I'm running newer trochoidal tool paths. I have solidcam / imachining. I overrun my buffer pretty quick when using it. running from memory vs. dnc makes a tiny bit of difference, but really running from memory isn't more usable than dnc in a practical sense - i still overrun my buffer. I'm only able to run at about 3/8 on the imachining settings without overrun.

another thing is that when running those kind of tool paths you get a lot of helical ramp too short errors. I've dealt with that by modding my post quite a lot to where I slow down the feed rate for moves with tight curvature. that works ok but you still come across the odd helical ramp error.

i also have an AFF option board. AFF only runs good with G1 moves. it goes all f*cked up if you program an arc. I typically run AFF with small G1 moves. I set the G1 move res to like 0.004", the aff is set to like 0.01", and I always leave 12 thou on the walls for a finish cut. this kind of setup seems to maximize the feeds that i can run without overloading the buffer.

I'm interested in knowing if you get more blocks/sec out of a -5 card setup. right now i have -4 cards. my machine is DC servos, 4 axis, and i put 101.3 software on it myself so i could run probe stuff over DNC


why else bother with -5 cards?