If I am not mistaken the card reader got an upgrade as well. Should be faster now, but I don’t know by how much. Also USB 3.0 that is attached to the x1 PCIE 2.0 should give up to 500MB/s. Not sure how much that usb 3.0 controller chip can do though.
[Edit]
363 /323 MB/s R/W usb ssd. 45/27 MB/s card reader. (theoretical max 100/62 MB/s)
The Raspberry pi site is running on new pi4s as well
The issue isn’t MMC read speed, it’s that MMC isn’t designed for primary storage use and tends to die. SSDs also use NAND but they run a whole bunch of them striped and have all sorts of reallocation and wear-leveling algorithms alongside extra slack space to increase longevity. MMC doesn’t have any of that, it’s just straight up NAND.
Yea, fair enough. Although some filesystems can be used to reduce the wear and tear on MMC. Not sure why Raspian uses ext4 by default. Maybe its for simplicity sake?
I’m mostly going to take a wait and see in hopes it’ll be possible to use one USB 3.0 port to boot the OS off a SSD, if the 4GB model ends up being the more “popular” version for development then it’ll be interesting. Cost point of view, I think the 2GB version will be more commonly used as some creators will usually want to stay within a specific project budget.
As far as my own usage of Raspberry Pi, a friend across the Atlantic more or less talked me into using such hardware. When it comes to shared work, it feels weird having a Raspberry Pi surrounded by nVidia hardware. (Beaglebone & Parallella were the two I preferred at the time as the hardware was much better–Parallella you could use the 16-core Epiphany chip for compute)
A good suggestion is figure out what kind of actual usage you need or want headroom, Jetson Nano can boot the OS via USB 3.0 which opens the door for faster SSD storage instead of the MicroSDXC slot. Hardware wise nVidia holds the lead if GPU/compute performance is necessary within a a low-power environment.
I don’t see why not. You would probably still have to use sd-card to load all the core components, but the rootfs can be on the external disk.
I’ve been using OpenSuSE with transactional-server role in my router for a while. It is a PCEngines APU2 that boots and runs from a 16GB SD card. This particular system places all the rootfile system on read only mode, enables a unionfs on etc (which is committed upon reboot). Other partitions are mounted read write but I’ve modified it to mount /var/log and /temp as tmpfs.
BTRFS is essential here because AFAIK updates are made by creating a snapshot, chrooting into it, installing the packages and marking the new snapshot as the main volume. The downside is that you have to reboot to commit any changes. I’ve also added some mount args to btrfs to minimize wear. Its been working like this for more than a year.
There is a variant called MicroOS which is a transactional-server system ready to deploy docker or Kubernetes. You can install other stuff if you want.
Thats the crutch of it, there was a tutorial that went off without a hitch to do a LAMP build. A year or two later when I needed to rebuild it, the ported Debian tutorial no longer worked as the kernel and repos were now different, and x86 version tutorials are not in paraity.
I’m not a grey beard, I need to leverage outside resources to get stuff done and its annoying if a Pi tutorial does not exists, and repos, commands etc do not match with the equivalent x86 build that has tons of documentation. Great if you want to go down the rabbit hole that leads to grey-beardom, but bad if you are trying to dabble in the kid section of the pool before going full on rescue swimmer.
It turned into such a head-ache. Yeah the hardware is reliable, but I didn’t feel like I needed an honorary CS degree to maintain the build as time went on. The joy of using x86 Ubuntu or Debian is the ocean (ha! especially digital ocean) of documentation and help out there.
From what I’ve read, booting off USB and network boot are not yet possible but are in the works.
A few hoops to jump through but nothing heroic… enable USB boot for RPI 3 and forward
https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/msd.md
edit, also NET boot:
https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/net_tutorial.md
It says " Raspberry Pi 3B, 3B+, 3A+ and Raspberry Pi 2B v1.2 models only. ". RPI 4, from what I understand, has a different bootloader. Since this makes changes to some one time programmable memory, I would not recommend trying to follow those instructions on RPI 4.
Well poop, I was all excited. Might give it a whack on my stack of Pi 2s.
Thanks for pointing out the important gotcha.
It can still be done. You would just be required to have sdcard plugged in with bootloader on it. But the filesystem itself can be on the external hard drive. I think the only catch is that you have to make sure the USB stuff is loaded into memory before it tries to load files from harddrive.
Actually, if there is anything I should be looking for, it’s a system that can run NAS storage.
yes with beefier graphics and an octocore
I mean, I’m just happy to see 4GB as an option.
I’d also like them to hop on the EMMC bandwagon. So much better than SD.
It looks interesting. Especially the 4GB version.
Although everything depends on the needs and what we want to apply. As a small NAS I would rather still choose Odroid HC-2 but for other needs …
My Pi 3 does a great job with my home Python web server.
Rather have nvme… even if it ups the price.
