ive been doing some reballing lately ,and i faced many 0003 error boards after reballing , i found what exactly was the problem , after a few tests , i found that when removing or soldering the gpu ,you have to be very careful with the heat temperature , too much heat will fry the gpu ,you might do a successful reballing and soldering , and the gpu fits perfectly , but you still get an error , the reason for that is that you applied too much heat on the gpu so it got bad and its not working any more , if you did a gpu and it got bad youll get 0003 , and the bad cpu will give 0002 , i read a post by some guy in the xboxhacker.net forum about the resistance between the ceramic condensers that fits on the power source of the gpu current like in
, it has to be around 2 ohms , if you got a 0 ohm your gpu is short and will give a buzz sound and a 0003 error when board is powered on , bridged balls under the gpu in the power area that is the center of the gpu will give the same error but im talking about the error that appears after a successful soldering or reflow ,i tried soldering the same gpu on another board and it gave me the same result , i tested the resistance and found it 0 , i removed the gpu ,and turned on the board , it powered but gave a 0020 as expected because there is no gpu , to be sure that your gpu is not bad before soldering it , test it with an avometer on those points , around 2 ohms is the right mesure , if 0 then its dead ,and no use soldering it ,
i face another error 0020 , this one is for the power of the small memory thing on the gpu , it gets its power from this trace pic0020.jpg, if you had this error in a virgin board then its a bad solder ball , try reflawing but if you got it after a successful reball then the memory thing is dead , and you need to replace the gpu with a working one.
i lost 5 boards to get to this point of knowledge , so i dont want any other people to suffer as me ,
good luck every body

Nice post, but this probably isn't news to anyone. You can get 0003 easily by heat gunning or reflowing at too high of a temp/too close to the GPU as well. Popcorning on the GPU will usually give 0003, and so sill a bridged solder joint underneath it. 0002 isn't always the CPU either, it can just be a short somewhere, or it can be the ethernet chip as well.

One thing though, if you get popcorning - the GPU might not be as dead as you think it is, you can actually take a Q-tip and cut off one of the buds (or use something else that's firm, but not jagged or sharp), and basically rub the crap out of the affected areas, and the top layer of the wafer will sometimes re-laminate itself, and there's a pretty good chance the system will work again (assuming any other issues are fixed as well). I've done this myself a couple times. Under normal circumstances the GPU won't get hot enough to re-popcorn afterwards either, and it won't work on every chip, but you might be amazed that it can actually work.

brandog , i didnt get all what you said about fixing the dead gpu's , can you explain a little more ,like what is exactly the pop-corning and what is it that you cut , if you have pictures please , im interested

Popcorning's when you've applied so much heat for too much time - taking the GPU above it's thermal stress level (can also be caused by ramping the heat up too quickly in your thermal profile - 1-4 degrees C max per second is it's limit).
A popcorned chip has bubbles over its surface - Brandogg suggests getting the plastic end of a cotton bud/q-tip and using it to press down the bubbles (or re-laminate). Never tried it myself, but it may work.

Also thanks for the info on testing the GPU's - I've got a few here that I'm going to test (I replaced them with new GPU's), although I agree with Brandogg that 0003, 0002, 0001 can also be caused by bad mosfets, or shorting on the board (basically a power issue).

I only have one popcorned GPU-hand (from getting my clean lift profile), but it's not really noticeable when I take a picture, and I can't find any pics through Google image search...but thwack pretty much described it perfectly. I actually just changed my reflowing profile since I had a couple slightly-poopcorned GPUs recently, now I just to 330C max on my 968. Got a clean lift and a perfect reflow last night, looks like a brand new motherboard.

When I used to use my 968 (and they're a great machine ) I'd have to use a profile that ran up to 450C on the machine. After drilling out a mobo to the BGA under the GPU, I found that @450 would really equal 230C, so my profile was (drumroll for boring bit) this is on a preheater that gives 140C to the base of the board, 130 to the top:

Temp On Aoyue Actual temp
90C - 30 Seconds - 115C
200C - 1 minute- 133C
275C - 1 minute - 157C
350C - 30 seconds - 179C
400C - 20 seconds - 196C
425C - 20 seconds - 217C
450C - 20 seconds - 224C - raising to 230C around 10 seconds

Thats 1 click every second up (so 3C per second) and temps left for that time after temp reached. Then back down to 200C on the Aoyue at 2 clicks per second, then straight down to 90C. Turn off preheater and leave for 20 minutes. Basically it sits in the reflow zone for no more than 60 seconds. These temps vary from rework station to rework station, only true way to work out your temps is by drilling to the BGA.



Poopcorn

Well i go mainly by my clean-lift profile, but I use thermocouples placed adjacent to the GPU get get a good temperature. Distance from the nozzle to the GPU makes a huge difference as well, I keep mine about 3-4mm above the wafer (from wafer to bottom of nozzle). My profile is like this (always going 3C per sec between holds, and these are the air temps from the AOYUE, no the thermocouple readings):
90C - 1 min
180C - 1 min
270C - 30 sec
315C - 30 sec
330C - 15-20 sec
315C - 10 sec
300C - 10 sec
then all the way back down to 90C, 1 click per second. Preheater is set to 400F, keeps the top of the board at about 160C, which from what I've read is pretty ideal, I turn it off when I get back down to around 240C on the AOYUE. Actual temps will vary from station to station like thwack says, but I've had a ton of success with the setup, about to do another one right now.

thanks for the info guys , it helped a lot ,
i just wanted to know , what flux do you use , i mean the soldering paste , i heard some people doesnt use any , is that possible without burning the board
there is one piece of info that i heard that might explain why the popcorning is a very bad thing on the gpu while it might not be that bad on the mainboard if happened ,
how is the silicon die mounted on the small fiber board of the gpu part thing , its soldered like the bga , balls , very small ones , so if popcorning happens , the balls will bridge somehow , because the surface isnt even under it ,
i wish i can reball the silicon die and solder it back on the small board ,this will help fix lots of bad gpu , but i guess this is impossible ,
did one of you say "replaced with new ones" , where do you get new GPU's , is there a source you could tell me about
thanks every body

great test for gpu

d I can't find any pics through Google image search...but thwack pretty much described it perfectly. I actually just changed my reflowing profile 。。。。。。。。。

The weird thing is, some people have reattached GPUs which measure 0.4 ohms and still got green lights.

as long as its not zero , then it still might work , zero means short circuit and dead for sure.

Good job on reading the date that was posted. LMAO. Dumbass...

cant you find any other joy in life other than insulting other people , you are sick , i just happen to pass by an old post of my own work , and i decided to leave a note so that other people would make use of the experience , other forums are making use of this post a source for an experience and information , and they do say thanks , polite people , not like you codec192

I had a Falcon last week with 0003, good reading on the GPU too. Reballed it and still 0003. Fitted a new GPU, green

there might be a problem with the small capacitors on the top of the gpu board , some times the gpu might not be short but wont work because the silicon is broken

Didn't look further in to it, just removed the GPU, binned it and fitted a new one

0003 is a problem with bridging of the gpu or ram or even faulty gpu

we did a test on about 9 0003 boards, removed about 60 balls from the gpu and placed it on and it booted fine
the rest all booted as well, trying new gpus on all of them

as i said earlier , if its not bridged balls under the gpu , it will be for sure a dead gpu , this is noticed when you solder the gpu to the board perfectly , but still get the 0003 error,
thanks for the post though