When the new Retina MacBook Pro launched, Apple had this image floating around which displayed the extra intake vents and new cooling system layout. Those of us who push our MacBook Pros hard are very aware of the space-heater quality they take on. Did the new chassis fix this?
Apple updated the old chassis with Kepler and IVB alongside the new Retina model. While all the focus was on the Retina, I've not seen anybody pit the chassis against each other to see whether it was the new parts that caused a thermal improvement, or the revised cooling system.
To cut the suspense short, the Retina MacBook Pro is substantially cooler under my testing. Read on for details.
Up until a year ago I lived in the San Francisco Bay Area, and would zip over to Apple Store Palo Alto when something new and shiny launched to run tests. Since then I have moved to Adelaide, Australia. There is no Apple Store here, and all of the retailers tend to get new shiny toys quite awhile after launch. I attempted to crowdsource this data from the MacRumors community, but nobody supplied anything more than anecdotal evidence (thanks regardless). Recently, Myer at Marion got a Retina model in, and I pounced on the chance to irritate the staff there by running their machines very hot and loud.
The models tested...
- Base model Retina MacBook Pro - 2.3GHz IVB, 8GB RAM, Nvidia GT 650M (1GB), 256GB SSD.
- Base model traditional MacBook Pro 15" - 2.3GHz IVB, 4GB RAM, Nvidia GT 650M (512MB), 500GB 5400RPM HD.
These are not the higher-shelf machines, but they are close enough to compare like-for-like.
The testing methodology...
- Install and run Temperature Monitor. Specify CPU Temp Diode, dGPU Temp Diode, and Left Palmwrest sensors, then apply these to the chart.
- Launch Activity Monitor and display CPU History to be sure all threads were fully utilised.
- Launch OpenGL Driver Monitor. Attach it to the GT 650M. Monitor parameters GPU Core Utilization and CPU Memory Utilization. To make sure that the GPU is being stressed.
- Launch Prime95 and set up five torture test workers.
- Launch SmallluxGPU, set up the Glasstable scene, using GPU only, at 1024x768, and run it in interactive mode.
- Use Quartz Debug to confirm that the Nvidia GPU was being used.
I had some concern about SmallluxGPU not properly utilising the new GPU, but it seemed to identify and use it just fine. I also was not fully sure that Temperature Monitor would work properly on the rMBP, but it seems like the sensors it is reading work properly.
The results...
- The rMBP's fans seemed to ramp up sooner. As soon as the GPU got to ~87 C the fans starting ramping and were able to bring the GPU temp down to a stable 80 C.
- In the older chassis, the fans waited until the GPU temp was quite high, about 95 C, before ramping up. They then brought the GPU down to a stable 90C. That's a full ten degrees hotter than the Retina MBP.
- CPU temperature in the classic MBP chassis settled around 76 C. In the rMBP, the temperature was 66 C. Again, a 10 degree difference in favour of the Retina chassis.
- I played a bit with the Prime95 worker count in order to get the highest possible utilisation across both CPU and GPU. Five workers was generally the best balance allowing 98% CPU and 80-90% GPU. If too many CPU cycles went into Prime, GPU utilization dropped considerably.
- Both machines improve markedly from last year, largely on the laurels of Ivy Bridge's and Kepler's efficiency. Thanks Intel and Nvidia. However, the Retina chassis appears to allow better (and more responsive) cooling.
- On the classic unibody MBP, set to four Prime workers, I was able to achieve ~90% GPU utilization. However, the CPU was then around 85% total utilization. When I set Prime to five workers, GPU utilization dropped to around 85%, but the CPU was at 100%.
- On the Retina MacBook Pro, I was able to maintain five Prime workers and ~90% on the GPU. I am not sure why this is - theoretically the compute is identical. System memory is probably the differentiating factor.
- I was not able to read or set fan speeds due to not being able to install SMCFanControl without an Admin password, which I did not have.
Attached images... (see bottom of post)
- Retina MacBook Pro Fan Ramp and Last Generation MacBook Pro Fan Ramp
- Retina MacBook Pro and Last Generation MacBook Pro Process Utilization and Info
A status point to note as I finished up the test. Showing various metrics and info on both machines.
- Retina MacBook Pro Desktop image
This is just a shot showing the environment I tested in. I had the rMBP set to 1920x1200 (doubled) for maximum desktop space. It was very usable and excellent.
Any questions or comments, don't hesitate. Cheers!
Apple updated the old chassis with Kepler and IVB alongside the new Retina model. While all the focus was on the Retina, I've not seen anybody pit the chassis against each other to see whether it was the new parts that caused a thermal improvement, or the revised cooling system.
To cut the suspense short, the Retina MacBook Pro is substantially cooler under my testing. Read on for details.
Up until a year ago I lived in the San Francisco Bay Area, and would zip over to Apple Store Palo Alto when something new and shiny launched to run tests. Since then I have moved to Adelaide, Australia. There is no Apple Store here, and all of the retailers tend to get new shiny toys quite awhile after launch. I attempted to crowdsource this data from the MacRumors community, but nobody supplied anything more than anecdotal evidence (thanks regardless). Recently, Myer at Marion got a Retina model in, and I pounced on the chance to irritate the staff there by running their machines very hot and loud.
The models tested...
- Base model Retina MacBook Pro - 2.3GHz IVB, 8GB RAM, Nvidia GT 650M (1GB), 256GB SSD.
- Base model traditional MacBook Pro 15" - 2.3GHz IVB, 4GB RAM, Nvidia GT 650M (512MB), 500GB 5400RPM HD.
These are not the higher-shelf machines, but they are close enough to compare like-for-like.
The testing methodology...
- Install and run Temperature Monitor. Specify CPU Temp Diode, dGPU Temp Diode, and Left Palmwrest sensors, then apply these to the chart.
- Launch Activity Monitor and display CPU History to be sure all threads were fully utilised.
- Launch OpenGL Driver Monitor. Attach it to the GT 650M. Monitor parameters GPU Core Utilization and CPU Memory Utilization. To make sure that the GPU is being stressed.
- Launch Prime95 and set up five torture test workers.
- Launch SmallluxGPU, set up the Glasstable scene, using GPU only, at 1024x768, and run it in interactive mode.
- Use Quartz Debug to confirm that the Nvidia GPU was being used.
I had some concern about SmallluxGPU not properly utilising the new GPU, but it seemed to identify and use it just fine. I also was not fully sure that Temperature Monitor would work properly on the rMBP, but it seems like the sensors it is reading work properly.
The results...
- The rMBP's fans seemed to ramp up sooner. As soon as the GPU got to ~87 C the fans starting ramping and were able to bring the GPU temp down to a stable 80 C.
- In the older chassis, the fans waited until the GPU temp was quite high, about 95 C, before ramping up. They then brought the GPU down to a stable 90C. That's a full ten degrees hotter than the Retina MBP.
- CPU temperature in the classic MBP chassis settled around 76 C. In the rMBP, the temperature was 66 C. Again, a 10 degree difference in favour of the Retina chassis.
- I played a bit with the Prime95 worker count in order to get the highest possible utilisation across both CPU and GPU. Five workers was generally the best balance allowing 98% CPU and 80-90% GPU. If too many CPU cycles went into Prime, GPU utilization dropped considerably.
- Both machines improve markedly from last year, largely on the laurels of Ivy Bridge's and Kepler's efficiency. Thanks Intel and Nvidia. However, the Retina chassis appears to allow better (and more responsive) cooling.
- On the classic unibody MBP, set to four Prime workers, I was able to achieve ~90% GPU utilization. However, the CPU was then around 85% total utilization. When I set Prime to five workers, GPU utilization dropped to around 85%, but the CPU was at 100%.
- On the Retina MacBook Pro, I was able to maintain five Prime workers and ~90% on the GPU. I am not sure why this is - theoretically the compute is identical. System memory is probably the differentiating factor.
- I was not able to read or set fan speeds due to not being able to install SMCFanControl without an Admin password, which I did not have.
Attached images... (see bottom of post)
- Retina MacBook Pro Fan Ramp and Last Generation MacBook Pro Fan Ramp
These charts are interesting. It appears that the Retina MacBook Pro is able to respond to more points on the temperature curve than the older MacBook Pro.
The older MacBook Pro lets the temperature climb very high, then it engages the fans at their full ~6,000RPM and brings the temperature back down quickly.
The Retina MacBook Pro appears to spin the fans at progressively higher RPMs as the machine gets hotter. This makes the temperature gain more gradual. All the dips in the rMBP's chart are me starting and stopping Prime quickly to try and adjust worker counts. It makes the chart a bit less accurate.
The older MacBook Pro lets the temperature climb very high, then it engages the fans at their full ~6,000RPM and brings the temperature back down quickly.
The Retina MacBook Pro appears to spin the fans at progressively higher RPMs as the machine gets hotter. This makes the temperature gain more gradual. All the dips in the rMBP's chart are me starting and stopping Prime quickly to try and adjust worker counts. It makes the chart a bit less accurate.
- Retina MacBook Pro and Last Generation MacBook Pro Process Utilization and Info
A status point to note as I finished up the test. Showing various metrics and info on both machines.
- Retina MacBook Pro Desktop image
This is just a shot showing the environment I tested in. I had the rMBP set to 1920x1200 (doubled) for maximum desktop space. It was very usable and excellent.
Any questions or comments, don't hesitate. Cheers!
