sure, you have 4 pins to send data into the computer, but no way of synchronizing transfers

Obviously Logitech and Microsoft managed to solve any syncing issues to make their protocols work.

you cant just multiply 4 bits by ISA bus frequency and assume that to be the max

>For theoretical max it should come down to max port polling rate of the four button pins.

you need to waste cycles/bits on synchronization

(Shakes head) Port polling rate doesn't refer to ISA bus frequency, it's how many in/outs you can do.

no busy/wait cycle mechanisms on ISA gameports, you can hammer it as fast as your cpu allows

Right, that's what I'm thinking, as fast as the port can be polled in the confines of the isa/pci bus effective bandwidth.

P.S. Microsoft's sidewinder is slower than Logitech's ADI protocol, using only 3 bits for transmitting data. http://www.epanorama.net/documents/joystick/p … html#wheelpedal

On the plus side there's an AVR project for interfacing sidewinders. https://hackaday.com/2019/01/29/interfacing-t … ystick-to-avrs/

lpt, com & other joystick interfacing(lots of dead links): http://www.epanorama.net/documents/joystick/pc_misc.html

Effective bandwidth of ISA @ 8bit 4.77mhz being only 1mb after delays: https://www.evaluationengineering.com/home/ar … quisition-needs

By this account, polling a port takes a us: http://www.faqs.org/docs/Linux-mini/IO-Port-Programming.html, would make the gameport's theoretical max on the original pc/xt 488.28kb/s. I'm finding effective rate of PCI being 62-100mb/s depending, so roughly 7.5-12mb/s should be theoretical max for a gameport on a 33mhz PCI sound card.

Also note that the game port needs some time to "recharge" and that older game ports were speed-sensitive.
Some had variable resistors to correct that little issue.

Thats the analog part (558), afair speed sensitivity was due to stupid 16 bit nop busy loops in game code breaking on faster cpus.

558 measures the potentiometers, the buttons are directly connected to the ISA bus: http://www.epanorama.net/documents/joystick/pc_joystick.html so the estimated theoretical max holds. Logitech and Microsoft use the analog input to signify the start of a tx, so if mirroring how they operate, a following tx start could be limited to how soon it could follow the preceding one, but the tx itself could be the max button rate. If Logitech/Microsoft's 10us rate were taken as the 558 recharge delay, that'd be 40bits/5bytes, how much could be tx on the button pins in 10us. Even if you were starting a new transfer as fast as it could be done consecutively, every 10us, as long as you were transferring 40+ bits per tx there would be no loss in transfer rate from a recharge.