| With the recovery time of the tube-counter combination of the GMC 300E+ series counters established as around 200Ás (http://www.gqelectronicsllc.com/forum/topic.asp?TOPIC_ID=4598 ) the max pulse rate of the counter mainly due to the tube is thus limited to something near CPS=5000.
However, the counter's electronic and firmware has additional limits. With a 2 byte storage reserved for CPM, its maximum is 2^16-1 = 65535 (equiv to CPS=1092). Although CPS is also stored in 2 bytes, due to the upper 2 bits being unused, its numeric limit is 2^14-1 = 16383 (equiv to CPM=982980). But its true limit is far less!
I removed the tube of a GMC 300E+ and powered the device on. Now the anode clip has high voltage (!), but at the cathode clip the DC voltage is zero (DMV: +0.1mV against ground) and there is no AC voltage beyond some noise in the low mV. So it is safe to feed a pulse from a signal generator to the counter at the cathode clip, and let the counter count. My signal was a DC coupled 0 ...2V rectangular pulse of 10% duty cycle. See picture, blue is signal at cathode, yellow at audio out.
I then set signal frequencies from 100Hz to 2kHz, and logged "count rates" with GeigerLog in CPM and CPS mode, resp.. Results shown in next picture.
In CPM mode the expected numeric overflow occurred at 65535 after switching to 2kHz and instead of CPM=120000 only 120000-65353 => CPM=54647 was seen.
In CPS mode at 2kHz the correct CPS=2000 was seen, displayed as CPM=120000.
The next step up of the pulse generator is 5kHz, and when this was set - irrespective of CPS or CPM mode - the counter would crash and not show anything (see far right edge in picture). The counter could only be set back to normal by cycling power.
So, somewhere between CPS=2000 and 5000 the counter reaches its technical limits, perhaps due to the limiting computing power of the microcontroller. This limit may coincide or come even earlier than the tube limit.
Basically, the counter is technically limited to CPM~60000 and CPS~2000.
While the data as shown were obtained by logging with GeigerLog, the same results are found by downloading the history.