory that cannot be access through regular means which control which parts are disabled.

All you've done is make the drivers and VGA bios think that the chip has extra shaders which are still disabled so when it tries to initialize or access them, it screws up. The artifacting on desktop is more likely a memory issue.

---

Why would machined package cutting take "way more time" then flashing GPU NAND memory?

Hi HKPolice,

Quote:

---

Originally Posted by HKPolice

There's probably a few NAND flash cells built into the chip, preprogrammed from the factory that cannot be access through regular means which control which parts are disabled.

---

This is great stuff, we can talk for hours on this !

Quote:

---

Originally Posted by HKPolice

All you've done is make the drivers and VGA bios think that the chip has extra shaders which are still disabled so when it tries to initialize or access them, it screws up.

---

Yes I agree with you.

Quote:

---

Originally Posted by HKPolice

The artifacting on desktop is more likely a memory issue.

---

Well, before I flashed the card with the G92-400-A2.rom, I checked the speed settings 650-1625-972 and I thought that this would be OK.
I have overclocked to 702-1728-999 with no problems. However I have now read that there could be different memory timings, so this could be causing the artifacts.

• How did I kill my Xpertvision 8800GT ???

This was my train of thought:-


I found this in Riva Tuner 2.06 and thought that this was the setting that I had to over come !

• $ffffffffff ----------------------------------------------------------------
  $ffffffffff NVIDIA specific display adapter information
  $ffffffffff ----------------------------------------------------------------
  $0100000000 Graphics core : G92 revision A2 (112sp)
  $0100000001 Hardwired ID : 0610 (ROM strapped to 0611)
  $0100000002 Memory bus : 256-bit

• $0100000001 Hardwired ID : 0610 (ROM strapped to 0611)

• (ROM strapped to 0611)

OK, nvflash says something about straps ??

straps m <straps> Change soft straps. format is: AND Mask 0, OR Mask 0, AND Mask 1, OR Mask 1

The 0610 is being masked to show 0611. I can make the videocard output 0610 but not 0600, the difference is xx1x. "AND Mask 0, OR Mask 0" ????????

I cannot change the 3rd "Mask" ??, I thought I need to flash "OR Mask 2" !!

So I did this :-

nvflash -4 -5 -6 -j -r -m= OR Mask 2 G92-400-A2.rom

nvflash went through the usual procedure, usual or you sure Y/N etc. etc.

Flashing..........., Verifing.............,OK

So I reboot and the system will NOT Post.

I have still got the PCI video card loaded, I have cleared CMOS, I have taken 8800GT out. This way the Computer will Post.


The only way to get computer to Post with 8800GT in PCI-E is to disrupt the 6-pin power to the card and then plug it back quickly.


This way the system will boot, but nvflash says "NO NVIDIA CARD FOUND".


So that is it 8800GT is DEAD !! , Please read Disclaimer at the top of 1st Post.





In carpentry they say "Measure Twice, Cut Once." Of Bios flashing they say "If it ain't broke don't fix it."


These are very wise words my friends, VERY WISE.


Off to the Pub.

Hal2710 (Retired)

Quote:

---

Originally Posted by wittekakker

Why would machined package cutting take "way more time" then flashing GPU NAND memory?

---

Do you have any idea how tiny the traces are inside a GPU? One tiny deviation and the GPU is ruined. There's also multiple layers of transistors and other devices in the chip, calibrating the laser to cut through only a single layer is also unrealistic. There is just too much risk and expensive equipment involved.

Whoever made up the 'laser cut' theory is full of S.H.I.T. Probably spread the rumor intentionally so that people would give up on trying to unlock things.

Quote:

---

Originally Posted by HKPolice

Do you have any idea how tiny the traces are inside a GPU? One tiny deviation and the GPU is ruined. There's also multiple layers of transistors and other devices in the chip, calibrating the laser to cut through only a single layer is also unrealistic. There is just too much risk and expensive equipment involved.

Whoever made up the 'laser cut' theory is full of S.H.I.T. Probably spread the rumor intentionally so that people would give up on trying to unlock things.

---

I did not mention any traces inside the GPU, I was talking about the GPU package. 8800GTS 512 and 8800GT 512 have G92 GPU's and probable all come from the same production line(s). After JTAG tests chips are separated:

good chips: G92 GTS
half good: G92 GT
-no match with real-life, just giving an example how chips are binned during manufacturing-

After that, the GPU gets either 8800GT or 8800GTS chip package.

I've compared GF8400 and GF 8500 GPU's, which both are G86 chips, their GPU packages don't match. Oké, that's no evidence of whatever I've mentioned above, it's just what makes most sense up to me.

Get a 2nd 8800GT, put the borked on in the slave position. Use nvflash to flash both.


Worked for me.


Btw, I've already tried unlocking this, since it is curious it is rom strapped on some 88GT but not others. My attempts failed. OH well.