Performance reduced after EFI update on MacBook Pro Retina

[Junkyman]

Of course it's only a temporary fix. Think about logically what's happening here. The new EFI update has more aggressive throttling parameters to reduce heat build up. The MBP CPU can easily hit 99 degrees Celsius. Even though this is just below T junction (max temp) it is still too high. The temps for these mobile quad CPUs should be kept below 90 degrees Celsius for throttling to not occur. Unfortunately the cooling system is not adequate enough to do this. The same problem is on the GPU side.

The SMC controls the cooling system and also sets parameters when throttling should occur, based on the temperatures that it sees in the system. When you reset it, those values are reset so SMC jumps in and checks out what's going on - so you see less throttling. Then it registers the hot temperatures and applies more aggressive throttling. So you can keep resetting the SMC, but you'll be back to the same point.

People that say they have no issue are either not noticing it or not pushing the thermal limits. There is no magical EFI that some people have and some people don't. If you've updated to the latest firmware, the parameters will be the same, until the day we have fully self-aware AI-like software.

Thanks for the elaborate answer. Appreciate it.

However, the thing is that (at least for me) the issue only surfaces if I use the sleep function. I have been shutting down my Mac for an entire week now and effectively not once used the sleep function of my Mac. By doing so, I have effectively eliminated the issue. That's why I'm certain that the issue only comes up when I use sleep.

I also push my Mac to higher temperatures on a regular basis (videos, gaming etc.) on both Mac OS and Windows (BootCamp). There is no "active throttling", that is that the performance drops during gaming. In fact, it runs very smoothly and without any problems. But again, once I use sleep - the issue is back and I have to do a SMC-reset.

So what exactly is causing the problem? I really like to use sleep again. Shutting down the Mac constantly is very annoying.

 

[theSeb]

Thanks for the elaborate answer. Appreciate it.

However, the thing is that (at least for me) the issue only surfaces if I use the sleep function. I have been shutting down my Mac for an entire week now and effectively not once used the sleep function of my Mac. By doing so, I have effectively eliminated the issue. That's why I'm certain that the issue only comes up when I use sleep.

I also push my Mac to higher temperatures on a regular basis (videos, gaming etc.) on both Mac OS and Windows (BootCamp). There is no "active throttling", that is that the performance drops during gaming. In fact, it runs very smoothly and without any problems. But again, once I use sleep - the issue is back and I have to do a SMC-reset.

So what exactly is causing the problem? I really like to use sleep again. Shutting down the Mac constantly is very annoying.

I am not 100% sure. I can only make educated guesses based on my knowledge of hardware and software that I've dealt with. Unfortunately I am not an Apple engineer and have not been involved in the EFI development. I can only draw on experience of what I've done. It could be that SMC is not writing the values in correctly when you set the MBP to sleep or that something gets slightly corrupt when it goes to sleep and not read back properly so you have a permanent throttling effect, regardless of the temperatures. It's difficult to say without seeing the code.

I can however tell you that there are serious issues with the latest EFI for the rMBP and many of them are related to sleep. I ordered my rMBP about a week after it was launched and it's been solid and reliable. A bit slower and noisier than I had expected or hoped, but certainly reliable. Ever since the EFI and 10.8.2 update my rMBP kernel panics when going to sleep / waking up. It does not happen all of the time - only occasionally. This and the issues you guys describe may be related, or they may not be, but I am hoping that there is a fix for this soon.

 

[Junkyman]

I am not 100% sure. I can only make educated guesses based on my knowledge of hardware and software that I've dealt with. Unfortunately I am not an Apple engineer and have not been involved in the EFI development. I can only draw on experience of what I've done. It could be that SMC is not writing the values in correctly when you set the MBP to sleep or that something gets slightly corrupt when it goes to sleep and not read back properly so you have a permanent throttling effect, regardless of the temperatures. It's difficult to say without seeing the code.

I can however tell you that there are serious issues with the latest EFI for the rMBP and many of them are related to sleep. I ordered my rMBP about a week after it was launched and it's been solid and reliable. A bit slower and noisier than I had expected or hoped, but certainly reliable. Ever since the EFI and 10.8.2 update my rMBP kernel panics when going to sleep / waking up. It does not happen all of the time - only occasionally. This and the issues you guys describe may be related, or they may not be, but I am hoping that there is a fix for this soon.

Thanks again for your answer. Let's hope that Apple is aware of this issue and taking care of it.

 

[0x000000]

Hmmm, in pre ivy/sandybridge macs throttling was osx-stuff. If anyone remembers, there was this nice app called CoolBook which used Intel's speedstep stuff to undervolt your cpu, reducing heat and improving battery time. That was all done within osx. Not sure what the EFI update did or whether throttling is now done by EFI now or not, but the problem sounds really weird.

If the EFI has settings about when to throttle, it would be a list where the computer looks up when to do what. Those values should stay the same and not change after sleeping...

Well, I hope apple gets that solved.

 

[theSeb]

Hmmm, in pre ivy/sandybridge macs throttling was osx-stuff. If anyone remembers, there was this nice app called CoolBook which used Intel's speedstep stuff to undervolt your cpu, reducing heat and improving battery time. That was all done within osx. Not sure what the EFI update did or whether throttling is now done by EFI now or not, but the problem sounds really weird.

If the EFI has settings about when to throttle, it would be a list where the computer looks up when to do what. Those values should stay the same and not change after sleeping...

Well, I hope apple gets that solved.

The SMC does all of this. Coolbook modified parameters that SMC uses. Coolbook can not be used since Lion when Apple changed how things work and getting that info is a lot harder. It was either the 10.8.2 update or the EFI update that changed the throttling parameters. I am not 100% clear on where SMC gets its parameters from these days, but I have to assume it is the EFI seeing as each model will have its own parameters.

 

[pandaen]

I will be following this thread.

Anyone have a justified geuss as to if this problem will ever be fixed and how?

Do you think Apple will fix the EFI firmware for it to be like before?
Do you think that other updates will fix the actual problem?
If Apple does not fix the problem(ever?), do you think it will be possible to fix it with 3rd party software/mods/tweaks?

I am willing to wait for a fix, but not forever :/

Resetting SMC does not work for me. If I stress my rMBP it will start throttling with reset SMC and not, and it is the same whether or not it has been to sleep. Note that I have disabled deep sleep AND powernap, so it never REALLY sleeps

I can run Starcraft 2 on decent settings without FPS drops and such though and that is my primary game, but I think it is pretty dissapointing that the performance drops so bad when stressed!

 

[bill-p]

...It could be that SMC is not writing the values in correctly when you set the MBP to sleep or that something gets slightly corrupt when it goes to sleep and not read back properly so you have a permanent throttling effect, regardless of the temperatures. It's difficult to say without seeing the code.

And I think this might actually be the case.

I re-enabled Power Nap (code word for "allow the computer to not immediately go to sleep while it still has things to do") and have not gotten EFI bug back for a whole week.

Prior to this, I'd get the EFI bug probably once a day at least since I use clamshell mode a lot when I'm at home.

I did notice that it would always happen every time I closed the lid while the fan was blasting away. It was usually after doing some intense debugging session involving Eclipse and AVD.

 

[0x000000]

Do you think Apple will fix the EFI firmware for it to be like before?
Do you think that other updates will fix the actual problem?
If Apple does not fix the problem(ever?), do you think it will be possible to fix it with 3rd party software/mods/tweaks?

I am willing to wait for a fix, but not forever :/

1. No, because they did it to fix something and probably had good reason to do so. They will not switch back to the old version, but they might fix the issues.

2. Maybe. It might help to file complaints or make topics in Apple's support forums so they know about these bugs and that people aren't happy.

3. Likely no. CoolBook was never updated, despite being a really great piece of software and was tweaked and updated for a very long time. It stopped very abruptly with Lion and the Sandy/Ivy-Bridge CPUs, so it's also very unlikely someone else will find a third-party-solution.

That may be different when using windows, though.

 

[pandaen]

1. No, because they did it to fix something and probably had good reason to do so. They will not switch back to the old version, but they might fix the issues.

2. Maybe. It might help to file complaints or make topics in Apple's support forums so they know about these bugs and that people aren't happy.

3. Likely no. CoolBook was never updated, despite being a really great piece of software and was tweaked and updated for a very long time. It stopped very abruptly with Lion and the Sandy/Ivy-Bridge CPUs, so it's also very unlikely someone else will find a third-party-solution.

That may be different when using windows, though.

Sounds pretty unlikely that we will see our rMBP's back to their old strenghts then

Already did file a complaint, but obviously no response..

What about the silent tweaks that it sounds like they are doing for the 15"'s while launching the 13"'s? Could that possibly be implemented in the "old" ones?

 

[inket]

<ignore this>

 

[inket]

theSeb is right.

Geniuses at Apple Stores say it's a normal and expected behavior (not a bug). Which ultimately means it's a big "**** you" to gamers, from Apple.

What I don't understand is why Apple bothered putting such powerful processor/gpu combo if they were going to nerf the **** out of them with an inconspicuous update. Like they didn't thoroughly test their products before releasing them.

Not to mention throttling a CPU 15°C below Tjunc is really aggressive. Actually, Apple should not even throttle CPUs/GPUs as they already have their own thermal throttling threshold set by Intel and Nvidia.

Anyway, we're trying ThrottleStop at the moment and we're having mixed results. Would be great if anyone can also try it. Set Multiplier to Turbo (31), enable Log File, Turn ON and game away or stress test then upload the .txt from ThrottleStop's folder.

ThrottleStop Guide for the interested.

 

[bill-p]

Have had my rMBP on 24/7 continuously stressing through various scenarios (Diablo 3, StarCraft 2, XCode, iPhoto, iMovie, etc...) now and still haven't seen the EFI bug back.

I think it's worth a try. Those of us who are affected by this bug may want to try toggling "Power Nap" ON/OFF.

I just did that, and my rMBP has not had the EFI bug back in 2 weeks. Prior to this, I had to reset EFI probably every few hours, which is crazy!

 

[Junkyman]

theSeb is right.

Geniuses at Apple Stores say it's a normal and expected behavior (not a bug). Which ultimately means it's a big "**** you" to gamers, from Apple.

What I don't understand is why Apple bothered putting such powerful processor/gpu combo if they were going to nerf the **** out of them with an inconspicuous update. Like they didn't thoroughly test their products before releasing them.

Not to mention throttling a CPU 15°C below Tjunc is really aggressive. Actually, Apple should not even throttle CPUs/GPUs as they already have their own thermal throttling threshold set by Intel and Nvidia.

Anyway, we're trying ThrottleStop at the moment and we're having mixed results. Would be great if anyone can also try it. Set Multiplier to Turbo (31), enable Log File, Turn ON and game away or stress test then upload the .txt from ThrottleStop's folder.

ThrottleStop Guide for the interested.

The "good" thing about Geniuses is that, contrary to public belief and their suggestive name, they rarely fulfill the original meaning of that title.

I have visited the genius bar several times now and every single genius had some different answers for me when I confronted them with my issue that revealed one common basis: blatant ignorance and zero knowledge. They had no idea what I was essentially talking about.

I'm not saying that they are not competent. But I think one can safely assume that they don't necessarily know about this very delicate and special issue. Or that they know of every issue and bug that haunts Macs.

Let's just hope that he was wrong and that we can count on Apple to solve this issue in near future. 'cause one thing is for sure: This is not a "featuere" in my eyes. They better fix this or else this is the worst possible way to destroy the experience of a rMBP.

I mean: It worked fine before the EFI, didn't it? They brought this mess on us, they surely can relieve us from it.

----------

Have had my rMBP on 24/7 continuously stressing through various scenarios (Diablo 3, StarCraft 2, XCode, iPhoto, iMovie, etc...) now and still haven't seen the EFI bug back.

I think it's worth a try. Those of us who are affected by this bug may want to try toggling "Power Nap" ON/OFF.

I just did that, and my rMBP has not had the EFI bug back in 2 weeks. Prior to this, I had to reset EFI probably every few hours, which is crazy!

I tried your proposed solution. At first I thought it worked, but it eventually came back after some occasions of sleep.

Can you describe in detail how you exactly eliminated the bug? Just uncheck/check power nap in the energy savings settings? Both battery and plugged in? Enable or disable power nap in the end?

 

[Eduardboon]

I never installed the EFI update but had to reset the SMC as well when Borderlands 2 began to run at 10fps max. Tad weird. Still, all of the sudden my Mac has become slower than usual again, SMC reset did NOT fix it. Too bad.. res drop from 1920 to 1440 in games was in order.. tempted to install the efi patch in the hopes it'll work better.

 

[bill-p]

I tried your proposed solution. At first I thought it worked, but it eventually came back after some occasions of sleep.

Can you describe in detail how you exactly eliminated the bug? Just uncheck/check power nap in the energy savings settings? Both battery and plugged in? Enable or disable power nap in the end?

I had "Power Nap" disabled for a long time... and then I just decided to re-enable it (only on power) since we just set up a new Time Capsule in the house.

Low and behold, it's been 3 weeks since I last had the EFI bug. I'm sailing happily now. I do short bursts of Diablo 3 grinding about 10 - 15 minute occasionally (at 2309 x something resolution), and I haven't seen my fps dip to sub 10 like before. Just did a 4-player MP run with 2 wizards firing off today, too. Shiny stuffs... but no fps dip whatsoever.

Of course everything else is snappier, too.

 

[Junkyman]

I had "Power Nap" disabled for a long time... and then I just decided to re-enable it (only on power) since we just set up a new Time Capsule in the house.

Low and behold, it's been 3 weeks since I last had the EFI bug. I'm sailing happily now. I do short bursts of Diablo 3 grinding about 10 - 15 minute occasionally, and I haven't seen my fps dip to sub 10 like before.

Of course everything else is snappier, too.

Still nothing on my side. :-/ The fix doesn't work for me.

I have now spontaneously decided to downgrade my OS. Running 10.8.1 now. Let's see if that has some impact.

 

[reery]

Same issue here on my rMPB.

I reset SMC every time before I boot into Windows, otherwise it's a 50/50 chance if throttling will occur or not.

 

[Junkyman]

So I made some very interesting observations today and I wanted to share those with you guys.

I did a downgrade yesterday from 10.8.2 to 10.8. Just to rule out that the OS itself is responsible for the throttling issues. And that indeed the EFI update caused all this mess.

I have excessivley tested the entire system, meaning somewhat between 20-30 sleeps (close lid, open lid), long sessions of sleep (more that one hour), with battery & AC power. Basically the entire day.

My result: Where the issue with throttling would surface after I used sleep with 10.8.2, I was now unable to produce the throttling. I booted into BootCamp (Windows 7) at least 10 times and playing Guild Wars 2 just to make sure. Everything went smoothly as hell.

Now, I'm not saying that it proves that 10.8.2 is indeed the culprit. However, I'm at this point inclined to take a bet that 10.8.2 may be causing the issue.

That said, I have just updated back to 10.8.2 again. While not having the throttling is great, the overall system in OS is definitely more snappy with 10.8.2. 10.8 and 10.8.1 both feel very sluggish to an extent that it's just not fun to work with it when you know how snappy the rMBP can be under 10.8.2.

Let's see how it goes from here.

One more thing: All of you who experience a long shut down now with 10.8.2: I can positively confirm that it is a bug that came along with 10.8.2. Under 10.8 and 10.8.1 the system shuts off immediately. Like at once. 1 sec. No white screen with spinning circle. Just bam. Off. <-- like to see that back, Apple!

 

[inket]

Interesting observation. Can anyone try and confirm ?

 

[Junkyman]

Alright, I will blame 10.8.2 from now on. Had a good night of solid sleep (that goes for both me and my rMBP ) unplugged. And guess what: Throttling is back!

So surprisingly: It's not EFI, it's 10.8.2. I'm not sure whether that's good or bad news though...

 

[Junkyman]

If it wasn't for the overall sluggishness of 10.8.1, I would recommend everyone suffering from this bug to downgrade. For me, however, the overall performance under Mac OSX is the most important part so I'll stick with 10.8.2 for the time being.

10.8.1:

+ super fast shut down
+ no throttling bug

- overall performance of the OS slower than 10.8.2

10.8.2:

+ very snappy (goes for Safari, Mission Control etc.)

- slow shut down (white screen, spinning wheel)
- throttling bug when put to sleep, thus SMC now and then in order

It's really a choice between the devil and beelzebub.

 

[timmerk]

Junkyman, thanks for all your work and reports! My Retina is getting here Tuesday. I will report how it acts with 10.8.2 preinstalled and the latest EFI update (assuming it ships with the latest of both).

 

[Junkyman]

Junkyman, thanks for all your work and reports! My Retina is getting here Tuesday. I will report how it acts with 10.8.2 preinstalled and the latest EFI update (assuming it ships with the latest of both).

You are very welcome!

That would be great if you could share your experience!

 

[inket]

We know that:

- Resetting SMC may fix the issue.
- Messing with the Power Nap toggle can, too.
- Sleep mode sometimes makes the bug reappear.
- The problem appears after updating to 10.8.2/EFI.

Isn't it safe to assume Power Nap is the source of the throttling since it was added with 10.8.2 ?

Let's assume the rMBP has a "power state" called Regular and installing 10.8.2 adds another one, say Power Nap, that's activated automatically.

It makes sense for the Power Nap state to have aggressive throttling so that the Mac doesn't get hot when it's off and doing "background" tasks.

What if the rMBP gets stuck on that state and never goes back to Regular ? Making the SMC or EFI think it's just Power Nap that's running instead of your brand new PC game ? And coincidentally, resetting the SMC makes the computer forget the state it was on, or default to Regular ?

Maybe this is a stretch but it makes sense to me. If anyone else gets rid of the throttling by going back to 10.8.0 we can definitely blame the issue on Power Nap/10.8.2.

 

[Junkyman]

We know that:

- Resetting SMC may fix the issue.
- Messing with the Power Nap toggle can, too.
- Sleep mode sometimes makes the bug reappear.
- The problem appears after updating to 10.8.2/EFI.

Isn't it safe to assume Power Nap is the source of the throttling since it was added with 10.8.2 ?

Let's assume the rMBP has a "power state" called Regular and installing 10.8.2 adds another one, say Power Nap, that's activated automatically.

It makes sense for the Power Nap state to have aggressive throttling so that the Mac doesn't get hot when it's off and doing "background" tasks.

What if the rMBP gets stuck on that state and never goes back to Regular ? Making the SMC or EFI think it's just Power Nap that's running instead of your brand new PC game ? And coincidentally, resetting the SMC makes the computer forget the state it was on, or default to Regular ?

Maybe this is a stretch but it makes sense to me. If anyone else gets rid of the throttling by going back to 10.8.0 we can definitely blame the issue on Power Nap/10.8.2.

Correct me if I'm wrong, but as far as I know the rMBP had power nap already with 10.8. The update 10.8.2 added power nap only to the MacBook Air.

Also, I tested toggling power nap on and off. I have it now on for both AC and battery. As far as I can tell it indeed prevents the throttling bug from surfacing. But there's a minor caveat: It only works if the Mac really has a solid wifi connection. Apparently if the Mac loses the wifi connection it just goes to sleep the way it properly is supposed to when power nap is deactivated.

I can tell this because when I use my Mac at home the bug does not surface when I activate power nap. However, when I go to university and come back, the bug is all of a sudden there. I can only explain it by the fact that I don't have a wifi connection when commuting.

Does that make sense? xD