duet4/5... -> high-speed sd-socket

A Former User

Hi,
would in the future a higher speed working duet4 or 5 maybe benefit from a high-speed sd-socket connection so that machines like a laser (which can work really super fast) or e.g. also the "simulation-time" for printing/milling/lasering/penceling/whatever be improved or is the bottleneck e.g. for simulation the algorithms on the cpu and or the clockspeed?
Just wondering... (this might be more an educational thing then a real wishlist thing)

Veti

@LB

duet3 with rp4 will offer high speed.

dc42

All Duets already use a high speed SD card interface. The factor that limits read and write speeds is the amount of memory available for use as buffer RAM. SD cards only maintain high sustained transfer speeds when reading/writing large blocks of data, as happens when they are used in cameras.

M122 shows the SD card transfer speed in Mbytes/sec.

I have done some work in RRF 3.2beta3 which I hope will make simulations faster, and I can probably do more to speed them up too.

A Former User

@dc42
Thanks so much - always learning
Interesting! Looking forward for 3.2 alpha! Fast enough for me now anyway but of course prints will get bigger and so over the next time I hope (original thought was for a future generation of hardware - didn´t know that you were already on to it via RRF and more...)
(for folks reading so far you might just like to check out the other side e.g. this https://www.youtube.com/watch?v=5nszCrXb-7I for sd-cards (performance & comparison) what helped me)

dc42

The RPi has far more RAM than a Duet so it can use very large buffers. Also it has a disk cache, so not all reads/writes involve a transfer to the SD card (writes will eventually go to the SD card, but depending on the test they may not complete until after the test has finished).

This is why uploading GCode files over Ethernet to a Pi+DSF+Duet is faster than uploading them directly to the Duet.

On the Duet 3 series we have some scope for improving the upload speed, due to the extra RAM compared to Duet 2. But I have more important things to work on before I address this.