15" Retina MBP: CPU t under full load

[maratus]

Nice feedback, folks.

Now we need someone to confirm the 2.3GHz being somewhat cooler. Stay tuned!

[DTKblaster]

I'm on the 2.3ghz base model, and with 8 terminal windows running that command, the CPU temp hovers around 100 degrees C.

But this seems to be the target temperature as defined by intel or apple, as the fan speed was only at 4200 (left)/3900 (right). It reached this number very slowly, after many minutes at full load (it spent the first few mins with inaudible fan noise, even at 100+ degrees).

I put it on my bed (block the air vents a little) and the fans hit 5000+, which brought it to the target temp around 100 again. Freeing the air vents reduced fan speed to the original values.

The intel tool reports a constant 3.1ghz, while consuming just under 40 watts (hovers around 39.88w).

To help with the comparison, I forced both fans to max (6K/5.5K) and left it for about 10 mins. The temp settled around 83 degrees. What was really interesting was that the reported power consumption was reduced by a couple of watts: down to ~37.92w, which makes sense when you consider that the resistance of a wire increases with temperature.

The enclosure base temp was 29 degrees.

[swerve147]

Originally Posted by DTKblaster

I'm on the 2.3ghz base model, and with 8 terminal windows running that command, the CPU temp hovers around 100 degrees C.

But this seems to be the target temperature as defined by intel or apple, as the fan speed was only at 4200 (left)/3900 (right). It reached this number very slowly, after many minutes at full load (it spent the first few mins with inaudible fan noise, even at 100+ degrees).

I put it on my bed (block the air vents a little) and the fans hit 5000+, which brought it to the target temp around 100 again. Freeing the air vents reduced fan speed to the original values.

The intel tool reports a constant 3.1ghz, while consuming just under 40 watts (hovers around 39.88w).

To help with the comparison, I forced both fans to max (6K/5.5K) and left it for about 10 mins. The temp settled around 83 degrees. What was really interesting was that the reported power consumption was reduced by a couple of watts: down to ~37.92w, which makes sense when you consider that the resistance of a wire increases with temperature.

The enclosure base temp was 29 degrees.

Same here, 2.3/8/256, but my average CPU temps under load using the OP's Terminal commands were about 89 degrees C. I did however, use the Temperature Gauge app from the App Store (which claims to be updated for rMBP). If I eyeball the smcFanControl app while running it, it seems to be a little higher than that.

Like DTK, though, running this test does not force max fans to kick in, they hover around 4000k as mentioned. I had to use smc to force max fans, and the average temp then dropped to about 85 degrees C under full load.

[maratus]

Nice, it seems that other 2.3GHz models perform in line with my findings. It's very interesting though I'm not exactly sure why 0.3GHz makes that much difference. It could be that in the beginning only 2.3GHz chips were stable enough to operate at lower voltages, while first 2.6GHz and 2.7GHz chips required more, which resulted in even higher power consumption.

I believe that newer steppings of 2.6/2.7GHz chips may work at lower voltages as well (even 2.8GHz chip was finally introduced) but Apple didn't update their built-in frequency/voltage curves. That's why even if you buy the newest machine you won't benefit from the "cooler" processor unless you undervolt manually.


P.S. Yes, forcing fans to the max will result in data comparable with mine, so it is essential. And I recommend always running them at full speed for any prolonged activities that stress CPU a lot.

P.P.S. 2.3Ghz/768Gb/16Gb is coming to the test, but I might postpone it.

[Ploki]

Originally Posted by maratus

Nice, it seems that other 2.3GHz models perform in line with my findings. It's very interesting though I'm not exactly sure why 0.3GHz makes that much difference. It could be that in the beginning only 2.3GHz chips were stable enough to operate at lower voltages, while first 2.6GHz and 2.7GHz chips required more, which resulted in even higher power consumption.

I believe that newer steppings of 2.6/2.7GHz chips may work at lower voltages as well (even 2.8GHz chip was finally introduced) but Apple didn't update their built-in frequency/voltage curves. That's why even if you buy the newest machine you won't benefit from the "cooler" processor unless you undervolt manually.


P.S. Yes, forcing fans to the max will result in data comparable with mine, so it is essential. And I recommend always running them at full speed for any prolonged activities that stress CPU a lot.

P.P.S. 2.3Ghz/768Gb/16Gb is coming to the test, but I might postpone it.

Turbo boost?

[maratus]

Originally Posted by Ploki

Turbo boost?

Turbo boot what?

It should decrease the difference if anything as it can't be aggressive when overheating is detected! 2.3Ghz model on the other hand always can be at maximum boost due to much better temps.

P.S. Just to clarify, I know how turboboost works and it's dependance on the number of active threads.

[Ploki]

Originally Posted by maratus

Turbo boot what?

It should decrease the difference if anything as it can't be aggressive when overheating is detected! 2.3Ghz model on the other hand always can be at maximum boost due to much better temps.

P.S. Just to clarify, I know how turboboost works and it's dependance on the number of active threads.

True.
CPU power jumped almost to 60W.

it starts throthling after 15secs approx to 3.40-3.45ghz, which is still more than 3.1ghz of 2.3 cpu

fans floating around 5k rpm afte 40secs, increasing after 1 minute to 5.6k and rising slowly towards 6k. (Left fan, right is at 5,4k!)
Right fan is always about 400rpm behind...

[maratus]

Originally Posted by Ploki

True.
CPU power jumped almost to 60W.

it starts throthling after 15secs approx to 3.40-3.45ghz, which is still more than 3.1ghz of 2.3 cpu

fans floating around 5k rpm afte 40secs, increasing after 1 minute to 5.6k and rising slowly towards 6k. (Left fan, right is at 5,4k!)
Right fan is always about 400rpm behind...

Right fan has lower maximum rpm by design, so it's normal.

[Schranke]

Originally Posted by maratus

Right fan has lower maximum rpm by design, so it's normal.

With smcfancontrol you can max both fans at 6000rpm
but the mac will normaly keep the 1 fan a little slower.
why i do not know

[maratus]

Originally Posted by Schranke

With smcfancontrol you can max both fans at 6000rpm
but the mac will normaly keep the 1 fan a little slower.
why i do not know

If it reads that way in smcfancontrol, it doesnt mean that it actually runs at 6000rpm

[maratus]

So, I finally have got my hands on 2.4GHz model and I can confirm that this CPU doesn't go beyond 85C (ambient is ~23C). The difference between 2.7GHz model (small cache version) is unknown. Generally, it seems to be very similar to 2.3GHz

[laurihoefs]

Here are the temperatures from my mid 2012 base model, if you are still collecting stats:

2.3Ghz, 8Gb RAM
Production year: 2012
Production week: 37 (September)

CPU: 78C
Battery/Enclosure: 29C / 30C
Ambient temp: ~19C

Seems pretty similar to the temperatures in the first post, considering my fairly low ambient temperature.

[Ploki]

Originally Posted by maratus

So, I finally have got my hands on 2.4GHz model and I can confirm that this CPU doesn't go beyond 85C (ambient is ~23C). The difference between 2.7GHz model (small cache version) is unknown. Generally, it seems to be very similar to 2.3GHz

What about idle temperatures?

[Dyno-Mike]

I wonder if the lower CPU temperatures on the 2.3ghz CPU are the reason I haven't had the dreaded SMC bug at all on my base rMBP?

[Ploki]

Originally Posted by Dyno-Mike

I wonder if the lower CPU temperatures on the 2.3ghz CPU are the reason I haven't had the dreaded SMC bug at all on my base rMBP?

what SMC bug

[IGregory]

Originally Posted by maratus

So I will probably get a fresh 2.3GHz model to double check that difference. Surprisingly, I see a lot of threads and discussions about heat but yet no one is bothered to do a test so that we can have at least general view on this problem. C'mon people, do some tests and post the results! Everyone will benefit out of this!

No, I don't go looking for a problem where none exists.

[agaskew]

Originally Posted by Ploki

what SMC bug

You can't be serious. Search the forum.


Mid 2012 2.7Ghz/16Gb/768 - same results: 6k fan, 103C CPU after about 5 mins of running 8 'yes' processes.

You don't need 8 terminal windows, you can run each 'yes' command in the background if you put an ampersand on the end:

$ yes > /dev/null &

And when you're done, a way to kill them all off could be:

pgrep yes | xargs kill -9

What impresses me is how useable the machine remains when running flat out, and how quickly things return to normal once the tests are completed.

[Ploki]

Originally Posted by agaskew

You can't be serious. Search the forum.

What impresses me is how useable the machine remains when running flat out, and how quickly things return to normal once the tests are completed.

sorry. I haven't encountered it. Seems it has nothing to do with CPU type.

How come 2.3 is so much cooler than ALL other cpus? 2.6/2.7 is pretty on par with 2.7/2.8

I agree with the last paragraph. My old MBP when cpu got @75%, fans started spinning and it wouldn't come down... Ever. just kept being hot until I left it alone a bit.

[maratus]

Originally Posted by Ploki

sorry. I haven't encountered it. Seems it has nothing to do with CPU type.

How come 2.3 is so much cooler than ALL other cpus? 2.6/2.7 is pretty on par with 2.7/2.8

I agree with the last paragraph. My old MBP when cpu got @75%, fans started spinning and it wouldn't come down... Ever. just kept being hot until I left it alone a bit.

Maybe because Intel stays on the safe side and only allows reduced voltage supply for their base processor? Otherwise, why the hell they released 2.4GHz model in September (as an upgrade to 2.3Ghz!!!) when 2.6GHz was already present on the market?

Usually, they only add higher frequency models with nothing in between.

[Ploki]

Originally Posted by maratus

Maybe because Intel stays on the safe side and only allows reduced voltage supply for their base processor? Otherwise, why the hell they released 2.4GHz model in September (as an upgrade to 2.3Ghz!!!) when 2.6GHz was already present on the market?

Usually, they only add higher frequency models with nothing in between.

I thought the 2.3/ higher clocked are just about "which CPU goes into the which bin" in the factory. Seems odd it runs so significantly cooler.

[maratus]

Originally Posted by Ploki

I thought the 2.3/ higher clocked are just about "which CPU goes into the which bin" in the factory. Seems odd it runs so significantly cooler.

Yes, but otherwise Intel wouldn't be selling them at the same price. I mean, why would any manufacturer order a lower clocked model for exactly the same price when you can get +300MHz to brag about at no cost like extra heat?

P.S. Apple charges 100$ for this upgrade just because it can.

P.P.S. Unlike dual-core chips, these are operating on the verge of the cooling system capabilities - it's not mainly TDP issue, but much higher heat flux density for the CPU core. This is why everything stays cooler when gaming despite much higher total load from CPU + GPU combined.

In addition to that, they are also designed to run @ up to plus 700-1000MHz boost even under 100% / 8 threads. So we deal with at least 3.0GHz / 3.3GHz constant speed and corresponding extra voltage boost.

[walkie]

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.

[eezacque]

Originally Posted by walkie

Come on! stop the paranoia about CPU temperatures and enjoy your rMBP, if it get toasted Apple will replace it for you,

That is, only if you're lucky enough to see it toasted within the Apple Care period. Don't expect it to last much longer, as defects usually mean the logic board needs to be replaced, which is not economical after three years. After guarantee expires, it is economically dead.

[Ploki]

Originally Posted by maratus

Yes, but otherwise Intel wouldn't be selling them at the same price. I mean, why would any manufacturer order a lower clocked model for exactly the same price when you can get +300MHz to brag about at no cost like extra heat?

P.S. Apple charges 100$ for this upgrade just because it can.

P.P.S. Unlike dual-core chips, these are operating on the verge of the cooling system capabilities - it's not mainly TDP issue, but much higher heat flux density for the CPU core. This is why everything stays cooler when gaming despite much higher total load from CPU + GPU combined.

In addition to that, they are also designed to run @ up to plus 700-1000MHz boost even under 100% / 8 threads. So we deal with at least 3.0GHz / 3.3GHz constant speed and corresponding extra voltage boost.

but 2.7/2.8 is more expensive isn't it?

thanks for the post, quite insightful.

[Queen6]

Interesting thread and one that matches my own observations; my own 2.3 Mid 2012 Retina does indeed run very cool and irrespective of load will not throttle, it can be pushed over 100C, however this takes a very concerted effort. The system tops out at 3.1GHz and it will stay at this level indefinitely without any indication of throttling.

I use UltraFan to help the fans cut in sooner and reduce the effect of thermal shock, as we are all aware Apple tends to compromise on cooling over noise intrusion. UltraFan in general spools up the fans sooner at a user predetermined value, for the most part the fans run at default value unless the system comes under a load that pushes temperatures beyond the preset.

Apple`s own cooling algorithm does equally work, however as ever the emphasis is on quiet system, I find that by the time the fans do spool up to higher rpm`s the system is already thermally saturated, therefore reducing the cooling efficiency. With UltraFan you can get ahead of the curve and lessen the thermal saturation, net result is lower temps, lower fan rpm`s, and a shorter duration.

When i tested my 2.3 Retina i disabled UltraFan, i was attempting to induce the much discussed SMC bug, even with higher loads than requested here on both CPU & GPU no bug and no throttling. Same test with UltraFan enabled, resulted in lower temps as the cooling system cut in far sooner, preventing thermal saturation of the case, components.