Octoprint keeps on impressing  

RE: Octoprint keeps on impressing

I think I may give it a go, I've just ordered an RPI4 with heatsink and HyperPixel 4 screen (With additional standoffs due to the heatsink covering both sides of the RPi).

Let the fun begin 😀  Start of with loading Diet-Pi then modify code to boot from usb micro usb storage drive (Sandisk Cruiser 8Gb - as this has really good 4k block RW performance, which is what you need for an OS, not the single large  fast performance figure which you only get with most drives. Don't look at the top rated speed as you will find 4k performance is abysmal. Those speeds only work for one file transfer, not the finding and getting the file ready to transfer times)

RE: Octoprint keeps on impressing

I've also used the Hyperpixel 4.0 - excellent choice. Should you encounter a miscalibration of the touch area this is fixable (contrary to the info on pimoroni website that states screen needs replacement..."

RE: Octoprint keeps on impressing

Posted by: @chocki

I think I may give it a go, I've just ordered an RPI4 with heatsink and HyperPixel 4 screen (With additional standoffs due to the heatsink covering both sides of the RPi).

Let the fun begin 😀  Start of with loading Diet-Pi then modify code to boot from usb micro usb storage drive (Sandisk Cruiser 8Gb - as this has really good 4k block RW performance, which is what you need for an OS, not the single large  fast performance figure which you only get with most drives. Don't look at the top rated speed as you will find 4k performance is abysmal. Those speeds only work for one file transfer, not the finding and getting the file ready to transfer times)

An example ... Samsung EVO Select 128GB uSD ... 

RE: Octoprint keeps on impressing

This is for a Samsung Fit Plus 64Gb.

But for best performance an SSD would be needed, so if you happen to have one lying around, maybe fit it into a SSD to USB adapter and run the RPI from this, but currently it still requires booting from a microSD card, but then directed to the USB memory device, native USB booting is not supported on the RPI4 - Yet!.

RE: Octoprint keeps on impressing

Other than for a NAS or desktop replacement, I don't find SD speeds overly confining on my RPis. Once everything is loaded up on a more-or-less static server, most operations run in RAM or transfers in which the network is the constraint. I just buy whatever Samsung Evo card is $15 at the time.

RE: Octoprint keeps on impressing

So Octoprint does not really benefit from a faster storage medium?, I'm just used to using my RPI's for uses that have a fair amount of disk usage and the USB drive really improved things.

But then again, I've never had an RPi4 yet.

RE: Octoprint keeps on impressing

Lets not get storage media confused here.

The SD card that is speed constrained on PC -> Einsy transfers is the one in the Einsy display unit. This delay is not the card so much as PC -> WiFi -> Pi -> USB Einsy -> Einsy storage. More the path and buffering than interface speed.

The SD (micro) card that the Pi boots off of and uses for storage is on the Raspberry Pi. (what Octoprint calls local). Faster media or flash card storage helps here if needed. It controls how fast the Pi can swap in your GCode for transmission to the Einsy.

PC -> Pi transfers, as from PrusaSlicer, are really fast. Couple of seconds for a 12M GCode file.

Drag a file to OctoPrint from PC storage and select Upload to Printer SD card - we are talking minutes for the same 12M file.

RE: Octoprint keeps on impressing

Posted by: @chocki

So Octoprint does not really benefit from a faster storage medium?, I'm just used to using my RPI's for uses that have a fair amount of disk usage and the USB drive really improved things.

Well, a RPi can certainly suffer from a microSD that is too slow. I finally gave up on using old Class 4 or 6 microSD cards for any RPi-based projects. You want a class 10 card or better, but most of the $15 SanDisk or Samsung cards perform very well. Once it boots up, it's not going to disk excessively. 

The only problems I've encountered or read about relate to the CPU getting busy on smaller RPis and having timing issues using USB to the printer -- presumably related to lack of serial flow control on the 8 bit controller.

I've monitored my RPi during prints and occasionally see one core spike briefly, but that's it. Memory utilization is typically < 300M, so I haven't seen it hit swap.

But then again, I've never had an RPi4 yet.

I just realized I've got mine running on a RPi 2B and it's done quite well streaming a Logitech C270 and RPi Zero running MotionEyeOS. I've got a couple of 3B, 3A+ and 4B to swap in (live near a Microcenter), so will look at the boards a bit more closely and actually put a 3B in this time. The 4 is definitely overkill, so I think I'll try a 3A+ first just for giggles. The 512M RAM should be just enough. I'm curious if the dual-band wifi will make much of a difference for streaming.

RE: Octoprint keeps on impressing

I'm in the playing stages with an Rpi4, and have a couple Rpi0W on the way for pet projects ... probably a waste of money now that I am doing research and finding they are almost less than adequate for Octoprint ... sigh. 

On m Rpi4, I tried the stock SanDisk SD and wasn't impressed with disk i/o, so tried the EVO Select. Numbers above were through two USB 3.0 hubs, so I may need to re-do that test. 

This Samsung disk isn't via USB ... but shows better performance ... but I have an earlier gen version that is USB mounted.  USB 3 can't achieve the peak speeds, but should allow the 4k numbers.

RE: Octoprint keeps on impressing

Interesting.  The Sandisk SD that Canakit shipped: 

The USB mounted Samsung 960 Pro M.2 (the faster numbers above are 970 EVO M.2):

RE: Octoprint keeps on impressing

Posted by: @tim-m30

I'm in the playing stages with an Rpi4, and have a couple Rpi0W on the way for pet projects ... probably a waste of money now that I am doing research and finding they are almost less than adequate for Octoprint ... sigh. 

The RPi Zero W is plenty useful for projects. It's just not well suited for running Octoprint. I use them for all sorts of things. I'm going to use one in my printer cabinet for an overview display (full bed, filament, etc.) running MotionEyeOS configured as a dedicated streaming webcam. High frame rates aren't required, so it's more than suitable.

On m Rpi4, I tried the stock SanDisk SD and wasn't impressed with disk i/o, so tried the EVO Select. Numbers above were through two USB 3.0 hubs, so I may need to re-do that test. 

I don't know that a faster microSD card will benefit Octoprint much. The RPi definitely suffers when booting with old slow cards. Linux tends to be pretty efficient with caching, and the USB transfer rate to the printer will be much slower, so I don't worry too much going with the Samsung Evo cards. It's a SBC, so I'm not expecting amazing throughput numbers.

RE: Octoprint keeps on impressing

Since the RPi Zero W has been seen as inadequate for running OctoPrint has anyone tried adding heatsinks or watercooling and overclocking it? 

RE: Octoprint keeps on impressing

Heatsinks for overclocking yes, water-cooling no. The performance gain is negligible.. Then you have an overclocked underpowered PiZero for the task. IMHO not worth the effort or the penny for doing it to run octopi.

RE: Octoprint keeps on impressing

Posted by: @jadedreddragon

Since the RPi Zero W has been seen as inadequate for running OctoPrint has anyone tried adding heatsinks or watercooling and overclocking it? 

You can be the first. Let us know how it goes.

Heatsinks alone do nothing unless the processor is heat throttling.

Is it even possible to change the clocks on the Zero? Faster clocks will not compensate for inadequate architecture.

That is like saying use a faster clock on a 3B - and you have a Pi 4. Not gonna happen.

RE: Octoprint keeps on impressing

Jup like that:

edit /boot/config.txt:

# 10% overclock

arm_freq = 1100

over_voltage = 8

sdram_freq = 500

sdram_over_voltage = 2

force_turbo = 1

boot_delay = 1

but as I stated - that will not turn the crawler into a runner.

Disclaimer: Overclocking might damage your hardware - the provided examples are just that and not a recommendation of values for YOUR device. Proper cooling of the components is solely the responsibility of the user.

RE: Octoprint keeps on impressing

Posted by: @christoph-e5

Heatsinks for overclocking yes, water-cooling no. The performance gain is negligible.. Then you have an overclocked underpowered PiZero for the task. IMHO not worth the effort or the penny for doing it to run octopi.

I have to admit that the idea of a water-cooled RPi Zero W is appealing. The idea evokes thoughts of a Fiat 500 Abarth angrily coming up to speed as it enters a highway. 

I haven't been following the RPi scene closely the last 2 years, but I recall that the newer RPis couldn't really be overclocked in any meaningful way, that the previous overclocking was now the standard setting. Regardless, I wouldn't expect much out of a single-core Zero, overclocked or not.

I'm looking at some of the other (non-Raspberry) boards in the same form factor offering quad core and more memory. For all that, it's a lot easier to just stick a RPi 3B behind the printer and be done with it.

RE: Octoprint keeps on impressing

More like a Volkswagen Bug coming up to speed on Nitro... Fun while it lasts.

And the 3B works so well...

RE: Octoprint keeps on impressing

RE: Octoprint keeps on impressing

My complete overkill rpi4 is coming along nicely, just got to figure out how to get a working version of octoprint installed onto Rasbian Buster all running from a 120GB Crucial 2.5" SSD. I know, complete overkill, no need for it, but hey, better to have and not need than need and not have.

Rasbian Desktop in 18 seconds from power on, on and all displaying on the rather nice Hyperpixel4 display.

How the heck do you rotate images in this forum!!!.

RE: Octoprint keeps on impressing

The additional horsepower under the hood, I plan on having a little tinker with a Movidius Neural Compute Stick 2 I have, and Logitech web cam to see if it is possible to detect failed printing. I have an idea to train a neural network to recognise a clean print head esp. around the nozzle area, and if it starts to get significant buildup of plastic, terminate the printing. This should be actually fairly straightforward. I'm not to bothered by the print maybe being a failure, but what I don't want is a blob of doom. If I can detect a spaghetti monster, then this would be a bonus, but I would need a lot of examples to train the model, whereas I already have a clean extruder and only need to create some G-code to move the head from left to right, up a bit right to left up a bit left to right etc, whilst taking snapshots with the web cam and put these through a trainer as good images, then just stick things (bluetack) to the nozzle and run the motion capture again with the fake build up of filament to train failures. It may even be better to have the camera fixed with the X axis, even less training required. AI uses black and white anyway, so this should work with all filaments if I can get it to work. The RPi4 may even be fast enough to not require a Neural Compute Stick, but I will have to test this.

The training would have to be done on my normal Windows machine as it has significantly more processing power than the RPi, The RPi4 would be used for infering the model against the video feed, and a % is returned against the model whether it is a good match or not, and at a certain threshold, code can be executed to say stop the print.

It's all hypothetical at the moment, but feel free to chip in or completely blow my musings out of the water so to speak. 😀