Q: 2011 MacBook Pro and Discrete Graphics Card

Yosemite puts a lot more demands on the GPU than Mavericks with the transparency option in particular, similar to Windows with the 'aero' mode but it uses more power with this translucent effect which needs rendering.  In fact with every release since snow leopard (which was by far the most lean and mean and fastest OS X on Intel) the amount of extra services have been increased with every new version but 10.10 has probably put more extra burden on the GPU than the earlier releases in terms of increments.  On earlier Mac's like 2007/9 iMacs for example I have turned some of these settings off to stop the UI lagging upgrading from Mavericks and I have quite a few more in the field still to do and they will stay on 10.9 until I do have time as work has gone crazy of late!

system preferences - accessibility reduce transparency

The 2012 though almost an identical chassis suffers far less from this problem and breakdowns are usually down to particulates blocking the fans. In fact the 3 broken 2012's with broken gpu I have opened up all had stuff blocking them.  I have had 4-5 times more 2011's with this problem and friends at apple service providers have also confirmed this to me.

Why - because unlike the 2011 the 2012 has a 35 watt cpu and 26 watt gpu. With 45 watts for the cpu and 40 watts for the gpu in the 2011 there is a lot more heat needed to be dissipated out of that chassis and thus the failure rates are much worse.  Then came the hugely improved cooling systems for the retinas which not only have pretty shiny copper surfaces to bind to the chips but much larger assemblies too to extract that thermal energy away from the chips.

Additional services are standard for new operating systems, if none were needed upgrading would not exist. The question is why does this not affect all Macs running Yosemite, none of mine run hotter than before, none run slower. is this magic?

Out of interest, which models do you have? And are they "all they can be" or just entry level?

My equipment list is included in my profile

Hi D3us,

Thank you again for the amazing job you did! My 15" Macbook Pro Late

2011 is working like new again!!!

The issue of the entire situation became so clear to me from your explanation!

And the problem is so logical actually after I heard your ideas....

All because of one poor choice of Apple in the production process.

Here's my attempt to try to help all other curious people understand what's going on with those 2011 machines (and who knows which other models).

The reason why so many of these 2011 macbook pro machines fail/die on us:

The logic boards, which are created in an amazing (to me, at least) way, contain very small to relatively large (BGA http://en.wikipedia.org/wiki/Ball_grid_array) components.

These components (incl the little tin balls) are all positioned perfectly on the contacts of the BGA/PCB (http://en.wikipedia.org/wiki/Printed_circuit_board) and then that whole thing goes through some kind of reflow oven. Think of the kind of conveyor belt oven Pizza Hut uses.

Goal: Heat the entire PCB incl all components to approx 240ºC to get all components soldered to the PCB, but to stay within the safety limits of ALL components and standardised solder timings/temps.

That "approx" is where the problem resides.

All the components on the PCB are different and have different max temperature (+duration @ that temp) specifications.

Different sized component can influence the heat penetration in the reflow oven, which can result in effective temperature differences at the solder balls. And remember: Some of these components are BGA’s, like the failing GPUs = These have small solder balls between the chip and pcb.

So Apple + the suppliers of the components had to "choose" a common "safe" max temperature that would serve the purpose of a) creating a good connection and b) not damaging any of the components..

And here lies some irony:

By trying not to damage some component(s), they kind of completely destroyed the reliability of this range of machines...

Let's say they chose to move the PCBs through that oven at max 240ºC (the melting temperature of tin is 232ºC) for “some” short period of time.

The AMD BGA in our macbooks has probably about 700-800 solder balls.

Now imagine those tiny balls heating up, but not long enough...

They are between the chip and pcb, so might need more time to get fully heated.

Some balls might turn completely into liquid phase, but some might not entirely turn into liquid phase and just go soft only!

Then you get the situation where probably most balls melted completely, forming a full soldering connection, while some balls might just have gotten soft and gotten "pressed flat" between the GPU and the PCB, making an electrical "pressure" contact, instead of a real soldering contact.

So you might say that Apple may have been operating too much on the “edge”, concerning choice of heating temperature in their ovens..

And there you go, problem created, but well hidden!

Immediate tests after production always showed perfect functionality, because the "pressure connections" made a fine electrical connection at that moment.

But then 2 main factors start to change that picture:

- TIME

- TEMPERATURE

As time passes and as you heat up your GPU nice and hot (basically by normal use of your nice, expensive Mac, even just by watching a movie), on each usage cycle, the GPU expands+shrinks a tiny bit.

Imagine this happening for a few years and also imagine how oxidation just loves to crawl in between those pressure joints.

Especially as these unibody models are a huge chunk of aluminium that cause plenty of condensation inside when moving the device from a cold car into a warm room, etc.

So there you go.

A wrong choice for the oven temperature and "baking" duration creates imperfect solder joints and time + specific high GPU temperatures ruines those bad joints further and our GPUs fail.

Why mainly the GPU and not other components?

Usually, the CPU doesn't operate on 100% of its capacity during its (normal) operational life, especially with those fast quad core CPU's nowadays.

Insufficient memory and other slower hardware components form a bottle neck anyway (usually) which prevent the CPU from making lots and lots of 100% occupation hours.

The GPU, on the other hand, is a component that gets it's rectangular butt kicked quite often, quite severely. The GPU is utilised in higher percentages much more then the CPU apparently.

Causing that 30W GPU to undergo many more "severe" temperature changes then the 45W CPU will experience.

Still, CPUs are also reported to have failed on these boards (and restored successfully after a "simple" reflow procedure).

So what was done to my MacBook was a REFLOW, not a reballing.

A reflow is basically repeating the process that Apple was supposed to be doing correctly in their oven in the first place, but then at the RIGHT temperature where all solder balls actually DO flow 100% completely.

The endresult?

What Apple would have supplied if they had risen the temperature just a little bit higher, or maybe for a little bit longer period of time.

I have paid a visit to D3us last saturday. He reflowed my GPU.

So no new chip, no reballing, just making the original solder balls melt good (for the first time ever).

And the result?

My Mac works like new!! And personally I feel that this whole story is super logical and the solution is just very practical.

And, UNLESS Apple changed the protocols for soldering drastically after the 2011 model, all following models might be up for similar problems after several years of service...? If they soldered everything according to the same protocols from 2011 on, I'm not surprised that Apple isn't responding to this very quickly. Maybe they are researching their behinds off, to get a picture of what's really happening here. Which might have a scary conclusion..

Since my Mac was out of warranty already, I didn't really care whether Apple would pay back the €125 I spent or not.

This was such a small price for bringing my Mac back better then ever before. And as I use my Mac to make a living, I really couldn't be bothered by spending €125 to get this thing working as it should.

My advice; think about the logic in this story, and consider this procedure, as it works in almost all cases. Of course, some component could be literally "broken", besides that the soldering is bad. Then it would mean replacing that too.

If it'd be impossible to find which component is broken, it's always possible to find a refurbished 2nd hand logic board and reflow that too.

Costs are still always lower then buying a nice new, replacement board from Apple Service which has the same problems built in!

The story above here also explains why the replacement boards keep showing the same defects, as they are all produced in the same way, with imperfect soldering connections. It also explains why nobody can find a successful "software fix". It also explains the variety in related symptoms, as it's of course unknown which of the 700-800 tin balls didn't melt/flow good!

Most of all: Applying this fix turns our dead Macs back into the reliable machines they're supposed to be. The hardware components are basically all just fine. The choice of combination of components is also just fine. Macs are in fact just fine concerning hardware.

Great even.

But if you do a bad job soldering good parts together, well... then this weird behaviour is really not a surprise, right?

Fixing the boards in this way, which Apple will NEVER be able to do in an affordable way (for them, on such a large scale), seems the most practical solution to me.

What Apple would officially have to go through, to fix this RIGHT, is this (and then imagine if you see that happening):

- create new logic boards

- produce them with different production standards, deviating from what they primarily decided was "safe for all components", heating them up to a higher temperature

- send every 2011 Mac owner (and any other who reports similar weird

problems) a nice new Logicboard...

Again: Do you see that happening?

Excuse my mistakes in english, it's not my native language.

I just hope that this story will help people to get an idea what's happening with their machine and what might be the most cost effective, durable solution..

D3us: Thumbs up and thanks again!!

Of course with every release they add new features on top, it's rare releases such as snow leopard and windows 7 where they take the existing os kernel from the previous version (leopard and Vista) and truly optimise it to make it much faster than the previous release.

Most users do not run fan control programs; the stock EFI does not ramp the fan controls up till almost 85c so most users do not notice. I do run these apps on most of my clients older Mac systems and an app like that running in conjunction with activity monitor you can see the system is being worked harder post 10.10 upgrade, though you should let spotlight to its work first as that hogs the system up heavily after a fresh Yosemite install.

Richard Feynman has some classic quotes about magic and they are worth googling

And like with every OS point upgrade you should boot off the recovery partition and repair the disk permissions on the primary Macintosh hd volume.  Command + r on power up, you will not see the recovery partition via option key anyway like Mavericks and earlier.  That fixes quite a few slow downs!

Yes, but it doesn't tell me what CPU and graphics you're running on what machine.

It would be interesting to know.

The model and year of every machine is there. They are all 13" laptops or 21 inch iMacs.

GavMackem wrote:

 

That fixes quite a few slow downs!

I don't get slowdowns, and my Macs are very standard, just loaded with fast drives and lots of Ram.

Then you'll probably find some speed ups after doing it instead, you will see the huge numbers of permissions that get repaired whilst the disk is unmounted.  Printers, Safari and a lot of the core services and tons in apps like iBooks etc.

My Macs are all pretty fast too, though I delegate all heavy gpu work to the Mac Pro because it can handle hard gpu work with aplomb. I now have a flashed 4,1 to 5,1 6 core 3.33 xeon with 24 gb of ram, but even in my old 8 core 3,1 with 32gb ram which had Yosemite beta running with all my upgrade parts I noticed that safari and the UI lagged less after I did the recovery permissions repair and that old 2008 was much faster than my 2011 MBP at almost everything.

Very happy with my 4,1 now, feels 25% quicker than my 3,1.but it looks like i may need to learn solid works so I think a dual processor tray, 12 cores and 48gb of ram is coming with 30,000 plus on geekbench.  Unlike my 2011 or any other Mac it's the true professional Macintosh for hard work!

Ah, so no 15-17" notebooks...

That is interesting.

If I may ask, did you configure them, or just buy the stock machines?

Configured them all (some from a master image) some upgrades, no machines actually came with Yosemite, most have been through several upgrades.

That will be an absolute beast of a machine!

GavMackem wrote:

 

Then you'll probably find some speed ups after doing it instead, you will see the huge numbers of permissions that get repaired whilst the disk is unmounted.  Printers, Safari and a lot of the core services and tons in apps like iBooks etc.

 

My Macs are all pretty fast too, though I delegate all heavy gpu work to the Mac Pro because it can handle hard gpu work with aplomb. I now have a flashed 4,1 to 5,1 6 core 3.33 xeon with 24 gb of ram, but even in my old 8 core 3,1 with 32gb ram which had Yosemite beta running with all my upgrade parts I noticed that safari and the UI lagged less after I did the recovery permissions repair and that old 2008 was much faster than my 2011 MBP at almost everything.

 

Very happy with my 4,1 now, feels 25% quicker than my 3,1.but it looks like i may need to learn solid works so I think a dual processor tray, 12 cores and 48gb of ram is coming with 30,000 plus on geekbench.  Unlike my 2011 or any other Mac it's the true professional Macintosh for hard work!

Other than the machines that I use these are all work machines, they have few services running (most never connect to the internet in normal use), they run proprietary software for device control based on external inputs. no mail services, network file storage and no games

My personal machines (the older ones) are more normal but no less reliable.