diff --git a/doc/tint2.1 b/doc/tint2.1
index c985bca..3d62190 100644
--- a/doc/tint2.1
+++ b/doc/tint2.1
@@ -831,9 +831,9 @@ execp_markup = 1
 .PP
 .RS
 .nf
-# "stdbuf -oL" noticed to give better performance than fflush() from awk code
+# Though unnecessary here, "stdbuf -oL" performs better than fflush() in awk code
 execp = new
-execp_command = free -b -s1 | stdbuf -oL awk '/^Mem:/ { printf "Mem: %.1fGi %.0f%%\n", $2 / (2^30), 100 * ($2 - $7) / $2 }'
+execp_command = free -b -s1 | stdbuf -oL awk '/^Mem:/ { c="numfmt --to=si"; print $2 |& c; c |& getline a; printf "Mem: %s %.0f%%\n", a, 100 * ($2 - $7) / $2 }'
 execp_interval = 1
 execp_continuous = 1
 .fi
diff --git a/doc/tint2.html b/doc/tint2.html
index 7266ab0..44cd3da 100644
--- a/doc/tint2.html
+++ b/doc/tint2.html
@@ -372,9 +372,9 @@ execp_continuous = 1
 execp_interval = 1
 execp_markup = 1
 </code></pre>
-<h5 id="memory-usage">Memory usage<a name="memory-usage" href="#memory-usage" class="md2man-permalink" title="permalink"></a></h5><pre class="highlight plaintext"><code># "stdbuf -oL" noticed to give better performance than fflush() from awk code
+<h5 id="memory-usage">Memory usage<a name="memory-usage" href="#memory-usage" class="md2man-permalink" title="permalink"></a></h5><pre class="highlight plaintext"><code># Though unnecessary here, "stdbuf -oL" performs better than fflush() in awk code
 execp = new
-execp_command = free -b -s1 | stdbuf -oL awk '/^Mem:/ { printf "Mem: %.1fGi %.0f%%\n", $2 / (2^30), 100 * ($2 - $7) / $2 }'
+execp_command = free -b -s1 | stdbuf -oL awk &lsquo;/^Mem:/ { c=&ldquo;numfmt &ndash;to=si&rdquo;; print $2 |&amp; c; c |&amp; getline a; printf &ldquo;Mem: %s %.0f%%\n&rdquo;, a, 100 * ($2 - $7) / $2 }&rsquo;
 execp_interval = 1
 execp_continuous = 1
 </code></pre>
diff --git a/doc/tint2.md b/doc/tint2.md
index 5b6b1b9..d1a27b1 100644
--- a/doc/tint2.md
+++ b/doc/tint2.md
@@ -694,9 +694,9 @@ execp_markup = 1
 ##### Memory usage
 
 ```
-# "stdbuf -oL" noticed to give better performance than fflush() from awk code
+# Though unnecessary here, "stdbuf -oL" performs better than fflush() in awk code
 execp = new
-execp_command = free -b -s1 | stdbuf -oL awk '/^Mem:/ { printf "Mem: %.1fGi %.0f%%\n", $2 / (2^30), 100 * ($2 - $7) / $2 }'
+execp_command = free -b -s1 | stdbuf -oL awk '/^Mem:/ { c="numfmt --to=si"; print $2 |& c; c |& getline a; printf "Mem: %s %.0f%%\n", a, 100 * ($2 - $7) / $2 }'
 execp_interval = 1
 execp_continuous = 1
 ```