diff --git a/Machines/Apple/Macintosh/Audio.cpp b/Machines/Apple/Macintosh/Audio.cpp
index 3dedb6d0a..3946553fc 100644
--- a/Machines/Apple/Macintosh/Audio.cpp
+++ b/Machines/Apple/Macintosh/Audio.cpp
@@ -12,7 +12,8 @@ using namespace Apple::Macintosh;
 
 namespace {
 
-// The sample_length is coupled with the clock rate selected within the Macintosh proper.
+// The sample_length is coupled with the clock rate selected within the Macintosh proper;
+// as per the header-declaration a divide-by-two clock is expected to arrive here.
 const std::size_t sample_length = 352 / 2;
 
 }
@@ -36,6 +37,7 @@ void Audio::set_volume(int volume) {
 	// Post the volume change as a deferred event.
 	task_queue_.defer([=] () {
 		volume_ = volume;
+		set_volume_multiplier();
 	});
 }
 
@@ -47,6 +49,7 @@ void Audio::set_enabled(bool on) {
 	// Post the enabled mask change as a deferred event.
 	task_queue_.defer([=] () {
 		enabled_mask_ = int(on);
+		set_volume_multiplier();
 	});
 }
 
@@ -58,7 +61,12 @@ bool Audio::is_zero_level() {
 
 void Audio::set_sample_volume_range(std::int16_t range) {
 	// Some underflow here doesn't really matter.
-	volume_multiplier_ = range / (7 * 255);
+	output_volume_ = range / (7 * 255);
+	set_volume_multiplier();
+}
+
+void Audio::set_volume_multiplier() {
+	volume_multiplier_ = int16_t(output_volume_ * volume_ * enabled_mask_);
 }
 
 void Audio::get_samples(std::size_t number_of_samples, int16_t *target) {
@@ -66,14 +74,24 @@ void Audio::get_samples(std::size_t number_of_samples, int16_t *target) {
 	// in fact it uses pulse-width modulation. But the scale for pulses isn't specified, so
 	// that's something to return to.
 
-	// TODO: temporary implementation. Very inefficient. Replace.
-	for(std::size_t sample = 0; sample < number_of_samples; ++sample) {
-		target[sample] = volume_multiplier_ * int16_t(sample_queue_.buffer[sample_queue_.read_pointer] * volume_ * enabled_mask_);
-		++subcycle_offset_;
+	while(number_of_samples) {
+		// Determine how many output samples will be at the same level.
+		const auto cycles_left_in_sample = std::min(number_of_samples, sample_length - subcycle_offset_);
 
-		if(subcycle_offset_ == sample_length) {
-			subcycle_offset_ = 0;
-			sample_queue_.read_pointer = (sample_queue_.read_pointer + 1) % sample_queue_.buffer.size();
+		// Determine the output level, and output that many samples.
+		// (Hoping that the copiler substitutes an effective memset16-type operation here).
+		const int16_t output_level = volume_multiplier_ * (int16_t(sample_queue_.buffer[sample_queue_.read_pointer]) - 128);
+		for(size_t c = 0; c < cycles_left_in_sample; ++c) {
+			target[c] = output_level;
 		}
+		target += cycles_left_in_sample;
+
+		// Advance the sample pointer.
+		subcycle_offset_ += cycles_left_in_sample;
+		sample_queue_.read_pointer = (sample_queue_.read_pointer + (subcycle_offset_ / sample_length)) % sample_queue_.buffer.size();
+		subcycle_offset_ %= sample_length;
+
+		// Decreate the number of samples left to write.
+		number_of_samples -= cycles_left_in_sample;
 	}
 }
diff --git a/Machines/Apple/Macintosh/Audio.hpp b/Machines/Apple/Macintosh/Audio.hpp
index c1348cbf1..cfec007d6 100644
--- a/Machines/Apple/Macintosh/Audio.hpp
+++ b/Machines/Apple/Macintosh/Audio.hpp
@@ -75,9 +75,11 @@ class Audio: public ::Outputs::Speaker::SampleSource {
 		// thread only.
 		int volume_ = 0;
 		int enabled_mask_ = 0;
+		std::int16_t output_volume_ = 0;
 
 		std::int16_t volume_multiplier_ = 0;
 		std::size_t subcycle_offset_ = 0;
+		void set_volume_multiplier();
 };
 
 }