Why?!
Printable View
Why?!
He RMAd it because CnQ didn't work anymore (and it seems that he lost the x64 instruction set as well). Move on, it was already discussed. Lets focus on the unlocking.Quote:
Originally Posted by Mehran
pirs (forum starter) didn't sell it yet
http://www.xtremesystems.org/forums/...light=unlocked
i hope you guy's will find it.
Can't wait to unlock mine :banana:
I'm surpised he hasn't! :stick:Quote:
Originally Posted by KoHaN69
surprized. .... not i 300 anything is too much i dunno how much 300 euro is in american dollars .... but .... yea
so have we determened if it's bios error, c&q error, manufacture error, combination of installing removing process error?
We haven't determined a damn thing yet :(
300 euro is about 366 in us dollars, not really interested in that. Also I really doubt we'll figure out how to unlock. I don't know cr@p about machine language or writing drivers or anything. I would say that a person would need to develop some kind of packet sniffer type of thing and actually wiretap the boot sequence as the comp talks to the bios. Then maybe you could pull some kind of man in the middle attack on chip or the bios. If for instance the bios gets the fid on pin x and then in order to set the multi it must send that data to pin y to set the multi you've got a chance. You must lie to pin y and you've got whatever multi you want. I have no idea if that makes any sense or not though, it's probably completely retarded. It may even have some error handling such as if no data comes from pin x then operate unlocked, so you could just hack off pin x. I've never even messed with wpcredit too much let alone know how to talk to the chip.
he lost the 64 bit instructions, and Cnq? who uses Cnq anyway?
People that rma unlocked venices...
The thread has already been cleaned from posts like these. Please do *not* make it be locked down.
Thank you.
Clocker, don't RMA your chip sell it here in the community.
He already did it. Move on. Please.Quote:
Originally Posted by serker
Back on the topic...
Ok, the cpu has internal "registers" which are in fact addressable static memory.
When you want to alter the multiplier, you tell the CPU "please change cpu-multiplier to X"
"the CPU takes X and compares it to its MAX_MULTIPLIER"
(MAX_MULTIPLIER) is a value stored in a register (static memory) in the cpu.
"if(X <= MAX_MULTIPLIER)
setRegisterValue(CURRENT_MULTIPLIER,X);"
Now.
There are 2 way in succeeding with this.
1. JTAG-interface, change value of MAX_MULTIPLIER
2. JTAG-interface, bypass the call to "if(X <= MAX_MULTIPLIER)"
Clockers CPU most probably got a power-surge which definately scewed up his CPU by scrabling some of the registers. The CPU most probably still is able to process x86-64-instruction, but the register saying that it can, has been messed up.
The CPU also is CnQ still, but same as above, the place that says it has CnQ, has been messed up.
Ok?
There is however a chance that AMD didn't burn the JTAG-bridge on all non-ES CPUs, this would make sense because if a customer like Clocker RMA:s a messed up CPU, AMD could easily connect to it via JTAG and see what was screwed up, flash the registers back to what they should be, and then sell it again.
But, then we also have the risk that AMD definately sends a destruct-signal to the JTAG-interface on the CPU which disables JTAG, prior to shipping the CPU.
So, in any of the cases, we need the JTAG-experiment to work, and for that to work we need the spec of the A64 939 JTAG interface.
Very interesting info there....thanks. I'm really not sure how we would procede from here though.Quote:
Originally Posted by Warship
BTW....about selling them back, I'm pretty sure there are laws/regulations against that ;) They could sell them to OEMs though, just not back through Retail.
They could either flash the CPU back to its original state, then ship it back to him, or give him a replacement and ship the rma'ed one to some third world country as a tray model.Quote:
Originally Posted by Vapor
But since AMD is phasing out their bulk/tray-sales, going for retail-only (so cusomers get a CoA etc), that wouldn't be likely.
In any case, the _only_ hope is JTAG. Because there is a specific register we need to tamper with, the one which stores the max-multi.
Yes, but HOW do we tamper with JTAG--any ideas?
On a PDF about AMD's Geode processor it mentions the JTAG Interface.
Maybe the K8's is similar?Quote:
JTAG interface and trace port compatible with First Silicon
Solution development tools: www.fs2.com/isa-geode.html
Tin-EOF has some plans from what I've read in this entire thread.
The hardware is not much of a problem, we know what pins go where.
The problem is the spec.
We can't talk to it unless we know what to say, therefore we need the spec.
The A64 939-JTAG spec.
How would we talk to it, exactly? Couldn't we listen to it in theory then, as well?
What we would do is that we would tell it "Set register MAX_MULTI to 010101(*or whatever value that x25 corresponds to)"
Tada. Fixed.
But to be able to do that we need to know what we have to do before we can tell it to change the values.
Perhaps it's only possible to flash _all_ registers at one time, rather than changing specific registries. If that's the case then we need the total spec of the registers, which values they need to have, so the CPU doesn't get fubared.
To simplify things a bit, this might be a colorful illustration of how it _may_ work:
A = we through JTAG
B = CPU through JTAG
A: Hi there, are you alive?
B: Yes.
A: I want to set a new value to register MAX_MULTI, and that value is X.
B: No can do, that register is read-only.
A: Ok, how about I want to flash your registers, since there are some erronous values in them.
B: I guess that's ok, go ahead.
A: flash-registers, set data 0101010100101010100101010101010...
The thing is, for us to flash the entire CPU, we need to know what goes where, in terms of register values.
And for that we need the spec.
Simple as that.
(This is pure speculation, but seems rational enough compaired to the BIOS-ideas and so on)
Well if the Jtag could be read on an FX could they be copied? Then flash them to a 3200 or whatever..
That's not entirely true.Quote:
Originally Posted by njkid32
We'd use the venice 3000+.
We'd probably need to read one register value at a time, then put them in one large update-chunk, change MAX_MULTI from 9 to 25, then overwrite _all_ registers on the cpu at the same time.
If we'd copy info from an FX to the Venice, we'd have to change the entire CPUID-info.
It's easier to use the same model, extract all data, modify MAX_MULTI, and then shovel it back into the cpu.
But for us to get a copy of the spec from AMD, we'd need to sell some souls to Satan. :fact:
Ahh got it! Sorry I know nothing when it comes to this stuff. I just thought it might work. Well I hope Tin_Eof figures it out! I have already donated my dead DFI mobo to this so I hope someone figures it out.Quote:
Originally Posted by Warship
Really?! How bad! That's no problem. Someone should find a way to unlock these :banana::banana::banana::banana:s!!! :DQuote:
Originally Posted by Daved+