Late 2013 rMBP never reaching "peak turbo"

[cryptocat]

I recently bought a used late 2013 15" rMBP, been really happy with this machine it so far, especially since coming from an early 2008 MBP which had served me well but has become unbearably powerless.


What's odd (to me) is its turbo boost behaviour:

This is a CTO unit with the fastest CPU available at the time: i7 4960HQ. 2.6GHz, boosts up to 3.8GHz for single core loads according to Intel specs.

When I run Cinebench (in SC mode), it never quite reaches 3.8, only 3.6-3.7 very briefly and then drops to 3.4GHz where it stays for the entire run. Power draw is only around 20W and CPU temp is far from thermal throttling. (Did a repaste and let the fans spin at max rpm to make sure it's not thermal throttling that's holding it back. See attatched screenshot.)

Now, is this "intended" behaviour to never sustain peak turbo, even though the power draw is far from TDP and temps are low enough?

I know there's much ado lately about the new six-core MBPs not being able to boost due to thermal restraints, but in this scenario of low temps and less-than-TDP power-draw?

My Gaming PC also has a "Haswell" CPU, which peaks up to 3.7GHz in SC, but stays there forever. In theory, this 4960HQ should be faster in SC, given it's 100MHz clock speed advantage. Cinebench R15 says otherwise: 136 vs. 153 Points in SC. Granted, it's Win10 vs. OS X, but still.

Thanks.

 

[leman]

IMO, the clocks you are observing are a bit low, but at the same time it can be difficult to maintain optimal conditions for maximal turbo boost. The 3.4Ghz you are observing is incidentally the max turbo for that CPU when all 4 cores are active. You might be running Cinebench in single-core mode, but it's very possible that the OS is running background processes on other CPU cores, which might limit the max achievable turbo. Of course, it's all pure speculation on my side. You can test it by disabling all but one core (either using Xcode instruments utility or setting boot-args="cpus=1").

As to why your desktop CPU behaves differently, my guess would be a) different core management on Mac and Windows (if I remember correctly, Mac tries to split tasks across cores more evenly) and b) much higher thermal headroom of the desktop CPU.

 

[cryptocat]

I understand that the desktop CPU has much better cooling headroom, that's what desktops are for, after all.

Just tried boot-args="cpus=1", to no avail. Still only boosts to 3.6 for a couple seconds, then remains at 3.4GHz.

What else could affect turbo behaviour?

When temps are low enough, with a PSU connected still this thing won't turbo higher.

Anyone with a 2.3GHz late 2013 here willing to run Cinebench for me?