820-3662 possible fix? Simple script

Had a 820-3662 come in for random shut down. No liquid. Checked all the rtc stuff and caps normally checked. Seems fine.
Found a script online claiming the shut downs are caused by a lower power state bug in the CPU that causes failure when the CPU enters a lower power state.
Before running the script I was getting shut downs ever few mins. Been going for a few hours now with no issue.
Is this legit? Or am I getting a placebo effect.
 

JohnB8812

New member
Hmmm please let us know! This sounds interesting if it helps. I still wonder if replacing the FDM6808s with FDM6708s is a know fix yet? Any luck dukefawks ?
 
I think the CPU Power Controllers could be a possibility but I would think they would fail under high load not low load. I have done those before on boards with obvious shorts on output.
 
realmacmods didnt know you were on here. Yes seems to be working great. If you dont mind spilling a few beans what seems to be the root cause? Is there proof in older firmware that they dont drop voltage that low?

Is this literally a flaw in Intel Power States where the CPU is put at such a low voltage it crashes because of not having enough power to sustain the processes required? Like when you undervolt a PC intentionally to save battery or reduce heat? Or in the opposite spectrum not enough power when over clocking a chip and causing a crash in the same respects.
 
Last edited:

realmacmods

Member
Yeah just recently got on here. long time lurker. Yeah its basically like undervolting instability. It actually panics (im almost certain), but there is no time to log it anywhere before the thing dies. Im quite certain there was an efi update in 2016/2017 that helped battery life by a few percent by dipping into that low power state. Watch it in Intel Power Gadget without the tool running and you will see what i mean. If you can find one on original EFI it wont ever go that low. My speculation is that the CPU was not designed to handle that power state, and a decent portion of the 3662 and 3787 are affected. Not enough to warrant a recall or an efi rollback of course. That only took 3 year and every 2011 ever made ;) . This is in use on a few thousand MBPs right now with nothing but positive results. Patient zero is a local customer here about a year and half ago (if memory serves) and he is still up and running just fine.


Also, watch power gadget in windows, and voltages from a terminal in linux. Never touches it, never crashes. Absolutely OSX telling it to go low.
 
Last edited:

dukefawks

Administrator
Don't do this. The problem is the DrMOS FET on phase 1 of the CPU power supply. If the CPU has no load only phase 1 out of 3 will be active and the failing FET will make it shut down. Having the CPU not enter idle anymore will just mask this problem until the FET deteriorates more or the FETs on the other 2 phases will start to get the same issue.

Replace all 3 DrMOS with the 6708. 6708 is available new from mouser or China and you won't get sanded pulled garbage as with all the 6808 trash from China.
 

dukefawks

Administrator
No offense meant, it is a working solution for the people that cannot or do not want to fix their machine or spend money.

I mean don't do this if you have the means to do the proper repair. Do not sell this "hack" as a real fix to a customer unless they are aware of their options and choose to go for the hack instead of the real repair. It is kind of similar to reflowing/reballing a GPU when you should replace the chip with a new one.
 

realmacmods

Member
No offense taken, i have a great deal of respect for you and your work. All i can offer is that of thousands, i have had no reports the issue returning or worsening. Not saying that the FET replacement is not a possible fix, but i am being honest with my results with a pretty large sample size over a decent time span. Not sure its quite on the same level as a reflow ha
 
Don't do this. The problem is the DrMOS FET on phase 1 of the CPU power supply. If the CPU has no load only phase 1 out of 3 will be active and the failing FET will make it shut down. Having the CPU not enter idle anymore will just mask this problem until the FET deteriorates more or the FETs on the other 2 phases will start to get the same issue.

Replace all 3 DrMOS with the 6708. 6708 is available new from mouser or China and you won't get sanded pulled garbage as with all the 6808 trash from China.

Ok thanks for the info Duke. Ill be buying some now. I think I had some of these that I replaced on 2 other water damaged boards but currently out of stock. Ill see if the customer wants to wait or just roll with the software.
Thank you both
 

dukefawks

Administrator
I saw you replace these in a video and you were using 6808. No idea where you got them, but you can assume with 95% certainty that they were pulled crap from China. They sand them up real nice and they look almost new until you compare it to a real new one under the microscope. Getting the 6708 from Mouser is the safe option here, but I have gotten 6708 from China that were real new too.
 
Last edited:
Top