Unlocked Venice 3000+

Seems like the easiest solution would be to edit the bios. If I had experience with modifying bioses, I would give the following a shot:

The FIDVID_STATUS register contains 001010 for bits 21-16 (MaxFID). If it was possible to OR those bits with 100000 or OR all 64 bits with ...00000100000...(making 101010, or 25x for MaxFID), it should be tricked into thinking 25x was the max multi. This would require an available register to write the new value to, and a modification to the address being used for FIDVID_STATUS in all locations (Although only the definition of the register probably has to be changed).

I am familiar with Motorola assembly and MIPS assembly, and this would be very easy to code. However I don't have a clue about the limitations of a BIOS.

Edit: If anyone is willing to point me in the right direction of finding a dissassembler, an uncompression utility, and a spare bios chip for my K8N Neo2, I'll see what I can do...

your FIDVID_STATUS register corresponds to a 10x max mult....wouldnt it be 001001 for the 9x multi on the 3000+, and we can do an bitwise or as you stated. By why 25 if the method even works.

I'm takin computer architecture class and learning about mips right now, so this has got me interested in addition to unlocking the multis

aznblueeboi, it can be set to anything desired, but why just set it to 10x when it would be no more difficult to set it to 25x, freeing all multi's inbetween.

Edit: What I am proposing IS a bitwise OR. Also, you need to look at the table in the AMD documentation (posted a page or two back) to see what the multipliers are for corresponding codes. You can't just convert the binary value into decimal.

I get what you are saying and I knew you were talking about a bitwise OR. I just don't think we need multi up to 25...but if we can do it...why the hell not.

:)

Exactly my thinking. :)

The main problem with what you're suggesting is simple:
The FIDVID_STATUS MSR is read only, and you can't write to it, doesn't matter whether its an OR or anything, its the same as trying to copy a file to a CD or edit an exiting file.
What I'm suggesting is this:

Find a guy who's good with editing BIOSes. Make a modded, specific BIOS for this application and have a separate ROM chip with this BIOS flashed to it.
The BIOS should perform a simple write operation to this register on startup.

Pretty simple, its just a few lines of assembly code, if this MSR becomes somehow, temporarily writable, this mod can be done.
On the other hand, what I suspect is really happening is that the FIDVID_STATUS might not contain these values, but pull them from some other place in the processor when this MSR is being read, so if this is the case, and even if the MSR becomes unprotected for writing at POST, it won't matter much, as the values you'll write to it will be useless as they won't get to the place they need to get.
************************************************** **************

Now, again this is just a guess, but I just think that what happened to this Venice is accidental, and was just probably caused by a glitch in this specific CPU.
I really doubt that motherboards have the ability to use the pins that AMD used to program all the values into the CPU.
Maybe 2 or more traces were short-circuited on that chip, causing the max multi value go from 00_1010 to 11_1010.
I'm not saying its impossible to change this value, because it is probably possible with the right equipment, but I'm just saying that us finding a way to do this is probably not going to happen, at least not anytime soon anyway.

I also still don't understand the reason for people being so obsessed to unlock these chips, you have ram dividers, and motherboards that are capable of running 300+ HTT speeds. I understand its a nice thing to have, but I most certainly wouldn't spend much, if any time and energy on this.

People! :rolleyes:

If you need to test modded BIOS on mobo I can do it.

I have

A8N-SLi Regular (1011-02 Bios) Fully vmodded
Venice E3 LBBLE 0519DPIW or E6 LBBWE 0529
both stable 2680MHz 1.8V
1x256MB BH-6 250MHz 3.47V Stable 1.5-2-2-5-1T

So fell free to ask.

BTW All A64 CPUs have JTAG. may be MaxFID programmed by JTAG?
If so it is possible to make S939 Programmator for unlocking CPUs.
I'm not sure may be K7-cpus also have JTAG for programming.
I don't have a lot of traffic to download PDF for them, so if anybody check it, it will be clear how AMD programm their cpus.

Waiting for results! :eek:

P.S. I'll be happy to programm unlocked CPU with CPUID "AMD Athlon FX-3G Selfmade Edition" ;)

Coolest, there may not be a need to write that register. If there is, then obviously it won't work since it is read only.

What I'm thinking is read the contents of FIDVID_STATUS, OR or XOR it depending on what max multi you want, and store the result in a generic register that is unused and can be written to. Then where ever FIDVID_STATUS is read, just make sure it reading from the address of the new register, and not the address of the FIDVID_STATUS

This should work on the software side, but I'm not sure how AMD designed the hardware side. If the multis are locked by hardware which also reads the FIDVID_STATUS, then again this won't work.

Mrk200

OK, mod the BIOS and I will try.

The Coolest..

I'm not sure for rock stable 24/7 work on my mobo with HTT=350
So i need more mult.
Also I have only 260MHz RAM, and I don't want to use any other divider except 1:1 (DDR400) but it no possibility on my 3000+
And I don't even have money to 3200+. So I need unlocking.

ALL

I have skill to make PCB and "programmator" if you guys get info what is need to do.

TiN, if I knew how, I would. I know nothing about modifying a BIOS, I am only familiar with assembly. :(

Quote:

---

Originally Posted by Mrk200

Coolest, there may not be a need to write that register. If there is, then obviously it won't work since it is read only.

What I'm thinking is read the contents of FIDVID_STATUS, OR or XOR it depending on what max multi you want, and store the result in a generic register that is unused and can be written to. Then where ever FIDVID_STATUS is read, just make sure it reading from the address of the new register, and not the address of the FIDVID_STATUS

This should work on the software side, but I'm not sure how AMD designed the hardware side. If the multis are locked by hardware which also reads the FIDVID_STATUS, then again this won't work.

---

You can't just do that. You're not reading a name, you're reading an address, you can't just try to read/write the wrong register, that won't do you any good.
Plus the CPU checks its internal values (possibly from the STATUS MSR) and then it decides on whether its going to apply that value or not, so basically you need to modify the values that the CPU is comparing your new values to. This means you need to modify the contents of FIDVID_STATUS, or the source this MSR gets the values from.

Coolest,

If there aren't any checks in place, I fail to see how reading from a different address/register will make the slightest bit of difference if the contents are correct.

Again, I don't know anything about the interaction of the BIOS with the CPU, besides reading registers, I only know assembly. So I'm just tossing out ideas that should work from the assembly side of things.

I don't understand you.
This doesn't have to do with the BIOS or any software, they way that the A64 works is that you set a specific value to the FIDVID_CTL, it compares it to the value in FIDVID_STATUS, and after that, if its within the range, applies it.
FIDVIS_CTL's address is C0010041, FIDVID_STATUS' address is C0010042, you can't change those addresses, the CPU will still try to read them and not values from some registers you decided to write to.

So you are saying the comparison is done through the hardware, and not the bios?

Ofcourse.

ALL

I measured resistanse between VSS (GND) and

STRAP_HI_AJ12 ... 86 kOm
STRAP_HI_AF12 ... 86 kOm
STRAP_LO_AJ10 ... 84.7 kOm
STRAP_LO_AH10 ... 84.7 kOm
STRAP_LO_AF10 ... 85 kOm
STRAP_LO_AG9 ... 85 kOm
STRAP_LO_C10 ... 0 Om
STRAP_LO_B13 ... 133 kOm
STRAP_HI_E11 ... 133 kOm
STRAP_LO_F11 ... 0 kOm
STRAP_HI_T3 ... >200 kOm
STRAP_LO_T4 ... >200 kOm

My CPU have MaxFID = x9 (0101)

BTW Socket 754 Package DO NOT have STRAP... pins...but it have NC..
may be they are necessary?

I'll try to measure their resistance with VSS

if you guys pulls this off you are going to be gods. im willing to donate my asus a8n-sli regular, when my dfi nf4-sli-dr mobo comes adn you can keep it

:rolleyes: I'm too far away :) And we need cpu and any 939-mobo to experiment with them. mobo - for powering cpu with power and testing voltages on pins, and cpu - to nake it and completely research it for any external solder places.

ALL

I count NC_... named pins in 754 and 939 processors

939 - 56 pcs
754 - 55 pcs.

After measurements mine CPU still alive.

Here I place some important (as I see) pins on image.

http://www.topmods.net/tin/s939.jpg

I hope this will help.

FX owners,

please measure your resistance between
1. STRAP pins and VSS.
2. NC pins and VSS (I know, it is awesome work, but it is worth doing)

May be tomorrow I will change my E3 Venice to E6 3000+ I will compare resistances.

See you.

Not sure if it's useful, but for people looking into the pins i'd like to post this for easy reference.

External Required Circuits:

Socket 754

NC_AJ23 Tied to VDDIO_SUS through resistor
NC_AH23 Tied to VSS through resistor
NC_A19 Tied to VDDIO_RUN through resistor
NC_C18 Tied to VDDIO_RUN through resistor
NC_D20 Tied to VSS through resistor
NC_C21 Tied to VSS through resistor
NC_D18 Tied to VSS through resistor
NC_B19 Tied to VSS through resistor
NC_C19 Tied to VSS through resistor

Socket 939

STRAP_HI_E11 Tied to VDDIO_SUS through resistor
STRAP_LO_F11 Tied to VSS through resistor
STRAP_HI_AF12 Tied to VDDIO_RUN through resistor
STRAP_HI_AJ12 Tied to VDDIO_RUN through resistor
STRAP_LO_AH6 Tied to VSS through resistor
STRAP_LO_AG9 Tied to VSS through resistor
STRAP_LO_AF10 Tied to VSS through resistor
STRAP_LO_AJ10 Tied to VSS through resistor
STRAP_LO_Ah10 Tied to VSS through resistor

Socket 940

NC_G14 Tied to VDDIO_SUS through resistor
NC_H14 Tied to VSS through resistor
NC_AE14 Tied to VDDIO_RUN through resistor
NC_AF13 Tied to VDDIO_RUN through resistor
NC_AE10 Tied to VSS through resistor
NC_AE11 Tied to VSS through resistor
NC_AF11 Tied to VSS through resistor
NC_AE13 Tied to VSS through resistor
NC_AE12 Tied to VSS through resistor
NC_T3 Tied to VSS through resistor
NC_T4 Tied to VLDT through resistor

soooo you guys got it figured out yet or what :p:

why are you measuring resitsance?.....are you just randomly comparing to see if anything sticks out or what

Not easy to read the resistances with sharp Fluke probe. If I had a small cupped probe it would be very easy to do.

On a dead FX-55 SD I got these for the first 4 values going from top to bottom on TiN's pic - VSS to Strap pins (Yellow)

.908 M Ω
.852 M Ω
.1 M Ω
.1 M Ω

Then for the next two at mid package

4.28 M Ω

I'll go for the rest tomorrow or a little later.

does athlon 64 mobile have unlocked multi?

because I know that they're on s754 but the 754 pins have the same function as 939 ones. the 185 pins of more is just for dual channel...

I saw that to unlock athlon xp it was necessary to mod it to a athlon xp-m.
Maybe it's the same way for a64.
We can see first if it's possible to make a s754 a64 to a a64-m and try with a s939 a64...

but that's just an idea... If athlon 64 mobile have locked multi it's not possible in that way

ha, replied to an earlier post. sorry.