Mods: I've created this thread separate from the Core i9 thread as I thought it would be less of headache. The Core i9 is definitely throttling more than that of the i7, but I thought it was still worth mentioning.
I upgraded my MacBook Pro from the base model 2015 15" to the top of the line off-the-shelf 2018 15". This is the 2.6GHz Core i7 paired with the Radeon 560X GPU. My previous MacBook had no dGPU and only had a 2.2GHz Quad Core Haswell i7. One thing that I always found amazing about that machine was that the CPU could almost max out and hold it's turbo boost in clamshell mode while rendering or exporting in Final Cut. I knew this new machine would have more to cool having a dGPU, so I was expecting some possible throttling. For the record, I normally use the machine in clamshell mode hooked up via DisplayPort to a Dell P2715Q 4K display.
As soon as I got the machine home, I noticed that while running a test in Final Cut the machine was getting extremely warm. When I opened Power Gadget, I saw some pretty remarkable throttling. Since then, I've been using Cinebench to test and see just how bad it gets. With the machine in clamshell mode hooked up to my 4K display and doing a cold test in a 78F room, the machine started throttling after about 25-30 seconds. Now, by throttling, I mean Power Gadget was reporting the current speed below the base clock speed. We'll start by seeing a quick spike to <4GHz, then a quick decrease to ~3.3GHz, when it starts steadily declining to ~3.0GHz, where it finally falls off the cliff and drops to about ~1.7GHz. From there I would see occasional peaks up to ~3GHz, followed by a drop to at least 2GHz, with an average drop to ~1.8GHZ, and at it's worst point hitting ~1.3GHz. FWIW, this CPU is supposed to be able to Turbo to 4.3GHz, but the fastest I have seen so far is 4.1.
Since no fan control software yet supporting this machine, I have to 'manually' cause the fans to spin up be running the test multiple times. The first three tests provided varying results: 804cb on the first test, 749cb on the second, and 771cb on the third, which is where the test score started to stabilize. By this time, it was throttling after around 20-25 secs, dropping to ~1.4GHz and jumping back to ~2.5GHz. It hit ~3.3GHz about 3 times in one test, before dropping to ~2.2. After the throttling, it's lowest point was 1.4GHz and highest was 3.4GHz, hitting these speeds 3 times each. Before throttling, it's highest speed was 4GHz which it maintained for 1-2 secs.
Finally, I decided to try the machine by itself, without a display connected. On the first pass, it turboed to ~3.7GHz, and after 5 secs dropped to ~3.3GHz. From there it began a steady decline to 3GHz over the course of about 20 secs, before nosediving to ~2.2GHz. This time it was certainly easier for it to maintain more speed, with the lowest point being ~1.6GHz reached 3 times, and the highest point being ~3.3GHz reached once. The overall graphic mostly showed the CPU running below the base clock speed of 2.6GHz. It scored 849cb in Cinebench.
The second test started off a lot stronger. It first turboed to ~3.3GHz then jumped to 4GHz, before declining rather quickly to ~3.1GHz. Over a now longer period of about 30 secs, it declined again to 3GHz before dropping below base clock, this time ~2GHz. The lowest speed it reached was ~1.5GHz reached once, and highest was 3.5GHz, also reached once. Unsurprisingly, this returned my best CPU score in Cinebench to date for this machine: 870cb.
By this point, the fans were kicked into high gear by the computer and I ran a third test. This one was very, very interesting. It peaked ~4GHz, then dropped to ~3GHz which is maintained for the shortest stint yet - less than 20 seconds. From there it dropped it's speed in half to 1.5GHz, and stayed below the base clock for a good ~30 seconds, save for one peak to ~2.7GHz. It played jumprope with the base clock speed line for the remainder of the test, ultimately landing a score of 781cb on this pass.
I ran the test multiple times after this, ultimately landing at the conclusion that most likely all 2018 MacBook Pro 15" models are prone to thermal throttle because of the extra cores and Turbo Boost speed provided by Coffee Lake and Apple's lack of a decent cooling design. I also blame Apple's reluctance to let the fans spool up to even half speed until a good minute or more of thermal throttling. While the i7s are no where near as bad as the i9s, thermal throttling does still exist. Other than that, I am pretty happy with the machine. I just hope Apple pushes an update that allows the fans to spin up quicker to assist. It might just make a marginal improvement, but marginal is better than nothing.
I upgraded my MacBook Pro from the base model 2015 15" to the top of the line off-the-shelf 2018 15". This is the 2.6GHz Core i7 paired with the Radeon 560X GPU. My previous MacBook had no dGPU and only had a 2.2GHz Quad Core Haswell i7. One thing that I always found amazing about that machine was that the CPU could almost max out and hold it's turbo boost in clamshell mode while rendering or exporting in Final Cut. I knew this new machine would have more to cool having a dGPU, so I was expecting some possible throttling. For the record, I normally use the machine in clamshell mode hooked up via DisplayPort to a Dell P2715Q 4K display.
As soon as I got the machine home, I noticed that while running a test in Final Cut the machine was getting extremely warm. When I opened Power Gadget, I saw some pretty remarkable throttling. Since then, I've been using Cinebench to test and see just how bad it gets. With the machine in clamshell mode hooked up to my 4K display and doing a cold test in a 78F room, the machine started throttling after about 25-30 seconds. Now, by throttling, I mean Power Gadget was reporting the current speed below the base clock speed. We'll start by seeing a quick spike to <4GHz, then a quick decrease to ~3.3GHz, when it starts steadily declining to ~3.0GHz, where it finally falls off the cliff and drops to about ~1.7GHz. From there I would see occasional peaks up to ~3GHz, followed by a drop to at least 2GHz, with an average drop to ~1.8GHZ, and at it's worst point hitting ~1.3GHz. FWIW, this CPU is supposed to be able to Turbo to 4.3GHz, but the fastest I have seen so far is 4.1.
Since no fan control software yet supporting this machine, I have to 'manually' cause the fans to spin up be running the test multiple times. The first three tests provided varying results: 804cb on the first test, 749cb on the second, and 771cb on the third, which is where the test score started to stabilize. By this time, it was throttling after around 20-25 secs, dropping to ~1.4GHz and jumping back to ~2.5GHz. It hit ~3.3GHz about 3 times in one test, before dropping to ~2.2. After the throttling, it's lowest point was 1.4GHz and highest was 3.4GHz, hitting these speeds 3 times each. Before throttling, it's highest speed was 4GHz which it maintained for 1-2 secs.
Finally, I decided to try the machine by itself, without a display connected. On the first pass, it turboed to ~3.7GHz, and after 5 secs dropped to ~3.3GHz. From there it began a steady decline to 3GHz over the course of about 20 secs, before nosediving to ~2.2GHz. This time it was certainly easier for it to maintain more speed, with the lowest point being ~1.6GHz reached 3 times, and the highest point being ~3.3GHz reached once. The overall graphic mostly showed the CPU running below the base clock speed of 2.6GHz. It scored 849cb in Cinebench.
The second test started off a lot stronger. It first turboed to ~3.3GHz then jumped to 4GHz, before declining rather quickly to ~3.1GHz. Over a now longer period of about 30 secs, it declined again to 3GHz before dropping below base clock, this time ~2GHz. The lowest speed it reached was ~1.5GHz reached once, and highest was 3.5GHz, also reached once. Unsurprisingly, this returned my best CPU score in Cinebench to date for this machine: 870cb.
By this point, the fans were kicked into high gear by the computer and I ran a third test. This one was very, very interesting. It peaked ~4GHz, then dropped to ~3GHz which is maintained for the shortest stint yet - less than 20 seconds. From there it dropped it's speed in half to 1.5GHz, and stayed below the base clock for a good ~30 seconds, save for one peak to ~2.7GHz. It played jumprope with the base clock speed line for the remainder of the test, ultimately landing a score of 781cb on this pass.
I ran the test multiple times after this, ultimately landing at the conclusion that most likely all 2018 MacBook Pro 15" models are prone to thermal throttle because of the extra cores and Turbo Boost speed provided by Coffee Lake and Apple's lack of a decent cooling design. I also blame Apple's reluctance to let the fans spool up to even half speed until a good minute or more of thermal throttling. While the i7s are no where near as bad as the i9s, thermal throttling does still exist. Other than that, I am pretty happy with the machine. I just hope Apple pushes an update that allows the fans to spin up quicker to assist. It might just make a marginal improvement, but marginal is better than nothing.
