The new Raspberry Pi bootloader is out which makes these instructions only necessary if you want to continue to use the SD card as a bootloader. Check out the new Raspberry Pi Bootloader Configuration Guide here to enable native USB booting support!
The Raspberry Pi 4* is finally here and has a lot of exciting changes. One very major downside is that it doesn’t support true USB booting yet out of the box (like the 3 series did). The Raspberry Pi foundation states that it is being worked on and will be added back with a future update. No timeline has been given yet for that to happen but they state it’s one of their top priorities.
Most of my projects heavily depend on having good performing storage so sitting and waiting was not an acceptable solution. In this guide I’ll show you a workaround to use USB devices as your rootfs device and use a Micro SD card as bootloader only which gives us full SSD performance after boot! To see exactly how much of a performance difference this makes (spoiler: it’s gigantic) check out the Raspberry Pi Storage Benchmarks.
I highly recommend doing this on a completely new install. If you try to upgrade your old ones and something goes wrong there’s a good chance you might lose data. We will be modifying the boot partition, resizing partitions, etc. so don’t use a drive with any data on it unless you are positive you have all of the steps down!
Compatible USB Adapters
The Raspberry Pi 4 is proving to be picky about what SATA, M.2, etc. adapters will work in the USB 3.0 port. The USB 3.0 ports are the ones in the middle that are blue inside. The black ones are USB 2.0 and won’t give you the faster speeds the new Pi offers.
It’s very likely that some of these will be fixed via software and firmware updates and the Raspberry Pi Foundation has several open known issues related to USB 3. Until that happens though I will maintain a list here of known working ones and known problematic ones. It’s still very early in the release of the Pi 4 so we still have a lot to learn about which adapters work / don’t work. If you have working and nonworking adapters leave a comment and I’ll add it in this list.
If the adapters worked before on older Pis then one thing you can try is putting them in the black USB 2.0 ports. Obviously this is stupid because we all want the Pi 4 performance gains but if you end up needing to buy a new adapter this will give you a workaround until a replacement arrives!
Find USB adapter chipset
There are certain chipsets used in adapters that are known to be working/not working.
pi@raspberrypi:~ $ lsusb Bus 003 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub Bus 002 Device 002: ID 174c:55aa ASMedia Technology Inc. Name: ASM1051E SATA 6Gb/s bridge, ASM1053E SATA 6Gb/s bridge, ASM1153 SATA 3Gb/s bridge, ASM1153E SATA 6Gb/s bridge Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub Bus 001 Device 002: ID 2109:3431 VIA Labs, Inc. Hub Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
This is a lsusb dump of all my connected USB devices. I have bolded the line with the USB bridge device. We can see that the chipset is ASM1153E. This is a really common one that works well with the Pi.
The new USB-C 3.1 Type A version of the Sabrent adapter is reported as working in the comments by UEF. DO NOT get the USB 3.0 version as that one is below on the naughty list and won’t work!
It’s bold. It’s beautiful. It’s also confirmed working by TADRACKET and Steve B. However, be warned, it takes a *lot* of power! Steve B. reports that even with the oversized 3.5A CanaKit adapter* it does not work. If you have the standard 3.0 adapter you can be practically certain it won’t power this enclosure. Does work with a powered USB hub*.
Returning legendary commentary Frank Meyer reports: Does not work with a 3.0A power adapter (also reported by TTE). It’s not enough power for this enclosure. Does work with a powered USB hub*.
M.Yusuf has given the first ever report of a working Orico adapter! Make sure it’s the USB 3.1 Gen 2 version that says “Support UASP for NVMe SSD”. This is the way.
Canakit has been making very reliable power supplies for several Pi generations now. Using a 3.5A power supply will give enough extra power for your Pi to power the drive without causing instability
Known Working Adapters
Known Problematic Adapters (Naughty List)
Here is a list of common USB adapters that are known to have problems with the Raspberry Pi 4. You can get some of these adapters working by using quirks mode (see the “Fix (some) USB Adapter Problems Using Quirks” section below).
Only works in the USB 2.0 ports. Will not boot in a USB 3.0 port. I have two of these and can confirm they don’t work. RIP to Sabrent, our previous king of the Pi 3 era of adapters.
Despite earlier reports as working Ryan and one other have reported this adapter does not work unless you enable quirks mode! Don’t make Ryan’s sacrifice in vain and avoid this one.
Extremely cheap adapter from AliExpress — MADATALIEXPRESS bought 5 of them and none worked, PPCM had one working, very unreliable and slow when it does work, not recommended even if you get lucky!
So far we have not found a single ORICO adapter that has worked correctly so I would avoid that brand completely for the Raspberry Pi.
You may use other types of drives with the Pi such as M.2 SATA to USB 3.0, M.2 NVMe to USB 3.0 and mSATA to USB 3.0. Here’s some adapters I’ve used for those types of drives:
Power can be a serious problem with these drives. We are learning from the comments that you are especially likely to run into power issues with NVMe enclosures. A powered USB hub or a power adapter that puts out 3.5A comes not only just strongly recommended, it may actually be required that you choose one option or the other for your drive to function.
The specific requirements of how much power you’ll need depend on the adapter/enclosure and the model of your drive itself. As a very rough guideline, older models of drives tend to use more power than newer models of drives. 3.5″ form factor drives also use more power than 2.5″ drives. The earliest SSD models like first and second generation models are also well understood to use significantly more power than newer models. This is due to changes and improvements in technology over the years and even using different more efficient memory like 3D NAND. Some super high end performance drives will consume more power as well.
Here’s the current recommendations based on everyone’s comments combined with stuff I’ve personally used with the Pi:
Sabrent 2.5A Externally Powered USB 3.0 Hub gives a whopping dedicated 2.5A of power to satisfy even the thirstiest of drives
Known Working Adapters
Fix (some) USB Adapter Problems Using Quirks
Some of the very common adapters on the naughty list above (such as the Sabrent) can be made to work by using USB quirks to disable UAS mode on the drive. This lowers performance, but it’s still much faster than a SD card and your adapter won’t go to waste.
To find out the quirks we need to find the device ID string for your adapter and then add an entry to cmdline.txt telling the kernel to apply them on boot.
Find Your Adapter
To apply the quirks we first need to get the adapter id. We will use the sudo lsusb command:
$ sudo lsusb
Bus 003 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 002 Device 002: ID 174c:55aa ASMedia Technology Inc. Name: ASM1051E SATA 6Gb/s bridge, ASM1053E SATA 6Gb/s bridge, ASM1153 SATA 3Gb/s bridge, ASM1153E SATA 6Gb/s bridge
Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
Bus 001 Device 002: ID 2109:3431 VIA Labs, Inc. Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
On line 2 we can see my ASM1051E SATA 6Gb/s bridge adapter (it’s the known working StarTech.com 2.5″ SATA to USB* adapter). You will see something very similar to mine when you run the command and it shouldn’t be too hard to figure out which device it is. If you need more information add a -v switch to make the command sudo lsusb -v. This can sometimes add some additional details to make it easier to figure out which one is your adapter.
If you’re still not sure, we have another command that between the two that can narrow things down. Type / paste the following:
sudo dmesg | grep usb
[ 0.828535] usb usb3: New USB device found, idVendor=1d6b, idProduct=0002, bcdDevice= 4.19
[ 0.828568] usb usb3: New USB device strings: Mfr=3, Product=2, SerialNumber=1
[ 0.828597] usb usb3: Product: DWC OTG Controller
[ 0.828620] usb usb3: Manufacturer: Linux 4.19.75-v7l+ dwc_otg_hcd
[ 0.828644] usb usb3: SerialNumber: fe980000.usb
[ 0.830051] usbcore: registered new interface driver uas
[ 0.830182] usbcore: registered new interface driver usb-storage
[ 0.836488] usbcore: registered new interface driver usbhid
[ 0.836511] usbhid: USB HID core driver
[ 0.971598] usb 1-1: new high-speed USB device number 2 using xhci_hcd
[ 1.154217] usb 1-1: New USB device found, idVendor=2109, idProduct=3431, bcdDevice= 4.20
[ 1.154254] usb 1-1: New USB device strings: Mfr=0, Product=1, SerialNumber=0
[ 1.154281] usb 1-1: Product: USB2.0 Hub
[ 1.301989] usb 2-1: new SuperSpeed Gen 1 USB device number 2 using xhci_hcd
[ 1.332965] usb 2-1: New USB device found, idVendor=174c, idProduct=55aa, bcdDevice= 1.00
[ 1.332999] usb 2-1: New USB device strings: Mfr=2, Product=3, SerialNumber=1
[ 1.333026] usb 2-1: Product: ASM105x
[ 1.333048] usb 2-1: Manufacturer: ASMT
[ 1.333071] usb 2-1: SerialNumber: 123456789B79F
This is the dmesg log showing the hardware detection as hardware is activated on the Pi. If your log is really long you can generate fresh entries by just unplugging a device and plugging it back in and running the command again. Here we can clearly see that the ASM105x is what our StarTech adapter is being detected as.
Now we can go back to our first lsusb command and we want the 8 characters from the ID field that comes right after the Device:
Bus 002 Device 002: ID 174c:55aa ASMedia Technology Inc. Name: ASM1051E SATA 6Gb/s bridge
Our adapter’s ID is: 174c:55aa
Applying Quirks
To apply the quirks to our USB adapter we are going to edit /boot/cmdline.txt. Type:
sudo nano /boot/cmdline.txt
We are going to add the following entry into the very front of cmdline.txt:
usb-storage.quirks=XXXX:XXXX:u
In place of the X’s above you will put in your adapter’s ID that we got before. With the example commands I gave above mine would look like this: usb-storage.quirks=174c:55aa:u. After this my cmdline.txt looks like this (everything should be one continuous line, no line breaks!):
Now reboot the Pi. If the Pi fails to boot you can plug the SD card into the computer and go to /boot/cmdline.txt and undo the change we did so you can boot back in with your SD card.
Verifying Quirks
Once you have rebooted after changing cmdline.txt we can verify the quirks have been applied by doing another dmesg | grep usb command:
sudo dmesg | grep usb
[ 1.332924] usb 2-1: New USB device found, idVendor=174c, idProduct=55aa, bcdDevice= 1.00
[ 1.332957] usb 2-1: New USB device strings: Mfr=2, Product=3, SerialNumber=1
[ 1.332983] usb 2-1: Product: ASM105x
[ 1.333006] usb 2-1: Manufacturer: ASMT
[ 1.333028] usb 2-1: SerialNumber: 123456789B79F
[ 1.335967] usb 2-1: UAS is blacklisted for this device, using usb-storage instead
[ 1.336071] usb 2-1: UAS is blacklisted for this device, using usb-storage instead
[ 1.336103] usb-storage 2-1:1.0: USB Mass Storage device detected
[ 1.336479] usb-storage 2-1:1.0: Quirks match for vid 174c pid 55aa: c00000
[ 1.336611] scsi host0: usb-storage 2-1:1.0
This time we can see in dmesg that UAS was blacklisted for the device and it has loaded with the usb-storage driver instead. This driver tends to be more compatible with the “problematic adapters” but the performance is usually significantly lower. It’s definitely worth a try though as some adapters do better with the quirks performance-wise. The only way to know for sure is to run a benchmark (see storage benchmark section near the end).
USB Boot Instructions
There are a lot of steps to follow to set everything up properly. If you make a mistake the first time don’t spend too much time trying to correct it or figure out what you did wrong. It’s usually faster to burn the images again and reconfigure again rather than try to figure out which step you might have made a typo on. It’s much easier the second time!
Note: Don’t attempt to use raw dd commands to write the images. Too many silly things can go wrong that are checked for/fixed by the recommended programs. Do yourself a big favor and use one of the recommends to avoid spending a ton of time troubleshooting basic imaging problems!
Preparing SSD
We are going to burn a second identical copy of Raspbian to the SSD. This ensures everything the Pi needs to boot is there so we can use the SD card as a bootloader but our actual system will be on our nice fast SSD drive.
Note: Make sure you create the empty file named “ssh” on the boot partition of both drives if you are headless or don’t have a mouse/keyboard attached so you can ssh in on the first boot.
First Boot (SD card only, SSD unplugged)
Insert your freshly imaged SD card into the Pi and connect the power. Sign into the Pi for the first time.
Don’t do an apt-get upgrade/dist-upgrade or any additional configuration yet. Just stick with the instructions until we have finished configuration (especially since if something goes wrong you may have to start over and any other setup you did will be lost).
Once the Pi has finished booting and you have signed in for the first time plug in your SSD to your Pi’s USB 3.0 ports. The USB 3.0 ports are the ones that have the blue plastic inside instead of the black plastic (the black ones are USB 2.0 ports).
Change PARTUUID
We need to change the PARTUUID of our SSD’s partitions so the Pi doesn’t get confused about what device to boot from. Right now the partitions on both the SD card and the SSD are an exact match and we need them to be different so we can tell the Pi to boot specifically from our SSD’s partition.
We are going to use fdisk to change the SSD’s PARTUUID to the hexadecimal d34db33f to make our SSD easy to identify. Use the following:
$ sudo fdisk /dev/sda
Welcome to fdisk (util-linux 2.33.1).
Changes will remain in memory only, until you decide to write them.
Be careful before using the write command.
Command (m for help): p
Disk /dev/sda: 238.5 GiB, 256060514304 bytes, 500118192 sectors
Disk model: ASM105x
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x6c586e13
Device Boot Start End Sectors Size Id Type
/dev/sda1 8192 532479 524288 256M c W95 FAT32 (LBA)
/dev/sda2 532480 500118191 499585712 238.2G 83 Linux
Command (m for help): x
Expert command (m for help): i
Enter the new disk identifier: 0xd34db33f
Disk identifier changed from 0x6c586e13 to 0xd34db33f.
Expert command (m for help): r
Command (m for help): w
The partition table has been altered.
Syncing disks.
Your /dev/mmcblk0 and /dev/sda devices should now be different from each other. The SD card’s ID is 6c586e13 and the SSD’s PARTUUID is now
Update /boot/cmdline.txt
We are going to change cmdline.txt to point to the SSD for booting instead of the SD card. First make a backup of your existing cmdline.txt file with the following command:
sudo cp /boot/cmdline.txt /boot/cmdline.txt.bak
We’ve now created a backup you can restore if something goes wrong. If you need to restore your backup plug the SD card into a computer/device and replace cmdline.txt with cmdline.txt.bak that we made above. Now your Pi should boot normally again.
Open up /boot/cmdline.txt using nano or your favorite text editor:
Make the change and double check the line is what it should be,then press Ctrl+X to save our changes.
Note: cmdline.txt should be one long solid line with no breaks — don’t add any line breaks or the system won’t boot and you’ll need to restore the backup we made earlier!
Test SSD
We are now ready to test booting from the SSD. Restart your Pi by issuing a
sudo reboot
The first boot with your SSD can be slow due to running fsck on the drive. If you have a really large SSD it can take surprisingly long to check all that space. Once the check completes it will mark the drive clean and skip the disk check from now on. It can take over a minute or two sometimes for really big drives so give it at least that much time before assuming it didn’t work.
After signing in we can verify that the SSD is being used like this:
$ findmnt -n -o SOURCE /
/dev/sda2
Verify that partition has switched over as shown below to /dev/sda2 (SSD) instead of /dev/mmcblk0p2 (SD card).
Update /etc/fstab
We are now ready to edit the /etc/fstab file to point to our updated drive. To edit the file type:
We want to change the root ( / ) partition (PARTUUID ending with -02) to load our SSD’s PARTUUID instead of the SD card. Replace the 2nd partition’s PARTUUID field on the last line in the file with the d34db33f label we applied earlier with fdisk. After making the change my /etc/fstab file looks like this:
Press Ctrl+X to tell nano to save our changes. Now type sudo reboot to restart the Pi.
Note: We want to leave the first partition (/boot) on the SD card. If you change this to the SSD then apt will update your SSD instead of the SD card so they won’t be used during boot! Remember that we are using the SD card as a bootloader and that is why the firmware updates (such as start.elf, etc) should go there instead of the SSD’s boot partition (which is never used).
Resizing Filesystem
By default the partition on the SSD / Flash drive will only be 1.8G. The Pi expands this automatically on micro SD drives but we will need to do it ourselves for a SSD / Flash drive. To do this we need to expand the partition and then resize the file system.
First let’s open fdisk and print the partitions:
pi@raspberrypi:~ $ sudo fdisk /dev/sda
Welcome to fdisk (util-linux 2.33.1).
Changes will remain in memory only, until you decide to write them.
Be careful before using the write command.
Command (m for help): p
Disk /dev/sda: 238.5 GiB, 256060514304 bytes, 500118192 sectors
Disk model: ASM105x
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0xd34db33f
Device Boot Start End Sectors Size Id Type
/dev/sda1 8192 532479 524288 256M c W95 FAT32 (LBA)
/dev/sda2 532480 4390911 3858432 1.9G 83 Linux
There is the line we need. Our start value for /dev/sda2 (rootfs) is 532480. Next we need to remove and recreate the partition as a larger size.
If you make any mistakes during this command just close fdisk by pressing q. The changes won’t be written to disk. If you mess up any of the commands the drive will no longer boot and you’ll have to start over again so be careful!
Command (m for help): d
Partition number (1,2, default 2): 2
Partition 2 has been deleted.
Command (m for help): n
Partition type
p primary (1 primary, 0 extended, 3 free)
e extended (container for logical partitions)
Select (default p): p
Partition number (2-4, default 2): 2
First sector (2048-500118191, default 2048): 532480
Last sector, +/-sectors or +/-size{K,M,G,T,P} (532480-500118191, default 500118191): 500118191
Created a new partition 2 of type 'Linux' and of size 238.2 GiB.
Partition #2 contains a ext4 signature.
Do you want to remove the signature? [Y]es/[N]o: N
Command (m for help): w
The partition table has been altered.
Syncing disks.
If everything went well then type “w” and press enter. Otherwise press “q” to quit and try again. Once you enter “w” the changes will be permanently written to disk!
Now reboot the system. Type “df -h” to view the current disk:
We can see our disk is still 1.8G even after resizing the partition. That’s because we still have one more step! We need to resize the filesystem to fill our new partition space. For this we will use “sudo resize2fs /dev/sda2”:
sudo resize2fs /dev/sda2
resize2fs 1.44.5 (15-Dec-2018)
Filesystem at /dev/sda2 is mounted on /; on-line resizing required
old_desc_blocks = 1, new_desc_blocks = 15
The filesystem on /dev/sda2 is now 62448214 (4k) blocks long.
And that’s it! You will now be using all of your space on your drive.
Update Pi using apt
Now that we’ve updated fstab it is safe (and highly recommended) to update your Pi’s software. Type “sudo apt-get update && sudo apt-get dist-upgrade” to update the system and firmware.
Your system will now be running completely from your USB drive! To verify this, run the command “findmnt -n -o SOURCE” / to ensure your root partition has switched over as shown below to /dev/sda2 instead of /dev/mmcblk0p2.
Verify SSD Performance
You can make sure everything is running correctly (and as fast as it should be) by running my quick storage benchmark. You can run the benchmark with the following one-liner:
This will give you a score you can compare to the other Raspberry Pi Storage Benchmark results and make sure that you are getting an equivalent speed to your peers with the same device!
Conclusion
The Samsung 950 Pro NVMe drive in the featured picture scored a 9189 on the Raspberry Pi Storage Benchmark. The previous all-time record score on a Pi 3B+ was 3561. The performance gains are very real and very dramatic.
For me getting this performance is well worth having to waste a micro SD card just to be a bootloader. I am largely after the USB 3.0 bus and gigabit ethernet performance improvements and using this method I am able to achieve the performance I was after without waiting an indeterminate amount of time for the feature to be added back in!
Although there are ongoing compatibility issues and we lack the super easy native USB booting support we had before I’m more than willing to go through the growing pains to finally get rid that ancient USB 2.0 bus! Just make sure if you are planning to build a system you plan your adapters and parts accordingly.
761 thoughts on “Raspberry Pi 4 USB Boot Config Guide for SSD / Flash Drives”
Walter
I just bought the ELUTENG enclosure seeing it in the list of working adapters. I tried it, it will NOT boot as a stand alone ssd using USB. I initially bought ORICO 2.5″ USB 3.0 External Hard Drive Enclosure for 2.5 Inch SATA HDD and SSD[Optimized for SSD. The board looks exactly the same as the ELUTENG, just the name on the case is different. This one will also not boot from USB. The one that I got working and boots off USB is ELEGIANT USB 3.0 to SATA, Super Speed USB 3.0 to SATA 22Pin Converter Adapter Cable for 2.5Inch Hard Drive Disk. This one sadly is not an enclosure but only a cable. (https://amzn.to/3rz5qUL) I am using the Raspberry Pi 4b – with the the bootloader dated Sep 3 2020. I started using the instructions on the page, I eventually used the page below. https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/msd.md
How are you powering your Pi? The enclosures in general seem to be pretty power hungry (much more so than the adapters). Are you using a powered USB hub or a minimum of a 3.5A power adapter? Do you have any larger USB-C chargers around the house you could test with like a cell phone charger / laptop charger / something that puts out more than 3.0A?
I don’t think it’s a coincidence that you’re having trouble with the enclosures and the adapters are working. The ELUTENG one in particular seems to be pretty popular and I suspect that one might work with some more power.
Sometimes firmware can affect these things but you said you’re using the Sep 3 2020 firmware so you should be covered there! I think you will probably be able to get that ELUTENG one working though with some experimentation here.
Hi I am using a 65 Watt USB-C adapter to power the raspberry pi 4 using the USB-C port on the pi. I have a USB Digital Tester. I put in series with the RP4 USB-C power input and it reads 5 V and only 0.8 amp using the Sata to usb connector as well as with the ELUTENG adapter. I do not have a USB-C powered hub. I am searching for a powered hub on Amazon.
That sounds like plenty of power. The Pi 4 can’t pass on all of that power but it should be passing on enough to power those enclosures. I was getting the impression that the NVMe ones were a lot more power hungry than the 2.5″ SATA variants although it depends on what drive you are plugging in.
I suspect the USB hubs may also be helping for another reason than just power. I think sometimes the USB hubs are a little bit better at talking universally to devices than the Pi is when it’s directly connected. There are actually some chipset level incompatibilities with the Pi using certain adapters that I think the USB hubs mask. They’re acting as a sort of translator or middleman!
I’m curious what yours is going to end up being because the ELUTENG as far as I’m aware has been working, but maybe there’s a new model or variant out. For example, I would say most of these adapters at this point have a USB 3.0 variant and a USB 3.1 variant. Can you tell us exactly which one you got so I can compare it with some of the previous results? The whole reason I made the list was to help people *not* buy the wrong adapters so I definitely want to figure out what the situation is with this one and update it accordingly!
There was a QR code on the ELUTENG board, scanning it with my phone got, pcba-05-000-0214 p010200523050013 The two chips on the board are, FM25004A the other JMS578.
I ordered a a powered usb hub, I will let you know if the enclosure works using it and the boots via USB. I just tried an old USB-2 sata enclosure with the same ssd, but the rpi 4 did not even recognize it. The drive I am testing this with is a Crucial BX500 2.5 SSD 120GB rated at 5 volt, 1.7 Amp, Model CT120BX500SSD1 FW: M6CR013
Hey James, regarding the EULTENG enclosure I definitely would NOT recommend it. Out of 5 tries, I got it to boot once off of USB and it took about 20 minutes to boot. The one time it booted was using the USB sata cable. The other times I tested it with a powered hub and EULTENG enclosure it did not help. It would look like it is starting the drive lights would flash, I got a Raspberry pi logo on the screen, I let it sit for about an hour and gave up, it never fully booted. I was testing it with a SP SSD A58 3 nano, 256 GB The OS i am running is the 2020-12-02-raspios-buster-armhf-full Using the Raspberry Pi 4b – with the the bootloader dated Sep 3 2020.
Raspberry Pi 4 Model B + 4GB using the official usb-c power adapter. Using the TDBT M.2 NVMe SSD Enclosure. Controller chip: JMicron JMS583(PCIe 3.0×2) Bus 002 Device 002: ID 152d:0562 JMicron Technology Corp. / JMicron USA Technology Corp.
Works good for the past 2 days without any quirks.
Hi Walter here, I am bad. I accidentally forgot the micro sd was also in the RPI 4. Right now without the micro sd neither enclosure boots with my ssd drive. I am going to reflash the ssd drive.
Hi here is an update, I bought the ELUTENG ssd drive enclosure and put my Crucial SSD drive in it and my RPi 4 booted quickly. Comparing the ELUTENG and the Orico enclosures the components are in the same place the part numbers look identical. Besides the name being different the two enclosures look identical. I put my SSD drive back into the Orico and it now also boots quickly and all four restarts, warm and cold have been quick. I also ran the original drive cable with both enclosures. End result now both the ELUTENG and Orico enclosures allow quick booting now. Did running the straight sata to USB3 cable do something?
I’m glad I found this page, I’ve been banging my head against a wall for a week now. I bought four different usb3 2.5″ enclosures and all have had their problems.
I’d like to use my Samsung 850evo 1tb ssd with my pi400. Can you recommend a good enclosure, ideally UsbC-UsbA, fast, that works equally well on both Windows10 and the Pi400 using trim and uasp?
Below my findings: Sabrent EC-UM30 and EC-UASP: Flashed both the latest Sabrent firmware on Windows, and also latest firmware as per https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=245931&start=50#p1570022. They don’t boot using usb3.0 on the pi400. They corrupt the image when used in conjunction with Win32DiskImager for writing .IMG files to the ssd. This results in a missing boot partition and difficulty in removing partitions using the Windows storage manager. Also ssd becomes too hot to the touch which is of course worrying. HOWEVER, the EC-UM30 has the fastest performance on Windows, it’s amazing, stable constant real 300Mb/s transfer read/write. Benchmark is 400+/400+ using CrystalDisk.
UGREEN 2.5″ 3.1gen2 Metallic and plastic enclosures: They work OK on the pi400 as you can see: https://storage.jamesachambers.com/benchmark/36047 HOWEVER, occasionally there seems to be a 20 seconds pause on the pi400 when copying files, or even when listing folder content. This worries me a bit but I can’t fully replicate it. On Windows both UGREEN enclosures are slower, and they drop down to 0.5/1Mb/s speeds once every 15 seconds. Speed oscillates regularly. Copying large images becomes impossible leading up to 14hours vs 50minutes using the Sabrent enclosures. Benchmark is 400/200 using CrystalDisk.
Many thanks Luca
P.s. Also, could some of these issues be caused by the official Raspberry Pi4 5.1v 3A power adapter? Please note I only have the ssd via enclosure and logitech k400 dongle (usb2) attached to my pi400.
Sorry for the slow reply. It’s definitely possible that it’s power supply. Different drives and adapters can change the power requirements.
If you have one of the 3.5A power adapters that should give a lot of overhead. They’re cheap and do the job!
There’s another possibility to consider. Have you ran the trim command on those drives? It’s:
sudo fstrim -av
to do all drives that are TRIM capable on the system (from Linux) and the Windows 10 built in “Defragment and Optimize Drives” utility can also run the TRIM command if it recognizes the drive as a “Solid state drive”. Some adapters don’t support passing through the TRIM command and it’s something you can usually find someone mentioning in the Amazon reviews for these adapters as it’s pretty important and a huge plus, if not a necessity to have.
The drives are definitely designed to be used this way and the TRIM command is usually taken care of by your PC’s periodic automatically scheduled maintenance in most modern environments. The equivalent for hard drives or HDDs would be automatic defragmenting on a schedule, although fragmentation is different and doesn’t happen on SSDs but SSDs still run best with periodic maintenance and for them that is the TRIM command.
The reason I wanted to suggest this is that when I’ve been benchmarking drives over and over for my Raspberry Pi Storage Benchmark I’ve ran into this issue before. Sometimes after a brand new fresh image the drive will be slower than it was before even though nothing was installed yet and more free space was available!
Eventually I learned that is because when you delete data from a SSD it isn’t destroyed on the drive for a multitude of reasons like wear-leveling but also that it takes time to destroy the data before you write there again. The trim command tells the drive to take the time and destroy ALL the data that is no longer in use. This restores the speed on a lot of devices! Each time you do one of the write tests on the drive even though the drive cleans up after itself the drive will start to have to overwrite which can limit your performance.
Since you are benchmarking the drives and writing a lot of data over and over for testing it’s pretty easy to fill them up very quickly. Much more quickly than would happen during normal usage. It’s a good habit to trim them with the Windows and Linux tool to just clean them up so they’re performing at their best and not overwriting old data as part of your routine and to rule it out right off the bat!
In Windows 10 they made the “Optimize Drive” running periodically the default on Windows 10 to maintain performance but it isn’t often is not automatically scheduled on Linux. However, even in Windows it doesn’t run automatically on external drives.
Every time I run fstrim on a Linux system it will give a large # of GB freed from the drive and benchmarking performance often increases. If all else is well this can be a bottleneck. The part you said about testing transferring large files also seems to make it more likely that this is at least part of the explanation for some of the problems you’re seeing.
Have you tried running the EC-UM30 with the quirks mode enabled plugged into a USB 3.0 port? Sometimes that will get it to boot (usually with a performance penalty, but occasionally it can be faster when there’s buggy chipsets at play) and I didn’t see you mention quirks in there.
The new beta bootloader firmware also has a bunch of fixes in it related to both chipset adapters and specific drives. If you don’t feel uncomfortable with trying the beta firmware it may also make a difference!
I am running this attached to Raspberry pi 4 B, 2 GB ram. It works, but incredibly slow. I takes about 4 minutes to boot, everything loads slowly, way slower then the micro-sd card. I am using a Crucial BX500 2.4 SSD 120GB. At first I thought I did the boot procedure wrong. I found this article about issues with some enclosures. I took the drive out of the ORICO enclosure and ran a straight sata to USB3 cable, I now get a 26 second boot time and a responsive system.
Have you tried the quirks mode at the bottom of this article by chance? Sometimes it can clean up the performance in the case of buggy chipsets.
You may also want to consider giving the beta firmware a try and see if the fixes which I covered more in depth in the comments of my most recent article here make any difference. I have some instructions on how to do that in this article if you’re interested.
Note: I don’t recommend updating to beta firmware for most people unless you’re having specific issues that are addressed in those updates, otherwise just wait for it to come to stable to avoid potential headaches.
I don’t own that adapter personally and my understanding is it has been working from the comments but depending on what you find and if we continue to see issues like this it may need to be moved to the danger zone section.
UPDATE: I did end up moving the Orico transparent enclosure to the trouble list. Avoid!
Samsung EVO 860 250 GB in a Orico 2569S3 (Banggood) with ASM1153E controller works like a charm. No quirks. Benchmark score 7998. Sweet!
The Orico 2139 with the Jmicron 578 of course doesn’t without setting the quirk flag. I tried a firmware upgrade but unfortunately flashed a wrong one which bricked it :-/
I had no success booting from a pair of cheap adapters. I could transfer the OS using rpi-imager on my PC but no luck with booting. I then plugged in the adapters to a powered USB 3.0 hub and they worked great. I’m guessing there was a voltage drop across the adapter cable that was the problem.
Dears, I just got the working adapter here in Czechia from ITEC – U3MYSAFE025 for 2,5″ and it is absolutly different experience using the drive without qirks – it does not use a lot of cpu for drive operations. I can drain the cpu to 100% and still watch movies and play music without any stutters sound distortion. It has ASM115X . I especially thanks to the customer support of that company for letting me know which adapters has this chipset. They told me that also MYSAFE35U401 for 3,5 drives has this chipset. James, please add it to the list.
I can confirm that the Orico DOES NOT work in uasp mode with the RPi4 (and 400) – it uses the Jmicron chipset (as reported by lsusb). Using the “quirks” workaround to turn off uasp mode does enable the interface to boot, but at a lower speed (about half the speed of uasp mode). Plugging the adaptor into an already running pi (booted from another drive) does recognise the unit in uasp mode and it does work at full speed, but of course only as a second drive.
James, great article for details. Hoping I don’t need all that…….. explains what is going on! Confused here tho. I have a microSD [vilros or canakit standard pkg] with noobs on a pi4, and I installed raspian. I made changes to do the usb boot, and I have raspi-config changed to boot usb. I connected via usb 3 a 64GB Samsung SSD, and I used “disks” utility to erase partitions, so I have free space now on the SSD. I am attempting to “SD card copy” from the microSD to the SSD and am having trouble. It creates identical partitions, but never seems to be able to complete the copy. failures of one kind or another after “copying partition 1.” The partition 1 is “recovery” and is noobs I assume. The rest is an extended partition with partitions 5, 6 and 7 as “settings” “/boot” and “/” -not sure what “settings” is……. Does “noobs” get in the way of this? I am assuming a boot always goes thru the noobs “recovery” partition since it gives me the option to reinstall. I was able to mount the partitions on the SSD and see them [before I erased it all], so I presume the usb driver is working. Thanks for any help! john
Thank you for your guide. Unfortunately I´m running into problems. I can change partuuid but after updating my cmdline.txt the system wont reboot. Is it possible that my USB to SSD adapter is able to detect the SSD but is not able to boot from it?
first off thank you for the article. I set up my pi 4 a while ago and it was working perfectly. But after the recent OS updates my sytem was shot. I was not able to boot normally. So I reset and followed your guide again. Now, I am having the same issue mkapsville is having: edited boot order. Now, it is not rebooting. No I/O LEDs blinking. I followed a different guide, which helped but my sequential read/write speeds dropped in half. Tested different adapters and M2s, same issue. Were there changes to newest Raspbian that might be causing problems?
There is a BOOT_ORDER field in the new bootloader. Depending on what this is set to it will attempt to boot from the SD card for 0x1, a USB device for 0x3, etc.
This support is new from after I wrote this guide (it didn’t exist at the time). It looks like maybe it is coming out of beta here and being released to everybody possibly!
It looks like the command to edit the bootloader is:
sudo -E rpi-eeprom-config --edit
This will only work if your system is fully patched and up to date it looks like but you can see the options here:
It looks like it is time for me to write the guide for the new bootloader here, but give editing the BOOT_ORDER field in there a try (0x4 for a USB mass storage device, other choices are in the link earlier in the post) and see if it works for you!
Thank you for the quick reply! Yeah, updates seemed to have rolled September which is right around the time I started having this issue. Right now I basically just reimaged my m2 ssd with the raspian imager using the newest OS build, enabled headless ssh mode, and plugged the ssd into one of the usb3 ports – no SD card involved. Works fine: I can restart, run updates etc. but I am only downloading with 29MBs/sec whereas before I was hitting 50MBs/sec. I ruled out my ISP and other potential bottlenecks (I have a 500mbit connection which I am fully getting on other devices).
I will backup my current setup and test your guide. Thank you again!
So, I realized changing the boot order does not really help since I reset everthing and I am already able to to boot from usb port. Still do not know why my download speed is so slow though. Basically cannot use my pi as my download manager anymore.
You said this drive had recently been reimaged. Have you ran the ‘trim’ command on the drive? This may be a bit of a long shot but during my storage testing I’ve noticed this sometimes. Trim will tell the drive to actually drop all data marked for deletion/no longer used from the device. You can try it with:
sudo fstrim -av
This keeps the SSD from having to write over this old data (which takes time). When you reimage a drive the entire old image is still left on there until the trim command is run. If you want to understand more there’s a pretty good writeup of SSDs and trim here.
After this give the device a reboot and try running your test again. You may also want to consider running my storage benchmark on it. This tests your drive’s speed using standard benchmarking tools/techniques and returns a score that we can easily compare against other submissions with the same drive. We can also look at the raw numbers and see where it’s struggling (things like random IO performance vs read / write performance, etc). It’s very likely other people have tested your same drive or one very similar we can compare it to!
There are some storage adapters and even SSDs that don’t support trim out there but they’re all older. If you’re using a modern one it should be able to do it! Adapter quality can also impact performance but if you’re comparing performance to how it was from before the updates it doesn’t make sense that the adapter would be the cause in this case.
I know your overall concern is network download performance but on the Pi a lot of the busses/lanes that carry this data are shared between the different components so it definitely can matter! At least on the Pi 4 they finally separated the ethernet from the main USB bus, but even so storage is still very likely to be the bottleneck here.
If you have any additional drives you can image and run the same network test on this would narrow things down. What download rate do you get with a standard SD card for example would be a good thing to check. It could definitely end up being related to a package or a change in one of the updates in which case you should see the same slowdown on anything running the latest version.
I am actually getting decent download speeds (300-350mbits) running speedtest-cli via terminal but big spikes. I have also been using your benchmark tool and my score is close to 7000 on average so those parameters look fine. I am using JDownloader as my main download tool which downloads to my DS413j. I ruled out my NAS as well by using a download folder on my pi as the destination. Maybe it is related to the JDownloader’s recent updates and its terrible Java code base… I have not found an alternative for Raspbian though which supports filehosters, etc. My pi is passively cooled and remains under 70 degrees Celcius under heavy load so CPU/GPU throttling should not be it. Also thank you for the trim tip. Just ran your cmd. Here are my most recent raw stats.
HDParm Disk Read 293.52 MB/s HDParm Cached Disk Read 278.41 MB/s DD Disk Write 169 MB/s FIO 4k random read 16370 IOPS (65483 KB/s) FIO 4k random write 7981 IOPS (31925 KB/s) IOZone 4k read 28309 KB/s IOZone 4k write 23576 KB/s IOZone 4k random read 17552 KB/s IOZone 4k random write 24261 KB/s
I definitely see it fluctuating there. I gave it a few runs on my Pi and it chose different servers a couple of times. Here’s what mine looked like:
Hosted by Think Smart Inc 7.558 ms Download: 341.58 Mbit/s Upload: 372.93 Mbit/s Hosted by UTOPIA Fiber 5.288 ms Download: 342.06 Mbit/s Upload: 364.29 Mbit/s Hosted by UTOPIA Fiber 5.757 ms Download: 338.92 Mbit/s Upload: 330.31 Mbit/s
So it looks like it can fluctuate but not usually by that much. Are you wired in or using wireless? Do you have access to another internet connection or network like at work / friend / family you could take it to and run this same test to see if you see the same type of network fluctuations?
It’s also possible it’s Java related. Have you tried it on the 64 bit Raspbian beta? That would run on a 64 bit Java Virtual Machine which can definitely benefit some applications. It can also be worse in some cases since it typically uses a little more memory so it’s good for 8 GB Pis but not so good for smaller capacities.
It definitely seems like there is some sort of network performance issue going on from your speedtest-cli numbers though!
Hi, I am wired directly into my WIFI Router (1Gbit). I only have the 4GB RAM Version and do not want to deal with Betas. I will double check my basic troubleshooting steps (reboot router, check cables, try different ethernet ports, maybe even try over wireless since my Pi sits next to the router, etc.)
Are there any JDownloader alternatives that you know of?
I’ll be honest and say I’m not super familiar with JDownloader itself. I have definitely had a ton of run-ins with Java and the JDK over the years via my Minecraft Pi server script.
I think you are on the right track troubleshooting and narrowing things down as much as possible. These types of issues can be really difficult to diagnose because something like network transfer speed has so many underlying pieces to it that have to all work together to achieve maximum speeds.
I would keep looking for alternatives if possible just as a comparison and keep trying to test different variables. I think you’re getting close here!
Mark
no other network I can try except a hotspot from my fine but that is that good reference.
I wonder a bit about that you seem to have posted your very good article on 19th of october this year while other sources stated about a month before that USB bootloading with the PI4 is officially supported from bootloader Version 2020-09-03? Apart from that your list of working/nonworking adapters is highly useful of course. Regards, jan
Sorry, that’s because it’s showing when it was last updated and that was the last time it updated the list. It’s on my to-do list to add the original published date below it but haven’t got to it yet, but you’re quite right! Thanks for the kind words on the article!
I just bought the ELUTENG enclosure seeing it in the list of working adapters. I tried it, it will NOT boot as a stand alone ssd using USB.
I initially bought ORICO 2.5″ USB 3.0 External Hard Drive Enclosure for 2.5 Inch SATA HDD and SSD[Optimized for SSD.
The board looks exactly the same as the ELUTENG, just the name on the case is different. This one will also not boot from USB.
The one that I got working and boots off USB is ELEGIANT USB 3.0 to SATA, Super Speed USB 3.0 to SATA 22Pin Converter Adapter Cable for 2.5Inch Hard Drive Disk.
This one sadly is not an enclosure but only a cable. (https://amzn.to/3rz5qUL)
I am using the Raspberry Pi 4b – with the the bootloader dated Sep 3 2020.
I started using the instructions on the page, I eventually used the page below.
https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/msd.md
Hey Walter,
How are you powering your Pi? The enclosures in general seem to be pretty power hungry (much more so than the adapters). Are you using a powered USB hub or a minimum of a 3.5A power adapter? Do you have any larger USB-C chargers around the house you could test with like a cell phone charger / laptop charger / something that puts out more than 3.0A?
I don’t think it’s a coincidence that you’re having trouble with the enclosures and the adapters are working. The ELUTENG one in particular seems to be pretty popular and I suspect that one might work with some more power.
Sometimes firmware can affect these things but you said you’re using the Sep 3 2020 firmware so you should be covered there! I think you will probably be able to get that ELUTENG one working though with some experimentation here.
Hi I am using a 65 Watt USB-C adapter to power the raspberry pi 4 using the USB-C port on the pi.
I have a USB Digital Tester. I put in series with the RP4 USB-C power input and it reads 5 V and only 0.8 amp using the Sata to usb connector as well as with the ELUTENG adapter. I do not have a USB-C powered hub. I am searching for a powered hub on Amazon.
Hey Walter,
That sounds like plenty of power. The Pi 4 can’t pass on all of that power but it should be passing on enough to power those enclosures. I was getting the impression that the NVMe ones were a lot more power hungry than the 2.5″ SATA variants although it depends on what drive you are plugging in.
I suspect the USB hubs may also be helping for another reason than just power. I think sometimes the USB hubs are a little bit better at talking universally to devices than the Pi is when it’s directly connected. There are actually some chipset level incompatibilities with the Pi using certain adapters that I think the USB hubs mask. They’re acting as a sort of translator or middleman!
I’m curious what yours is going to end up being because the ELUTENG as far as I’m aware has been working, but maybe there’s a new model or variant out. For example, I would say most of these adapters at this point have a USB 3.0 variant and a USB 3.1 variant. Can you tell us exactly which one you got so I can compare it with some of the previous results? The whole reason I made the list was to help people *not* buy the wrong adapters so I definitely want to figure out what the situation is with this one and update it accordingly!
I bought the ELUTENG enclosure from Bestbuy.ca, https://www.bestbuy.ca/en-ca/search?search=%2014856418
ELUTENG SSD External Hard Drive Enclosure 2.5 inch SATA to USB3.0 UASP Clear Portable Hard Drive Case Max 2T HDD
There was a QR code on the ELUTENG board, scanning it with my phone got, pcba-05-000-0214 p010200523050013
The two chips on the board are, FM25004A the other JMS578.
I ordered a a powered usb hub, I will let you know if the enclosure works using it and the boots via USB.
I just tried an old USB-2 sata enclosure with the same ssd, but the rpi 4 did not even recognize it.
The drive I am testing this with is a Crucial BX500 2.5 SSD 120GB rated at 5 volt, 1.7 Amp, Model CT120BX500SSD1 FW: M6CR013
Hey James, regarding the EULTENG enclosure I definitely would NOT recommend it. Out of 5 tries, I got it to boot once off of USB and it took about 20 minutes to boot. The one time it booted was using the USB sata cable. The other times I tested it with a powered hub and EULTENG enclosure it did not help.
It would look like it is starting the drive lights would flash, I got a Raspberry pi logo on the screen, I let it sit for about an hour and gave up, it never fully booted. I was testing it with a SP SSD A58 3 nano, 256 GB
The OS i am running is the 2020-12-02-raspios-buster-armhf-full
Using the Raspberry Pi 4b – with the the bootloader dated Sep 3 2020.
Hey Walter,
I ended up moving this to the don’t recommend list, thanks a ton for the feedback!
Hello
from the Samsung SSD Starter kit the USB 3.0 SATA III adapter works well. Boot from USB3.0 port with an 120GB PNY SSD9C120GS4DH08B-T (5VDC; 0.5A) on PI4 8GB
If you want I can share the details..
Starter kit : https://www.samsung.com/de/support/model/MZ-75E250RW/
Thanks Rene, I’ve added this to the list!
ICY-BOX IB-183M2 M.2 SATA you can put on the naughty list
Raspberry Pi 4 Model B + 4GB using the official usb-c power adapter.
Using the TDBT M.2 NVMe SSD Enclosure.
Controller chip: JMicron JMS583(PCIe 3.0×2)
Bus 002 Device 002: ID 152d:0562 JMicron Technology Corp. / JMicron USA Technology Corp.
Works good for the past 2 days without any quirks.
Thanks WorkHard, I have put it on there!
WD 1 TB My Passport Go Portable SSD – This also works
https://www.amazon.co.uk/WD-TB-Passport-Portable-SSD/dp/B07MVWG9KM/ref=as_li_ss_tl?dchild=1&keywords=wd+ssd&qid=1607287968&s=electronics&sr=1-23&th=1&linkCode=ll1&tag=codemallet-21&linkId=41cf5463dc1a6601992c27594517c51b&language=en_GB
UGREEN USB C Hard Drive Enclosure External 2.5 inch HDD Caddy – this works perfectly
https://www.amazon.co.uk/gp/product/B0851B6TCC/ref=as_li_ss_tl?ie=UTF8&psc=1&&linkCode=ll1&tag=codemallet-21&linkId=da5056ab9ce19bacdf8eb15580d9ca60&language=en_GB
Thanks, I’ve added it to the list!
Hi,
also ORICO 2588US3-BK won’t work.
https://www.orico.cc/us/product/detail/3266.html#link3
https://www.amazon.it/gp/product/B07G9K2YTK/ref=as_li_ss_tl?ie=UTF8&psc=1&&tag=&&linkCode=ll1&tag=codemallet04b-21&linkId=b9e1b8bc9f98c2a8588648b99748e314&language=it_IT
Thanks for the warning Andrea, I have added it to the danger list!
Hi Walter here, I am bad. I accidentally forgot the micro sd was also in the RPI 4.
Right now without the micro sd neither enclosure boots with my ssd drive.
I am going to reflash the ssd drive.
Hi here is an update, I bought the ELUTENG ssd drive enclosure and put my Crucial SSD drive in it and my RPi 4 booted quickly.
Comparing the ELUTENG and the Orico enclosures the components are in the same place the part numbers look identical. Besides the name being different the two enclosures look identical.
I put my SSD drive back into the Orico and it now also boots quickly and all four restarts, warm and cold have been quick.
I also ran the original drive cable with both enclosures.
End result now both the ELUTENG and Orico enclosures allow quick booting now.
Did running the straight sata to USB3 cable do something?
I’m glad I found this page, I’ve been banging my head against a wall for a week now. I bought four different usb3 2.5″ enclosures and all have had their problems.
I’d like to use my Samsung 850evo 1tb ssd with my pi400. Can you recommend a good enclosure, ideally UsbC-UsbA, fast, that works equally well on both Windows10 and the Pi400 using trim and uasp?
Below my findings:
Sabrent EC-UM30 and EC-UASP:
Flashed both the latest Sabrent firmware on Windows, and also latest firmware as per https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=245931&start=50#p1570022.
They don’t boot using usb3.0 on the pi400.
They corrupt the image when used in conjunction with Win32DiskImager for writing .IMG files to the ssd.
This results in a missing boot partition and difficulty in removing partitions using the Windows storage manager. Also ssd becomes too hot to the touch which is of course worrying.
HOWEVER, the EC-UM30 has the fastest performance on Windows, it’s amazing, stable constant real 300Mb/s transfer read/write. Benchmark is 400+/400+ using CrystalDisk.
UGREEN 2.5″ 3.1gen2 Metallic and plastic enclosures:
They work OK on the pi400 as you can see: https://storage.jamesachambers.com/benchmark/36047
HOWEVER, occasionally there seems to be a 20 seconds pause on the pi400 when copying files, or even when listing folder content. This worries me a bit but I can’t fully replicate it.
On Windows both UGREEN enclosures are slower, and they drop down to 0.5/1Mb/s speeds once every 15 seconds. Speed oscillates regularly. Copying large images becomes impossible leading up to 14hours vs 50minutes using the Sabrent enclosures. Benchmark is 400/200 using CrystalDisk.
Many thanks
Luca
P.s.
Also, could some of these issues be caused by the official Raspberry Pi4 5.1v 3A power adapter? Please note I only have the ssd via enclosure and logitech k400 dongle (usb2) attached to my pi400.
Hey Luca,
Sorry for the slow reply. It’s definitely possible that it’s power supply. Different drives and adapters can change the power requirements.
If you have one of the 3.5A power adapters that should give a lot of overhead. They’re cheap and do the job!
There’s another possibility to consider. Have you ran the trim command on those drives? It’s:
sudo fstrim -avto do all drives that are TRIM capable on the system (from Linux) and the Windows 10 built in “Defragment and Optimize Drives” utility can also run the TRIM command if it recognizes the drive as a “Solid state drive”. Some adapters don’t support passing through the TRIM command and it’s something you can usually find someone mentioning in the Amazon reviews for these adapters as it’s pretty important and a huge plus, if not a necessity to have.
The drives are definitely designed to be used this way and the TRIM command is usually taken care of by your PC’s periodic automatically scheduled maintenance in most modern environments. The equivalent for hard drives or HDDs would be automatic defragmenting on a schedule, although fragmentation is different and doesn’t happen on SSDs but SSDs still run best with periodic maintenance and for them that is the TRIM command.
The reason I wanted to suggest this is that when I’ve been benchmarking drives over and over for my Raspberry Pi Storage Benchmark I’ve ran into this issue before. Sometimes after a brand new fresh image the drive will be slower than it was before even though nothing was installed yet and more free space was available!
Eventually I learned that is because when you delete data from a SSD it isn’t destroyed on the drive for a multitude of reasons like wear-leveling but also that it takes time to destroy the data before you write there again. The trim command tells the drive to take the time and destroy ALL the data that is no longer in use. This restores the speed on a lot of devices! Each time you do one of the write tests on the drive even though the drive cleans up after itself the drive will start to have to overwrite which can limit your performance.
Since you are benchmarking the drives and writing a lot of data over and over for testing it’s pretty easy to fill them up very quickly. Much more quickly than would happen during normal usage. It’s a good habit to trim them with the Windows and Linux tool to just clean them up so they’re performing at their best and not overwriting old data as part of your routine and to rule it out right off the bat!
In Windows 10 they made the “Optimize Drive” running periodically the default on Windows 10 to maintain performance but it isn’t often is not automatically scheduled on Linux. However, even in Windows it doesn’t run automatically on external drives.
Every time I run fstrim on a Linux system it will give a large # of GB freed from the drive and benchmarking performance often increases. If all else is well this can be a bottleneck. The part you said about testing transferring large files also seems to make it more likely that this is at least part of the explanation for some of the problems you’re seeing.
Have you tried running the EC-UM30 with the quirks mode enabled plugged into a USB 3.0 port? Sometimes that will get it to boot (usually with a performance penalty, but occasionally it can be faster when there’s buggy chipsets at play) and I didn’t see you mention quirks in there.
The new beta bootloader firmware also has a bunch of fixes in it related to both chipset adapters and specific drives. If you don’t feel uncomfortable with trying the beta firmware it may also make a difference!
Hi I also recent bought the ORICO 2.5″ USB 3.0 External Hard Drive Enclosure for 2.5 Inch SATA HDD from Amazon.ca https://www.amazon.ca/gp/product/B01LY97QE8/ref=as_li_ss_tl?ie=UTF8&psc=1&&linkCode=ll1&tag=codemallet08-20&linkId=5a6cf3f57d6586f544e29aa20ce399e7&language=en_CA
Bought November 25, 2020
Someone else commented this enclosure works.
I am running this attached to Raspberry pi 4 B, 2 GB ram.
It works, but incredibly slow. I takes about 4 minutes to boot, everything loads slowly, way slower then the micro-sd card.
I am using a Crucial BX500 2.4 SSD 120GB. At first I thought I did the boot procedure wrong.
I found this article about issues with some enclosures.
I took the drive out of the ORICO enclosure and ran a straight sata to USB3 cable, I now get a 26 second boot time and a responsive system.
Hey Walter,
Have you tried the quirks mode at the bottom of this article by chance? Sometimes it can clean up the performance in the case of buggy chipsets.
You may also want to consider giving the beta firmware a try and see if the fixes which I covered more in depth in the comments of my most recent article here make any difference. I have some instructions on how to do that in this article if you’re interested.
Note: I don’t recommend updating to beta firmware for most people unless you’re having specific issues that are addressed in those updates, otherwise just wait for it to come to stable to avoid potential headaches.
I don’t own that adapter personally and my understanding is it has been working from the comments but depending on what you find and if we continue to see issues like this it may need to be moved to the danger zone section.
UPDATE: I did end up moving the Orico transparent enclosure to the trouble list. Avoid!
Samsung EVO 860 250 GB in a Orico 2569S3 (Banggood) with ASM1153E controller works like a charm. No quirks. Benchmark score 7998. Sweet!
The Orico 2139 with the Jmicron 578 of course doesn’t without setting the quirk flag. I tried a firmware upgrade but unfortunately flashed a wrong one which bricked it :-/
I had no success booting from a pair of cheap adapters. I could transfer the OS using rpi-imager on my PC but no luck with booting. I then plugged in the adapters to a powered USB 3.0 hub and they worked great. I’m guessing there was a voltage drop across the adapter cable that was the problem.
Dears, I just got the working adapter here in Czechia from ITEC – U3MYSAFE025 for 2,5″ and it is absolutly different experience using the drive without qirks – it does not use a lot of cpu for drive operations. I can drain the cpu to 100% and still watch movies and play music without any stutters sound distortion. It has ASM115X . I especially thanks to the customer support of that company for letting me know which adapters has this chipset. They told me that also MYSAFE35U401 for 3,5 drives has this chipset. James, please add it to the list.
I can confirm that the Orico DOES NOT work in uasp mode with the RPi4 (and 400) – it uses the Jmicron chipset (as reported by lsusb). Using the “quirks” workaround to turn off uasp mode does enable the interface to boot, but at a lower speed (about half the speed of uasp mode). Plugging the adaptor into an already running pi (booted from another drive) does recognise the unit in uasp mode and it does work at full speed, but of course only as a second drive.
James, great article for details. Hoping I don’t need all that…….. explains what is going on!
Confused here tho. I have a microSD [vilros or canakit standard pkg] with noobs on a pi4, and I installed raspian.
I made changes to do the usb boot, and I have raspi-config changed to boot usb.
I connected via usb 3 a 64GB Samsung SSD, and I used “disks” utility to erase partitions, so I have free space now on the SSD. I am attempting to “SD card copy” from the microSD to the SSD and am having trouble. It creates identical partitions, but never seems to be able to complete the copy. failures of one kind or another after “copying partition 1.”
The partition 1 is “recovery” and is noobs I assume. The rest is an extended partition with partitions 5, 6 and 7 as “settings” “/boot” and “/” -not sure what “settings” is…….
Does “noobs” get in the way of this?
I am assuming a boot always goes thru the noobs “recovery” partition since it gives me the option to reinstall.
I was able to mount the partitions on the SSD and see them [before I erased it all], so I presume the usb driver is working.
Thanks for any help!
john
Thank you for your guide. Unfortunately I´m running into problems.
I can change partuuid but after updating my cmdline.txt the system wont reboot. Is it possible that my USB to SSD adapter is able to detect the SSD but is not able to boot from it?
Hi,
first off thank you for the article. I set up my pi 4 a while ago and it was working perfectly. But after the recent OS updates my sytem was shot. I was not able to boot normally. So I reset and followed your guide again. Now, I am having the same issue mkapsville is having: edited boot order. Now, it is not rebooting. No I/O LEDs blinking. I followed a different guide, which helped but my sequential read/write speeds dropped in half. Tested different adapters and M2s, same issue. Were there changes to newest Raspbian that might be causing problems?
Hey Mark,
It definitely sounds like it has something to do with the new bootloader support. Edited boot order seems to point me toward this:
https://www.raspberrypi.org/documentation/hardware/raspberrypi/bcm2711_bootloader_config.md
There is a BOOT_ORDER field in the new bootloader. Depending on what this is set to it will attempt to boot from the SD card for 0x1, a USB device for 0x3, etc.
This support is new from after I wrote this guide (it didn’t exist at the time). It looks like maybe it is coming out of beta here and being released to everybody possibly!
It looks like the command to edit the bootloader is:
sudo -E rpi-eeprom-config --editThis will only work if your system is fully patched and up to date it looks like but you can see the options here:
[all]BOOT_UART=0
WAKE_ON_GPIO=1
POWER_OFF_ON_HALT=0
DHCP_TIMEOUT=45000
DHCP_REQ_TIMEOUT=4000
TFTP_FILE_TIMEOUT=30000
TFTP_IP=
TFTP_PREFIX=0
BOOT_ORDER=0x1
SD_BOOT_MAX_RETRIES=3
NET_BOOT_MAX_RETRIES=5
[none]
FREEZE_VERSION=0
It looks like it is time for me to write the guide for the new bootloader here, but give editing the BOOT_ORDER field in there a try (0x4 for a USB mass storage device, other choices are in the link earlier in the post) and see if it works for you!
UPDATE: New guide posted here: https://jamesachambers.com/new-raspberry-pi-4-bootloader-usb-network-boot-guide/
Hi, jamesachambers,
Thank you for the quick reply! Yeah, updates seemed to have rolled September which is right around the time I started having this issue. Right now I basically just reimaged my m2 ssd with the raspian imager using the newest OS build, enabled headless ssh mode, and plugged the ssd into one of the usb3 ports – no SD card involved. Works fine: I can restart, run updates etc. but I am only downloading with 29MBs/sec whereas before I was hitting 50MBs/sec. I ruled out my ISP and other potential bottlenecks (I have a 500mbit connection which I am fully getting on other devices).
I will backup my current setup and test your guide. Thank you again!
Cheers
So, I realized changing the boot order does not really help since I reset everthing and I am already able to to boot from usb port. Still do not know why my download speed is so slow though. Basically cannot use my pi as my download manager anymore.
Hey Mark,
You said this drive had recently been reimaged. Have you ran the ‘trim’ command on the drive? This may be a bit of a long shot but during my storage testing I’ve noticed this sometimes. Trim will tell the drive to actually drop all data marked for deletion/no longer used from the device. You can try it with:
sudo fstrim -avThis keeps the SSD from having to write over this old data (which takes time). When you reimage a drive the entire old image is still left on there until the trim command is run. If you want to understand more there’s a pretty good writeup of SSDs and trim here.
After this give the device a reboot and try running your test again. You may also want to consider running my storage benchmark on it. This tests your drive’s speed using standard benchmarking tools/techniques and returns a score that we can easily compare against other submissions with the same drive. We can also look at the raw numbers and see where it’s struggling (things like random IO performance vs read / write performance, etc). It’s very likely other people have tested your same drive or one very similar we can compare it to!
There are some storage adapters and even SSDs that don’t support trim out there but they’re all older. If you’re using a modern one it should be able to do it! Adapter quality can also impact performance but if you’re comparing performance to how it was from before the updates it doesn’t make sense that the adapter would be the cause in this case.
I know your overall concern is network download performance but on the Pi a lot of the busses/lanes that carry this data are shared between the different components so it definitely can matter! At least on the Pi 4 they finally separated the ethernet from the main USB bus, but even so storage is still very likely to be the bottleneck here.
If you have any additional drives you can image and run the same network test on this would narrow things down. What download rate do you get with a standard SD card for example would be a good thing to check. It could definitely end up being related to a package or a change in one of the updates in which case you should see the same slowdown on anything running the latest version.
Hi,
I am actually getting decent download speeds (300-350mbits) running speedtest-cli via terminal but big spikes.
I have also been using your benchmark tool and my score is close to 7000 on average so those parameters look fine.
I am using JDownloader as my main download tool which downloads to my DS413j. I ruled out my NAS as well by using a download folder on my pi as the destination. Maybe it is related to the JDownloader’s recent updates and its terrible Java code base…
I have not found an alternative for Raspbian though which supports filehosters, etc. My pi is passively cooled and remains under 70 degrees Celcius under heavy load so CPU/GPU throttling should not be it. Also thank you for the trim tip. Just ran your cmd. Here are my most recent raw stats.
HDParm Disk Read 293.52 MB/s
HDParm Cached Disk Read 278.41 MB/s
DD Disk Write 169 MB/s
FIO 4k random read 16370 IOPS (65483 KB/s)
FIO 4k random write 7981 IOPS (31925 KB/s)
IOZone 4k read 28309 KB/s
IOZone 4k write 23576 KB/s
IOZone 4k random read 17552 KB/s
IOZone 4k random write 24261 KB/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 15.632 ms
Download: 245.66 Mbit/s
Upload: 49.73 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 14.072 ms
Download: 299.02 Mbit/s
Upload: 49.33 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 13.416 ms
Download: 241.75 Mbit/s
Upload: 49.16 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 13.121 ms
Download: 224.04 Mbit/s
Upload: 49.46 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 13.606 ms
Download: 343.81 Mbit/s
Upload: 49.51 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 12.783 ms
Download: 213.43 Mbit/s
Upload: 50.43 Mbit/s
@raspberrypi:~ $ speedtest-cli –simple
Ping: 13.454 ms
Download: 425.74 Mbit/s
Upload: 49.09 Mbit/s
I tested a brand new WD Blue SN550 250 GB NVMe SSD in a UGREEN SSD M.2 NVMe enclosure. Same behavior.
Hey Mark,
I definitely see it fluctuating there. I gave it a few runs on my Pi and it chose different servers a couple of times. Here’s what mine looked like:
Hosted by Think Smart Inc 7.558 ms
Download: 341.58 Mbit/s
Upload: 372.93 Mbit/s
Hosted by UTOPIA Fiber 5.288 ms
Download: 342.06 Mbit/s
Upload: 364.29 Mbit/s
Hosted by UTOPIA Fiber 5.757 ms
Download: 338.92 Mbit/s
Upload: 330.31 Mbit/s
So it looks like it can fluctuate but not usually by that much. Are you wired in or using wireless? Do you have access to another internet connection or network like at work / friend / family you could take it to and run this same test to see if you see the same type of network fluctuations?
It’s also possible it’s Java related. Have you tried it on the 64 bit Raspbian beta? That would run on a 64 bit Java Virtual Machine which can definitely benefit some applications. It can also be worse in some cases since it typically uses a little more memory so it’s good for 8 GB Pis but not so good for smaller capacities.
It definitely seems like there is some sort of network performance issue going on from your speedtest-cli numbers though!
Hi,
I am wired directly into my WIFI Router (1Gbit). I only have the 4GB RAM Version and do not want to deal with Betas. I will double check my basic troubleshooting steps (reboot router, check cables, try different ethernet ports, maybe even try over wireless since my Pi sits next to the router, etc.)
Are there any JDownloader alternatives that you know of?
Hey Mark,
I’ll be honest and say I’m not super familiar with JDownloader itself. I have definitely had a ton of run-ins with Java and the JDK over the years via my Minecraft Pi server script.
I think you are on the right track troubleshooting and narrowing things down as much as possible. These types of issues can be really difficult to diagnose because something like network transfer speed has so many underlying pieces to it that have to all work together to achieve maximum speeds.
I would keep looking for alternatives if possible just as a comparison and keep trying to test different variables. I think you’re getting close here!
no other network I can try except a hotspot from my fine but that is that good reference.
*phone *not
you may add 0080:a001 ORICO mSATA SSD box 3.0 USB with UAS (MSG-U3) to the non working list as of today
https://www.orico.cc/us/product/detail/3686.html
🙁
I wonder a bit about that you seem to have posted your very good article on 19th of october this year while other sources stated about a month before that USB bootloading with the PI4 is officially supported from bootloader Version 2020-09-03?
Apart from that your list of working/nonworking adapters is highly useful of course.
Regards, jan
Hey Jan,
Sorry, that’s because it’s showing when it was last updated and that was the last time it updated the list. It’s on my to-do list to add the original published date below it but haven’t got to it yet, but you’re quite right! Thanks for the kind words on the article!
I truly appreciate this blog article.Much thanks again. Really Cool.