X800Pro NON-VIVO -> x800XT PE

[ViperJohn]

Originally Posted by MagicHawk

You can find those BIOS here:

http://www.techpowerup.com/bios/

Okay. That is the GeCube x800Pro VIVO bios. It is the same as the Sapprire x800Pro VIVO bios
P/N 113-A26107-100 dated 5/11/04.

Viper

[ViperJohn]

Originally Posted by MagicHawk

Thanks for that bit if hope Viper!

If we pound on the issue long enough hopefully there will be Joy in the future. Pretty sure
the true function of the bridge cut (AA cores) is known now.

I have a BbATI x800Pro here that cleans spotlessly clean at 594/579 stone stock with the
fan locked up on high speed using Atitool and running the 113-A26105-100 XT-PE bios.
The card will run well over 600 core without lockup (but dirty) so Vcore and good cooling
will really turn the core on but the card will need that last pipe quad to truely shine.

Viper

[_TiTaNiuM]

So that bios is worthless then?... the GeCube

[ViperJohn]

Originally Posted by _TiTaNiuM

So that bios is worthless then?... the GeCube

As far as unlocking a non VIVO yes. It and the Sapphire VIVO bios are one in the same.

The key is the ASIC ID. If that can be changed the drivers won't need to be modded to get
around it. On x800Pro VIVO's with "AD" cores and no bridge cut the ASIC ID will change from
4A49 (x800Pro) to 4A50 (x800XT) and back again when the bios is flashed with the right flash
tool. With "AA" cores it won't.

Pretty sure bridge cuts are not pipe locks per say but register programming bit count locks. It
is similar to the resistor hardmod on the old 9500NP's but with a twist. When you moved the
resistor on the R300 cores it unlocked the ASIC ID register to reprogramming with the 9700
ASIC ID which you got with the flash to a 9700 bios. The x800Pro VIVO's with "AD" cores do
the same thing so the flash tools are capable of changing the ASIC ID's for R420 cores unless
they are "AA" cores. That leads you down a path that tells you the bridge controls weither
or not a register check bit(s) is set (it would have to be non volitile) that either allows or
doesn't allow the register holding the ASIC or maybe more correctly will hold the ASIC ID to
be programmed above a certain bit count or 100101001001001 in this case.

It would be a rather simple thing at the Fab during testing/binning to set a check bit(s), using
propriatary software, to prevent the the ASIC register bit count from being programmed high
enough to hold an XT ASIC ID of 4A50 then cut the bridge and mark the cores as AA's. When
the cores are in the hands of the user jumping the bridge on an AA core and flashing the card
doesn't get you anywhere. While the flash tools have the ability to change the ASIC ID they
do not have the ability to change the protection check bit(s) even if the bridge is jumped to
allow it/them open to change. That stops the flash tools ability to change the ASIC ID dead
in it's tracks.

One thing ATI isn't is dumb and they learned their lesson on the 9500NP well. ATI knew OC'ers
would try to try and jump the bridge then flash the card to XT spec's. By using a rather simple
(read cheap) protection method the flash tools do not have the ability to change you are still
locked out of having 16 pipes.

This is all open to speculation of course at this point and may be completely wrong but it
definately fits with what is currently known about the cards. From ATI's engineering standpoint
it is easy to do, requiring only a few seconds during test/binning and very effective. It also lends
itself well for when the 8 pipe x800se hit the streets. With two bridges cut that would indicate
a double lock on them...one to get to 12 pipe and another to get to 16 pipe which is extremely
doubtful they could run with clean anyway.

Ati's bean counters would like as well because it is cheap to impliment and perform on a per
part basis. When you are going to do something a few million times that becomes a key
issue as well.

Time will tell HOPEFULLY!!!

Viper

[Wilko042]

Nice theory. Do you think it would be possible to hex edit the bios and change all AA to AD and AD to AA, and/or 4A49 to 4A50 and 4A50 to 4A49 if they exist and then reflash? or am I just whistling in the wind.

[ViperJohn]

Originally Posted by Wilko042

Nice theory. Do you think it would be possible to hex edit the bios and change all AA to AD and AD to AA, and/or 4A49 to 4A50 and 4A50 to 4A49 if they exist and then reflash? or am I just whistling in the wind.

Like I said it fits in with what is known but it is just educated speculation right now.

The AA and AD just refer to the laser marking on the die's surface. That is permanent and
is just a way to ID the cores without getting the magnifiers out and looking for cut bridges.
The stuffing robots probably use it to prevent a 12 pipe core from going on an x800XT-PW
by accident during assembly.

If I am correct what you want to change would be the check bit(s) so the register that
holds, or will hold, the ASIC ID can be reprogrammed to hold a bit count high enough to
represent 4A50h when the card is flashed.

Viper

[HH_Quazi]

Sounds like we may getting somewhere, eh? Come on all you brainiacts, lets figure this one out. Viper John may be on to something here. One thing is for sure, if we don't, no one else will. Thanks to Viper for his unique insight to our dilema. Where there is a will, there is a way m8's. I'll volunteer my card for any trial runs at getting this to work. I've already performed to hard mod so no RMA for me. So if anyone can come up with a way to get our cards to do what they need to do to accept the XT PE's bios and/or unlock the other 4 pipes, let me know. I'll test for us. Thanks again Viper!!

[ViperJohn]

Originally Posted by HH_Quazi

Sounds like we may getting somewhere, eh? Come on all you brainiacts, lets figure this one out. Viper John may be on to something here. One thing is for sure, if we don't, no one else will. Thanks to Viper for his unique insight to our dilema. Where there is a will, there is a way m8's. I'll volunteer my card for any trial runs at getting this to work. I've already performed to hard mod so no RMA for me. So if anyone can come up with a way to get our cards to do what they need to do to accept the XT PE's bios and/or unlock the other 4 pipes, let me know. I'll test for us. Thanks again Viper!!

You shold know there are two Surface mount resistor "switch" location differences between
X800Pro's and X800Pro VIVO's and XT-PE's PWA's too. That is over and above the Pro's lack
of VIVO components. I changed the location of those part on my card and it did not unlock the
last pipe quad (or the ASIC ID register). If a way to unlock the ASIC ID to changes is found
those SMR's may still need to be moved to do it. ATI put in multiple layers of hardware, firmware
and software (driver) checks from the looks of it to keep the pro's 12 pipe.

Time will tell.

Viper

[Wombat]

If only W1zzard would share more of his knowledge. He already made this modded x800pro 16pipe bios with modified flashrom and knows to change the clocks into these ROMs.

Anybody has connections to him?

Wombat

[ViperJohn]

Originally Posted by Wombat

If only W1zzard would share more of his knowledge. He already made this modded x800pro 16pipe bios with modified flashrom and knows to change the clocks into these ROMs.

Anybody has connections to him?

Wombat

W1izaards card was a VIVO if I remember right. They are no problem. There were a few
very early Pro non Vivo's that would change too but that is history.

Viper

[Wombat]

But he may clarify what´s all about and why the newer ones don´t do anymore.
He somehow managed to do an own flashrom that opens the pipes on the vivos so he may know more or can second your correct looking findings.

i personal am interested in the clock rate change. He did it for a german Mod reseller he wrote in another forum but didn´t tell how

Wombat

[ViperJohn]

Originally Posted by Wombat

But he may clarify what´s all about and why the newer ones don´t do anymore.
He somehow managed to do an own flashrom that opens the pipes on the vivos so he may know more or can second your correct looking findings.

i personal am interested in the clock rate change. He did it for a german Mod reseller he wrote in another forum but didn´t tell how

Wombat

He modded Flashrom v2.37 so it would no longer preserve the ASIC ID register and could
change the ASIC ID on X800 Vivo's the way GVflash.exe could.

Changing the bios clocks is a matter of finding where they are set in bios code. What
would nice if W1zzard someone else with the skills would come up with an editor for the
x800 bios images to change the clocks in them like Radedit does in 9800 and earlier bios
image's. W1zzard has the knowledge, tools and skill to do it but having the time to do it
may be another matter entirely for him.

Unwinder and W1zzard are the two people with the skill and knowledge to pull off an unlock
if they have enough accurate information about the cards, differences and behavoir to
flashing. Have tried contacting unwinder with no luck. Am trying to contact W1zzard now.

Viper

[Wombat]

When we are at it (Editors), a litle warning:
Rabit looks promising cause it is still worked on. I tried to change the HardwreID on my X800 vivo Bios from 4A49 to 4A50 cause this field for the X800 series is open in Rabit 1.3.

It messed up my Bios completely

Wombat

[ViperJohn]

Originally Posted by Wombat

When we are at it (Editors), a litle warning:
Rabit looks promising cause it is still worked on. I tried to change the HardwreID on my X800 vivo Bios from 4A49 to 4A50 cause this field for the X800 series is open in Rabit 1.3.

It messed up my Bios completely

Wombat

Doesn't surprise me as changing the ASIC ID directly could have all kinds of weird consequences.
It isn't needed either. On the Vivo's I i've had come through me the ASIC ID would change to 4A50
and back to 4A49 freely when flashed with the right flash tool. Haven't seen a Vivo yet that didn't
have an "AD" core and intact bridges but I have read they are out there.

Viper

[Wilko042]

They say misery likes company, so I thought I would let you know that my HIS X800 Pro non-VIVO failed the bridge mod


However, I did flash it to a HIS X800 XT BIOS and found that my 3D Marks went up from 11,308 to 11,783 with both runs at 551c/551m Furthermore, after reinstalling the HSF I can now get 575c/561m and a 3D Marks of 11,938. When I took the HSF off there was probably too much thermal paste on and I cleaned it up and replaced it with AS5.

Anyone else see a performance boost using the XT BIOS?
____________________________
P4C800-E Deluxe [email protected]
HIS X800 Pro non-VIVO 575c/561m
2x512MB Ballistix 2.5,2,2,5
2xWD 36G Raptors Raid 0

[HH_Quazi]

Originally Posted by Wilko042

They say misery likes company, so I thought I would let you know that my HIS X800 Pro non-VIVO failed the bridge mod


However, I did flash it to a HIS X800 XT BIOS and found that my 3D Marks went up from 11,308 to 11,783 with both runs at 551c/551m Furthermore, after reinstalling the HSF I can now get 575c/561m and a 3D Marks of 11,938. When I took the HSF off there was probably too much thermal paste on and I cleaned it up and replaced it with AS5.

Anyone else see a performance boost using the XT BIOS?
____________________________
P4C800-E Deluxe [email protected]
HIS X800 Pro non-VIVO 575c/561m
2x512MB Ballistix 2.5,2,2,5
2xWD 36G Raptors Raid 0

Am I reading you right. Your non-VIVO was able to accept the XT PE's bios? That's more than allot of can do m8. Congrats! You are half-way there. I am running XT PE speeds by oc'ing, but I have found no XT PE bios to work. I haven't tried the HIS yet though. I think it is the only one I haven't tried.

[Wilko042]

Yes, my non-VIVO did accept the XT PE BIOS. I Used the HIS one as my card is a HIS.

For reference I used Flashrom 2.37 ( http://www.overclockers.ru/cgi-ATI_flashrom_2.37.rar ) and the HIS BIOS from http://www.techpowerup.com/bios


I was following instructions from
http://www.warp2search.net/modules.php?name=News&file=article&sid=18198
but the BIOS there could not be downloaded. Should I have used a different one? I didn't realise your BIOS flash did not work properly.

Also, in the instructions above I could not use the 4.5 Beta Cats as when using them it said that it could not find any suitable hardware. I ended up using the ones that came with my card CD.

Thanks.

____________________________
P4C800-E Deluxe [email protected]
HIS X800 Pro non-VIVO 575c/561m
2x512MB Ballistix 2.5,2,2,5
2xWD 36G Raptors Raid 0

[johny]

My analysis of this Mod:



If pipeline do have some error during manufacturing, why it is always the bridge between pin1 and pin4 needs to be cut? The production line always have same error in a specific quad pipeline? Not very possible.

Conclusion: All R420 chips are ready for 16 pipeline operation. Disconnect pin1 and pin4 changed certain values in some register. (Someone reported that if pin 1 is connected to pin M, then it becomes 8 pipeline in atitool, I have not tried this yet)

[johny]

I used conductive pen first to connect the bridge, without any result. I doubt the effect of those liquid silver, since I have strong confidence from my previous analysis: If pin1 and pin4 are really connected well, like before the laser cut, it should appear as a real x800 XT core immediately.

So I decided to go the most certain way by soldering it.

I used 0.25mm thin solder, 1mm coper wire bundled on the top of the solder iron to work, applied the solder first to pin1 and pin4:



And then soldered two copper fiber (as thin as a hair, from shielded earphone wires) on each of the pin, and finally soldered them together.



Now I'm 100% sure pin1 and pin4 is connected as all the other X800 XT boards.

Then I bootted up the machine, open Atitool, guess what? Still 12 pipeline.

Conclusion: Besides connecting the bridge, there are more need to be done in order to make Pro work as a real XT.

[johny]

Now, there are 2 things can be sure:

1. Factory always cut the connection between pin1 and pin4 means that this operation has nothing to do with the production quality of the GPU. If some pipeline really appears as broken(Another strange but natural way of thinking: In software architecture, pipeline is totally a different concept from a water pipeline in the kitchen. Details I don't know, but I'm sure if the manufacturing process caused a certain pipeline broken, 99% of the chance is that the other part of the GPU has already broken by the same manufacturing process), it will be either pin1, pin2, pin3 or pin4 disconnected, not like what we have seen today: only pin4 is disconnected. No, personally I will never believe this, since this will cause too much work and verification for the laster cut operation.

2. Hardware mod to connect pin4 back to the other pins is not enough for a successful mod. I can be 100% sure that after soldering all the 4 pin is connected well together.

But is there any other things I might ignored? For example, the laser cut may be did a two layer cut which also cut a layer below the surface layer? I don't know how many layers actually located on the GPU, so I can not say anything about that.

Today, I got this two picture from a chinese site:
x800pro


x800xt


That site already found out that on this specific part PRO VIVO layout is the same as XT, but PRO is not. I think that viperjohn has tried this part without a difference in result.

Anyway, if I eliminate this last difference, from pure hardware point of view, there is no difference between X800 PRO non vivo and XT except the memory chip type.

ASIC ID is another part that is remaining unknown. I don't think this has something to do with the pipeline itself, since many successful PRO VIVO mod without change the ASIC ID can tell that this is not a big issue.

I have a feeling that the final solution will come out in software approach. We will see.

My card:
ASUS AX800PRO
215RAACGA11F
GB8820.1
0423AA
TAIWAN

(I like the pro fan very much, especially the lovely girl printed on that. Even ignored my GPU water block. ASUS XT does not have such a fan, so I just stick to PRO and want to have 16 pipelines. I hate some people destroy a perfect product just because they want a full product line.)

[MaxxxRacer]

Guys, Ive been reading alot about the x800's (as I am going to buy one soon [hopefully]).

Here is the scoop. You can hard mod them some way. A couple of review websites have done it. How they did it i dont know. But hte problem is that they all get artifacts like mad. The vivos dont have the artifact problem and can be softmodded Here is how it works

x800pro non vivo - these cards are ones that have broken pipelines but can reach XT core and clock speeds no problem usuallly
x800pro vivo - these cards are ones that have all pipelines working, but cant reach xt speeds on stock voltage and cooling (there are exceptions to this though, as alot of hte vivos ppl are getting up to xt speeds)

The vivos are basicly just retagged XTs. they share the exact same pcb and everything. there is no difference, except for how fast hte memory and core are.

[-Rogue5-]

I have read that you need to flash the bios after connecting the laser cut bridge.

Can anyone confirm that this works?

Also, not all X800Pro NOn-VIvos make it, but they can be flashed back easily enough (not like the 6800nu).

peace,
-Rogue5-

[MaxxxRacer]

connecting hte bridge and flashing the nonvivos is pointless. There is a 1 in a hundred chance that it will work. And connecting the bridge sometimes doesnt even work to turn on the pipes.

[-Rogue5-]

Originally Posted by MaxxxRacer

connecting hte bridge and flashing the nonvivos is pointless. There is a 1 in a hundred chance that it will work. And connecting the bridge sometimes doesnt even work to turn on the pipes.

These sites make it seem possible.
http://www.madshrimps.com/gotofaqlink.php?linkid=2577

http://www.maxitmag.co.uk/article.php?articleID=79

http://www.ixbt-labs.com/articles2/radeon/rx800-3.html

http://www.warp2search.net/modules.p...icle&sid=18198

Am I missing something, or are these just out of date?

peace,
-Rogue5-

[MaxxxRacer]

Who knows... From what ive read, the nonvivos are really not modable for the most part. you can do it but with a decent chance of Fing up the card... I just wouldnt suggest it.

[ViperJohn]

Originally Posted by johny

Now, there are 2 things can be sure:

1. Factory always cut the connection between pin1 and pin4 means that this operation has nothing to do with the production quality of the GPU. If some pipeline really appears as broken(Another strange but natural way of thinking: In software architecture, pipeline is totally a different concept from a water pipeline in the kitchen. Details I don't know, but I'm sure if the manufacturing process caused a certain pipeline broken, 99% of the chance is that the other part of the GPU has already broken by the same manufacturing process), it will be either pin1, pin2, pin3 or pin4 disconnected, not like what we have seen today: only pin4 is disconnected. No, personally I will never believe this, since this will cause too much work and verification for the laster cut operation.

2. Hardware mod to connect pin4 back to the other pins is not enough for a successful mod. I can be 100% sure that after soldering all the 4 pin is connected well together.

But is there any other things I might ignored? For example, the laser cut may be did a two layer cut which also cut a layer below the surface layer? I don't know how many layers actually located on the GPU, so I can not say anything about that.

Today, I got this two picture from a chinese site:
x800pro


x800xt


That site already found out that on this specific part PRO VIVO layout is the same as XT, but PRO is not. I think that viperjohn has tried this part without a difference in result.

Anyway, if I eliminate this last difference, from pure hardware point of view, there is no difference between X800 PRO non vivo and XT except the memory chip type.

ASIC ID is another part that is remaining unknown. I don't think this has something to do with the pipeline itself, since many successful PRO VIVO mod without change the ASIC ID can tell that this is not a big issue.

I have a feeling that the final solution will come out in software approach. We will see.

My card:
ASUS AX800PRO
215RAACGA11F
GB8820.1
0423AA
TAIWAN

(I like the pro fan very much, especially the lovely girl printed on that. Even ignored my GPU water block. ASUS XT does not have such a fan, so I just stick to PRO and want to have 16 pipelines. I hate some people destroy a perfect product just because they want a full product line.)

There are two resistor "switch" position differences between x800Pro's and x800 XT's or Pro Vivo's.

The resistor at R576/R577 is in the R577 position on Pro's and in the R576 position on XT's and
pro VIVO's.

The other is R360/R359. For Pro's it is in the R359 position and on XT's or Pro VIVO's it is in the
R360 position. Moving the resistors on the Pro cards did not get the last quad enabled.

Viper

[HH_Quazi]

Originally Posted by -Rogue5-

These sites make it seem possible.
http://www.madshrimps.com/gotofaqlink.php?linkid=2577

http://www.maxitmag.co.uk/article.php?articleID=79

http://www.ixbt-labs.com/articles2/radeon/rx800-3.html

http://www.warp2search.net/modules.p...icle&sid=18198

Am I missing something, or are these just out of date?

peace,
-Rogue5-

Yea. They might make it seem possible. And it may be. But I have tried it on two different X800Pro non-VIVO cards. Just like the directions say to do. And it did not work for me. If you get to work, that would be great and please let us know. But, don't hold your breath.

[-Rogue5-]

Originally Posted by HH_Quazi

Yea. They might make it seem possible. And it may be. But I have tried it on two different X800Pro non-VIVO cards. Just like the directions say to do. And it did not work for me. If you get to work, that would be great and please let us know. But, don't hold your breath.

WHat error did you get? Or were you just not able to activate the extra 4 pipelines?

peace,
-Rogue5-

[-Rogue5-]

Originally Posted by ViperJohn

There are two resistor "switch" position differences between x800Pro's and x800 XT's or Pro Vivo's.

The resistor at R576/R577 is in the R577 position on Pro's and in the R576 position on XT's and
pro VIVO's.

The other is R360/R359. For Pro's it is in the R359 position and on XT's or Pro VIVO's it is in the
R360 position. Moving the resistors on the Pro cards did not get the last quad enabled.

Viper

I say this at the risk of sounding stupid;

I assume that you flashed the card to an XT BIOS after doing all the hardmods right? Everything I have read says that you have to do that or you will still only get 12pipes.

peace,
-Rogue5-

[ViperJohn]

Originally Posted by -Rogue5-

I say this at the risk of sounding stupid;

I assume that you flashed the card to an XT BIOS after doing all the hardmods right? Everything I have read says that you have to do that or you will still only get 12pipes.

peace,
-Rogue5-

Of course LOL. If your card has an AA core you will not be able to get 16 pipes right now. There
were a rare few successful last qaud unlocks with x800Pro non VIVO's very early but that is history
now. The unlock issue is still being worked on at the software level but not with a lot of gusto.

Viper