I only tested few CPUs. My own observation, the key is to keep the heatsink clean, then simply spin up the fan a bit, keep the core (diode) temperature below 80°C, then Turbo Boost always available.Under *normal* conditions this is true, the big but is the 'no thermal constraint' bit, and that's where it get's a little complicated and nuanced.
I've done a fair bit of testing on this recently as part of my aftermarket fan project, and being lucky enough to test a number of different CPUs including the X5677 and X5690 I can confirm that the X5677 does stay on max boost for longer under under some workloads.
'Thermal constraint' on these CPUs manifests mostly in one of two ways
Single/Low-core count spike
High loads for short time periods where individual cores, or a small subset of them are heavily stressed. They reach max turbo but then if the load is sustained, even for a relatively short period and you watch the individual core temps you'll see the loaded cores spike close to their limit even though other cores and the overall die temp is cool. Once the individual cores get hot then the clock drops away from max.
How the threads are bound can have a big influence here, as if you have two threads locked to the same two cores you'll get a different response to threads starting, completing, and then starting on another core. So the workload plas a big part in this as well, but it is entirely possible to hit thermal issues on a single core.
What I observed is that this happens sooner on the 6 core CPUs that the 4 core at the same clock, presumably a result of more of the die being active and heat contribution from them even when not being specifically stressed.
Multi-core saturation
This is where the CPU is heavily loaded across multiple or all cores continuously. When this happens the CPU will start to drop below max-turbo quite quickly even when overall die temp is kept in check.
As above the individual cores bounce and hover below the thermal limit even when the die itself is not overheating , there's a limit to how quickly you can get the heat out of a processor, but if you're lucky and keep temps in check what you'll see is an all core clock ~3.5GHz or at least bouncing between base clock and 3.6(ish)GHz
This only gets worse as things get heat saturated. You can observe this by watching the individual core temps and the die temp. You'll see cores reach their max temp but still holding a boost, then as the die temp increases the boost drops, and then they all drop to base clock, all the while the individual cores will be sitting at max or just under (as that's the point they throttle). If the CPU die temp continues to climb due to poor or insufficient cooling thermal throttling itself down even further and go lower and lower...
Anyway, I can confirm that the native fan profile may not good enough to keep the Xeon at max Turbo Boost. That's why I mentioned "thermal constrain" in my last post. And only "almost" always available. This is actually for using native fan profile, and for those who never clean the Mac.
The cMP cooling system (hardware) itself definitely good enough, but to keep max Turbos Boost always available, we need to manage the fans by ourselves.
Last edited: