15" Retina MBP: CPU t under full load

maratus · Mar 8, 2013

walkie said:
Come on! stop the paranoia about CPU temperatures and enjoy your rMBP, if it get toasted Apple will replace it for you, if you spend a lot of money on a rMBP you'd better focus on doing something productive on it.

I can't be 100% sure, but for some of my long-time simulations I encountered memory-related errors. It may happen when IMC is locally overheated.

Modern CPUs won't fry even without heatsink. It seems that other nearby components with accessible sensors aren't subject to temperatures higher that +30C to ambient. Airflow is decent either. Therefore I'm not concerned with reliability, but rather random glitches that may occur. As a rule of thumb you should never use CPU operating at TJMax temperature, even though it won't damage itself.

alex4040 · Mar 12, 2013

Hi!
Could somebody run Intel Power Gadget on a fully loaded rMBP 2.3 please?
(as Ploki did for 2.7rMBP)

It is interesting to see the real value of the power consumption. However all rMBP have the same TDP=45W according to the intel official site. But just to make sure...

In the current rMBP design thermal engineers put (1) long 2mm heat pipe. and the CPU is in the center of it. So this (1) heat pipe acts like (2) short heat pipes for each radiator. So each of the (2) heat pipes take the half of thermal power 45/2=22.5W (if NVIDIA is turned off). As I know from talking with heat pipes' designers the critical power value for a 2 mm heat pipe is about 20W. With further power increasing the temperature begin to rise significantly (10 times faster for each W).

So I suppose that at this critical point the value of even 1W of power increasing is important and the difference of 300MHz can be significant.

dastinger · Mar 12, 2013

Ran this yesterday with Power Gadget up and running. Didn't log the results but I'll do it again once I get home and post here. I was trying to induce the SMC bug. I know my rMBP suffers from it but I haven't been able to, now that I want it, to test and discuss this with a hardware engineer.

dastinger · Mar 12, 2013

Here are my results.

Mid 2012 15" rMBP

Been running the command for 10 minutes when I took these screenshots:

These were taken some minutes after terminating the processes and with Chrome running 9 tabs (first and second were not taken exactly at the same time):

My UI started lagging after the test so I guess I managed to trigger the SMC bug. I'll play a bit under Windows in a minute and be sure about it.

alex4040 · Mar 12, 2013

Thanks dastinger

It seems i7-2.3GHz has 17% less maximum power consumption than i7-2.7GHz (39.33W and 46.65W respectively as reported in posts #32 and #54).
Yet we have not enough statistics for making conclusions.

I'll get my new rMBP 2.7GHz delivered on 03/22 and will run the same tests.

alex4040 · Mar 12, 2013

According to screenshots of post #54
I see that temps of GPU were 63C and GPU diod 69C while NVIDIA 650m was switched off. I believe NVIDIA GPU was heated by the hot heat pipe (CPU and GPU share the single heat pipe). So we know the temperature of heat pipe at the location of GPU. It is about 69C

Queen6 · Mar 13, 2013

dastinger said:
Here are my results.My UI started lagging after the test so I guess I managed to trigger the SMC bug. I'll play a bit under Windows in a minute and be sure about it.

Now this is really what bugs me, if i perform the same test on my 2.3 Retina there is absolutely no ill effects, so why are some machines triggering this "SMC bug" and some not, my own Retina will pump out over 3GHz all day every day, with or without the GPU`s being fully loaded, it only had the very slightest lag under Lion, and now with 10.8.2 nonexistent.

I have seen recorded temperatures up to 104C and still no "SMC bug" personally i am beginning to think that all Mac`s are not equal, which is concerning given the price point, right now it looks like if you get a good one "all fair and well" if you get a bad depending on circumstance you may well be stuck with it. The present issue with the 2012 Retina`s and the "SMC Bug" is longstanding yet there appears to be no solution on offer from Apple.

As my work commitment is now changing, i am very much of the mind to go back one 15" MBP and build a desktop "Hacintosh" as a Workstation it will be faster, and i will have control of whats in the case...

Mr MM · Mar 13, 2013

Queen6 said:
Now this is really what bugs me, if i perform the same test on my 2.3 Retina there is absolutely no ill effects, so why are some machines triggering this "SMC bug" and some not, my own Retina will pump out over 3GHz all day every day, with or without the GPU`s being fully loaded, it only had the very slightest lag under Lion, and now with 10.8.2 nonexistent.

I have seen recorded temperatures up to 104C and still no "SMC bug" personally i am beginning to think that all Mac`s are not equal, which is concerning given the price point, right now it looks like if you get a good one "all fair and well" if you get a bad depending on circumstance you may well be stuck with it. The present issue with the 2012 Retina`s and the "SMC Bug" is longstanding yet there appears to be no solution on offer from Apple.

As my work commitment is now changing, i am very much of the mind to go back one 15" MBP and build a desktop "Hacintosh" as a Workstation it will be faster, and i will have control of whats in the case...

the problem here resides on the belief that the EFI bug is triggered by the load on the cooling system or cpu/gpu

simply ask if this isnt the case at all

dastinger · Mar 13, 2013

Queen6 said:
Now this is really what bugs me, if i perform the same test on my 2.3 Retina there is absolutely no ill effects, so why are some machines triggering this "SMC bug" and some not, my own Retina will pump out over 3GHz all day every day, with or without the GPU`s being fully loaded, it only had the very slightest lag under Lion, and now with 10.8.2 nonexistent.

I have seen recorded temperatures up to 104C and still no "SMC bug" personally i am beginning to think that all Mac`s are not equal, which is concerning given the price point, right now it looks like if you get a good one "all fair and well" if you get a bad depending on circumstance you may well be stuck with it. The present issue with the 2012 Retina`s and the "SMC Bug" is longstanding yet there appears to be no solution on offer from Apple.

As my work commitment is now changing, i am very much of the mind to go back one 15" MBP and build a desktop "Hacintosh" as a Workstation it will be faster, and i will have control of whats in the case...

It's weird. Right after posting the results, I booted into Windows and played some FIFA without resetting the SMC and no lag at all. I've been playing now as well and still no lag. I know it'll suffer from it once again some time but I'm not managing to trigger it lately. I wish it stayed that way.

Queen6 · Mar 13, 2013

dastinger said:
It's weird. Right after posting the results, I booted into Windows and played some FIFA without resetting the SMC and no lag at all. I've been playing now as well and still no lag. I know it'll suffer from it once again some time but I'm not managing to trigger it lately. I wish it stayed that way.

As I have said "there is more to this, than meets the eye" and Apple`s typical "cloak and dagger" approach is helping no one...

----------

Mr MM said:
the problem here resides on the belief that the EFI bug is triggered by the load on the cooling system or cpu/gpu

simply ask if this isnt the case at all

I know, and this the point i am attempting to make why some systems trigger the "SMC bug" and not others. I have deliberately loaded my own Retina to the point where it`s at 100% capacity for days, yet no throttling, yet other identical machines throttle due to all intents and purpose an SMC issue, the only question is why and Apple`s typical "cloak and dagger" tactics are helping nobody...

alex4040 · Apr 7, 2013

Early 2013
rMBP 15" 2.7GHz 16GB RAM
MAC OS 10.8.3

My results:

maratus · Apr 8, 2013

Interesting.

Here're my updated results for comparison.

Early 2013
rMBP 15" 2.4GHz, 16GB RAM
Mac OS 10.8.2

First is 8 processes in terminal -> 37W
Second is 8 threads of Prime95 -> almost 44W

So far, the Palm rest / Battery temps seem to reflect ambient temps more or less consistently.

Performance of the thermal interface can be derived from CPU Die minus GPU Die - Analog, because when 650M is switched off, it's temperature is very close to that of the heat sink.

The efficiency of the fans/radiator can be derived from GPU Die - Analog minus ambient or, which is somewhat safe to say, Palm Rest/Battery average. Palm Rest is being heated up a little bit but it stays within 2C of idle temp for quite a long time.

These calculations aren't extremely precise but are good enough to do a direct comparison. My battery / palm rest temps are slowly rising to 32C now when idling as I switched it on only 30min ago, so we probably have the same ambient temperature. At the same time we both have 66C for the heat sink when CPU wattage is similar (mine Prime95 and yours terminal load)

silvetti · Apr 8, 2013

So basically the lower clocked rMBP work 20 degress lower while on max cpu usage ?

maratus · Apr 8, 2013

silvetti said:
So basically the lower clocked rMBP work 20 degress lower while on max cpu usage ?

Yes (it was the whole point of the thread), and 300MHz of difference in clock speed reduces down to 180MHz or less.

silvetti · Apr 8, 2013

Guess if it stays the same with haswell I'm getting the lower end model then...

Search

Search

15" Retina MBP: CPU t under full load

maratus

macrumors 6502a

alex4040

macrumors newbie

dastinger

macrumors 6502a

dastinger

macrumors 6502a

alex4040

macrumors newbie

alex4040

macrumors newbie

Queen6

macrumors G4

Mr MM

macrumors 65816

dastinger

macrumors 6502a

Queen6

macrumors G4

alex4040

macrumors newbie

Attachments

maratus

macrumors 6502a

Attachments

silvetti

macrumors 6502a

maratus

macrumors 6502a

silvetti

macrumors 6502a

Our Staff