Chosen Solution
It was working fine then the display got glitchy. So I booted into single mode and typed the nvram gpu-power-prefs code and it worked. Now I’m wondering if the GPU died… But then again I remember seeing a Louis’ video about bad soldering on 2012 models but I think it was the Retina one. Does the non-Retinas have bad soldering too and if so does the symptom look like it does in this picture…?
So I warmed it up the board to remove the silicone around the GPU and applied some flux.
Then I covered the surrounding components with tinfoil but honestly don’t know if it helps or not. I did it just in the sense that it can’t do any harm so why not…
Then I used the hot air gun at 300C for 5min timed. And waited till it cooled down to clean up with isopropyl alcohol. A ultrasonic bath would be ideal but I don’t have that so iso it is.
Then I assembled it back and… here it goes:
It worked! :D Even after hours on a stress test still wouldn’t go above 70C I did do a few ‘improvements’ to ensure lower temps though: not only I used high performance thermal paste but I also used some copper shims sandwiched between thermal pads on the back of the board transferring some of the heat to the aluminium bottom plate and essentially turning it into a heatspreader.
Word of caution though: this GPU comes covered with some sort of tape or whatever and it can be very difficult to remove. I knocked off a resistor and had to spend quite some time soldering it back on…
Carlos Ferrari totall y agree with @arbaman on the 2011 is absolutely commonly caused by the GPU chip itself. The processor is a flip chip design and the issue is commonly caused by the solder bumps between the IC and the substrate. The key to this is the design of the processor which is a flip chip design. The proper definition of that can be found at Wikipedia “is a method for interconnecting semiconductor devices, such as IC chips and microelectromechanical systems (MEMS), to external circuitry with solder bumps that have been deposited onto the chip pads.” So what happens is that the chip heats up (due to poor design of the ventilation and heatsinks. It does not matter if it is a MBP a PS3 or a XBox) and the bumps that connect the chip to the substrate lose contact and your chip (in this case processor) fails.
As you can see the “bumps” are what actually connects the die to the substrate to make the chip complete. If these bumps fail, the die does no longer make contact with the substrate and thus no contact with the circuit board. The chip has failed.
Here you can see the space where the bump has failed and no longer makes contact. We are talking microns of space here. Sometimes a bit of pressure on the top of the die potentially close the gap. Same with a reflow, which may allow some of material from the bump to reshape and starting to make contact again. The heating of cooling of the chip during use is what will eventually cause it to fail again. If you are going to do a reflow then the general guidelines for a good profile would be something like this: Temperature ramp up 1ºC/second Peak temperature should be 200C to 210C Remain above liquidus (183C) for 45-75seconds Do not heat any packages above 220C since this will most likely destroy the IC Here is a good document that will help you with the profile as well. BGA-Reflow-Rework.pdf This is not an answer just a quick and dirty explanation of what has commonly affected the flip chip designed GPU
Sometimes it shows different patterns, everything gets pink or green with shadows, other times it doesn’t show anything at all, but unfortunately yes, your Mac picture definitely looks like a faulty Gpu, bad luck. The 2012 release didn’t show such an epidemic like the 2011 models, but still a Gpu may go bad especially if used under heavy loads, videogames, or if the machine has never been cleaned on the inside. The higher the temperature it reaches for longer time spans the higher the risk a Gpu gets fried.
Let me correct @arbaman - answer a bit. Arberman is correct the 2011 models have a bad GPU and/or the VRAM chips. The issue is in fact the soldering which can be either the chips external connections or it can be the joints within the chip (FlipChip). Apple’s push for lead-free solder while noble got them into trouble. As the volume of Tin increases in the solder, the Tin can create whiskers! This is when the tin migrates material to an apposing charge (NASA - What are Tin Whiskers). This can happen within either solder join area. It is believed excessive heat aggravates this process. The 2011 GPU’s had this failure internally as the chip ran hotter and Apple failed to cool the GPU effectively. So using a Lead containing solder removed the Tin whiskers risk as well as solder breakdown (cold solder joints) which is also present. While we can clean off the old solder and apply Leaded solder around the external joints we just can’t get into the chip to fix the Flip-Chip joints. Its suspected NVIDEA was smart enough to add a bit of lead into the solder (or used a better formulation) they used within the chip so the issue is not present within their chips. We are also to blame here as well! We forget the design of a given system was based on the technology it was build with, not for the applications we desire to run on it today, which can push the system to the breaking point. A 1950 Packard Super 8 was a great car in its day, its top speed was about 85 MPH. But, if you pushed it over 60 MPH for long periods you would kill the engine! In the day the roads where still mostly dirt or gravel other than the cities which were mostly cobblestone and tar. Roads that you could maybe travel at best 50 MPH. Within 20 short years the interstates and most major roads where rebuilt into modern roads we have today able to support cars running 85 MPH or faster as the top speed most limited to 65 MPH. So going 60 in the old 1950 Packard would quickly kill the engine as the road was just too much for it. We of course don’t blame the car as it did what it was designed to run at, the roads just got better! So what does this have to do with the MacBook Pro’s? Basically, the same! The applications we run on these systems have just like the roads gotten better but in doing so we are expecting the system to keep up running these apps without failing us. Of course that is the same as the old Packard expecting it to work beyond what it was designed to run at. So running heavy graphics will kill your system over time. Is this fault present in the 2012 model as well? Yes, it is! just not as badly! The newer NVIDIA GeForce GT 650M GPU ran more efficiently from the AMD Radeon HD 6750M or 6770M GPU’s used in the 2011 models so it didn’t struggle as hard so the systems GPU ran cooler.
Thank you ALL for taking the time to assist the rest of us!
