Local AI and ML
Long inference sessions, repeated prompts, and other sustained GPU-heavy work where accumulated heat can reduce performance over time.
MacTC is built for Mac users who would rather trade silence for performance during sustained local AI, build, creative, and gaming workloads. It moves fan response earlier than quiet-first defaults when accumulated heat starts to threaten speed.
The GUI is free forever. Purchase unlocks background fan control after trial.
Downloads are temporarily unavailable. Purchase starts in MacTC and is completed through Paddle’s secure checkout in your browser.
Your Mac is tuned to stay calm and quiet through normal work. Under longer heavy loads, that quiet-first choice can let heat accumulate before the fans get assertive. MacTC changes that tradeoff: it asks for more fan earlier when sustained heat threatens performance, then backs off when the extra cooling is no longer earning its noise.
System mode remains available as the baseline, so the quieter default stays one click away.
Current MacTC builds are arm64 and require macOS 14 or later. Active fan control has been validated on the two-fan M4 MacBook Pro and one-fan M2 Mac mini configurations.
On a confirmed fanless Apple Silicon Mac, MacTC stays in monitoring-only mode. Monitoring remains available on other Apple Silicon fan configurations, but active fan control is not currently supported until that configuration is validated. Future macOS or firmware updates may require fresh validation.
MacTC is for power users who run long local workloads and would rather hear the fans earlier than watch those jobs slow down.
Long inference sessions, repeated prompts, and other sustained GPU-heavy work where accumulated heat can reduce performance over time.
Repeated local builds, test suites, indexing, and compile loops where small heat-related slowdowns can add up through the day.
Video exports, renders, encodes, and media jobs that stay hot long enough for fan behavior to affect completion time.
MacTC shows live temperatures, fan speeds, power, and the fan target it would use while you stay in System mode. During the seven-day free trial, MacTC mode can actually control the fans so you can compare it under real work.
The paid layer is MacTC fan control that keeps running in the menu bar, watches sustained heat, and changes fan behavior when you choose more proactive cooling.
The seven-day free trial exists so you can test that policy on your own machine, under your own workloads, before buying.
MacTC changes fan timing under sustained load. It is not an all-fans-all-the-time mode, and it does not promise better results every time.
MacTC uses more fan activity to pursue lower temperatures when performance matters more than silence.
System mode stays available, and the dotted fan line shows what MacTC would ask for before paid control enforces it.
Thermal results vary by machine, workload, ambient conditions, firmware, sensors, and current system state.
In rare or unforeseen cases, MacTC may be less effective than intended or may worsen thermal behavior.
I realized I was sometimes missing out on the Mac performance I had purchased. Learn more about the path to MacTC's creation here.
I've been using a number of LLMs on my M4 Max MacBook Pro using
the amazing
LM Studio. I
noticed some sustained workloads seemed to get slower and
slower. This is easy to detect in an LLM harness like LM Studio
where it can show you the tokens per second (t/s) rate after
every response. When I ran a bunch of queries back to back on
some models, the fans would spin up as expected, but I noticed
my t/s falling off substantially. Using the outstanding
mactop
tool (brew install mactop), I immediately saw that
GPU performance was starting at 100%, and then progressively
dropping several minutes into long-running jobs. While I had
paid an extra $300 to jump from the available 32-core GPU to 40
cores, under a few minutes of heavy load, I was sometimes left
with a 24-core GPU's worth of performance.
I experimented with some of the available Mac apps for adjusting the fans. What I found was that for my GPU-heavy LLM workloads, the set-it-and-forget-it fan options were sometimes noisy, and still would not prevent all thermal throttling in cases where blasting the fans at 100% did help. Nobody wants a noisy machine, and Apple did a fantastic job with their tradeoffs, to a point. But I wanted the performance I bought, and if achieving that was sometimes louder, that was the tradeoff I was willing to make.
After a lengthy investigation of how my system handled power, heat, performance, and fans, I found that the fixed-fan-percent-by-temperature model is simply unsuited for some heavy loads lasting more than a few minutes. The system has an inherent notion of the thermal soak properties of the SoC. In practice this means my Mac was quite lazy about ramping up the fans initially even when the CPU and GPU shot past 90°C, and it did not throttle speeds early. But if that heat load continued for a couple minutes, the fans would happily sit at a middling speed while reducing performance.
That is the problem I set out to solve for at least some workloads with MacTC. It defers to the system's fan choices by default, but when it observes power and heat rising, it should try to front-run accumulated heat with more fan earlier, while still being as quiet as possible. And when thermal throttling begins, it will take the fans higher only for as long as it takes to end the throttling, and then attempt to fall off to the lowest fan speed that can sustain full performance.
When MacTC does increase fan speed, it deliberately takes a few seconds. Ideally, the end result at your desk is the difference between noticing "oh, my fans are on," and "wow, my fans just came on." That small whoosh turns out to be really aggravating all by itself. As it came together, I immediately understood this was an app worth sharing with others. I hope it can benefit you as much as it benefitted me during testing.