2020’s Fastest Raspberry Pi 4 Storage SD / SSD Benchmarks

We’ve now received over 20,000 benchmarks at https://storage.jamesachambers.com/! These are submitted by regular people from all over the world.

In celebration of that milestone as well as the launch of the 8GB Raspberry Pi 4* take a closer look at which device you should choose for your Raspberry Pi 4 as determined by science and measurement!

Best Performing Devices

Solid State Drive (SSD)

(View: AliExpress*)

The Samsung 970 Evo Plus* is currently the king. It’s not cheap, but a lot of people have it with a total of 49 benchmarks so far!

Secure Digital (SD / MicroSD) Card

(View: AliExpress*)

The SanDisk Extreme A2* is barely edging out the SanDisk Extreme A1 variant* for the top spot. The difference is tiny though, so you should pick whichever one is cheaper at the moment.

Storage Adapter Performance Note

The type of USB adapter you use (if you aren’t using a SD card) can have a big effect on performance. I cover which adapters are good and which should be avoided more comprehensively in my Raspberry Pi 4 USB Booting Post.

The most popular one at time of writing is the StarTech USB 3.1 to 2.5″ SATA adapter*.

Top 20 Fastest Storage Devices

Top 10 Fastest Secure Digital SD / MicroSD Cards

Benchmarking Your Own Device

I highly recommend benchmarking your own storage device. To run the benchmark paste/type:

sudo curl https://raw.githubusercontent.com/TheRemote/PiBenchmarks/master/Storage.sh | sudo bash

The benchmark only takes a few minutes and will give you a lot of information about whether your Pi is performing well or not. Storage is definitely the biggest performance variation between otherwise identical Pi setups.

Compare your score with others at the full benchmark browsing site https://storage.jamesachambers.com/. Leave a comment letting us know how you did or if you need any help!

17 thoughts on “2020’s Fastest Raspberry Pi 4 Storage SD / SSD Benchmarks”

  1. Avatar for TruthOrNot

    Isn’t the i/o speed to the RPi4 limited by the USB 3.0 interface? So all the benchmarks for PCI type NVMe M.2 cards are bogus for the Pi, the most throughput via USB 3.0 will be about 500MB/s on some SATA based SSD, or what am I missing?

    1. Avatar for Michal

      Exactly. USB 3.0 is the Pi’s bottleneck which none of the recent SSDs fits through, thus the difference among them for a Pi user is minimal. Please correct me if I’m wrong.

  2. Avatar for Matthew

    Hello James –

    I have a Samsung 970 EVO Plus 500GB and I’m attempting to compare benchmark numbers to your site to make sure everything on my pi is properly configured. Your benchmark script is very helpful but its unclear to me what I should be looking for. I’m expecting a score somewhere near `9563` in my report – but where?

    Thanks,
    Matt

    1. Avatar for jamesachambers

      Hey Matthew,

      I think you actually took it when storage.jamesachambers.com was down! You’re right though, it’s pretty hard to know where to find it. I have a couple of upcoming improvements coming to help with this.

      For example, I am going to make it so that when your test is over it actually gives you a link to your result! This will be a little tricky for drives it doesn’t know how to identity yet but I’ll just add something in that says it couldn’t be identified etc. In your case I think the test was successful but my site was being rebuilt at the time (database crash) so it probably didn’t make it.

      Regarding your score itself, it’s great! One thing you can use to compare it is to pull up your drive by brand name:

      https://storage.jamesachambers.com/brand/Samsung_970_EVO_Plus

      I’m still right in the middle of creating this “brand” meta page so there’s only a couple of boxes of information right now but it does have the top 5 scores ever recorded for your drive. The highest one is 12,333 and #5 is 10,200. Many of the top 5 results are overclocked so never feel too bad if you didn’t get first place if you weren’t overclocking and shooting for it.

      The other important thing to look at is the average. This is exactly what it sounds like, an average of all the scores although there is also a “Pi 4” only average since Pis below the 4 are not capable of running these drives at full speed. The average for this drive is currently 9,171 so your score was a few hundred points above average!

      If your drive was scoring off from those scores by more than 20% (especially below the “average”) it would indicate that something is not right. This usually ends up being a problem with a storage adapter/occasionally a bad drive/configuration error etc. Yours looks good to go!

      1. Avatar for Matt Kime

        Thanks for confirming.

        To help others, here’s more info on my setup –

        HOGORE USB 3.2/ 3.1 Gen 2 Hub, 4-Port Powered USB Hub (USB-C to USB-A x3 & USB-C) w/ power adapter
        XAOSUN 3.1 USB C Female to USB Male Adapter
        TDBT M.2 NVMe SSD Enclosure with Heat Sink, 10Gbps USB-C to PCIe NVMe M.2 Hard Drive Enclosure with Thermal Cooling Pad

  3. Avatar for Olivier

    I bought specifically the Samsung T5 after reading the AnandTech article and its very low consumption.
    It works great out of the box with the a 5V-3A power supply. I still need to add a fan instead of the heatsinks, nevertheless it should be fine.

  4. Avatar for Tyler sojka

    hello everyone,
    im having some issues! after looking through this guide I recently bought a western digital blue sn550 nvme 250gb, and theo tdbt m.2 nvme enclosure, got my pi 4 to boot using this set up, and ran this test. my score was 2526, which is FAR below even the lowest recorded score. ive been trying to troubleshoot why this might be and i cant seem to find a solution. all of the videos ive watched with people booting from a ssd give them hparm speeds of 100-600 mb/s, while mine is hovering around 30. any ideas?

    1. Avatar for jamesachambers

      Hey Tyler,

      Sorry for the slow reply, I missed this one in the shuffle. Have you tried enabling quirks mode on your drive? I cover this a little bit more in depth in my New Raspberry Pi 4 Bootloader USB Booting Guide. Sometimes unpredictable things can happen and it can help.

      2500 sounds suspiciously close to the USB 2.0 limit. It may be worth checking to make sure it’s plugged into one of the blue ports. I recently updated the benchmark to tell you what mode your USB connection is running in if it’s able to be seen depending on your adapter.

      I’m not familiar with this enclosure. Do you happen to have a link to your benchmark id #? I would definitely be willing to take a closer look here and see if we can figure out what is bottlenecking you!

  5. Avatar for Austin Vojta

    James — seems we can no longer download the results to a CSV or .xlsx document? Did you make that change intentionally, or will the functionality be added back at some point? I would like to sort through results on my own.

    1. Avatar for jamesachambers

      Hey Austin,

      Wow! Did you actually use the benchmark back then or is there a reference to that somewhere still around?

      So it wasn’t an intentional change. What happened is I have been migrating away from a PHP forms submission software called “FormTools” to purely my own code. That project seems to have been abandoned. The author has a note up and the site is still up but it hasn’t been updated for years. It’s still partially used right now until I can finish the rest of my own custom stuff. The benchmarking bash script still uploads the results to a FormTools form.

      The site used to basically be a giant table that had limited sorting and was impossible to navigate very effectively. The “export” function was built into FormTools already. I’ve replaced the FormTools frontend entirely though at this point so there’s no way to get to the button for non-admins. At the time it was literally easier to read the data in Excel (most things are really when presented/sorted, but I mean you basically had to to even read it). I had actually forgotten about it until you mentioned it!

      There’s no reason I took it away or anything like that so I will definitely put it on the ToDo list to add an export button here at some point. I’m right in the middle of major changes on the site trying to get it more usable (it’s better, still horrible, but improving and you can kind of browse it now, yay!). If you check it today and check it tomorrow it may look completely different or have new pages than before so I may get an opportunity to implement this before too long!

  6. Avatar for Jim "JR"

    Greetings!

    Two questions:

    1. I have been running your benchmark and have submitted numerous reports on various SD and USB devices I have here, all as (using “European” quotes for clarity), and yet, interestingly enough, none – not a single one – has shown up on your list. I did a new test of a Seagate Expansion SSD (500G) on my Pi-4 test rig and submitted it on 2020-10-14, and it has not yet shown up. Nor bustin’ chops, but I am curious. Maybe other significant data is also being swallowed up in some black hole somewhere?

    2. When running tests on my devices, I have noticed that several tests on the same device, on the same port (USB/micro SD), on the same system, using an identical image, show what might be significant variation.

    For example:
    Today I purchased the Seagate Expansion SSD device – an external USB-3 device that contains a SSD instead of a standard mechanical hard drive.

    I ran a total of ten passes of your test on it and recorded all the results. (I turned off my WiFi at the end of the test to avoid spamming your site with tests.)

    The “score” for this device, run exactly the same way over and over again in succession, ranged from a low of 6477 to a high of 7394 – a difference of almost exactly 12.5%.

    I have noticed a similar effect when testing other devices though I have not run such a test as this because the other devices were too slow.

    I would hope to believe that the device itself doesn’t have a 12 to 13% variation in speed based on the phases of the moon or whatever. I also shudder to think that my Pi-4 is so inconsistent.

    Do you have any idea why there should be such a variation? Note that the variation seems to be random in nature instead of steadily increasing or decreasing.

    Thanks!

    Jim “JR”

  7. Avatar for Patrick

    This is great! FYI, the reason that that particular micro-SD card at the top is so popular is because it is sold in a 2-pack from Costco in the US for only $45, making it out to $22.50 per card! So they are popular because they are cheap!

  8. Avatar for DerFritz

    Hi James,
    thanks for providing such very deep information.
    Over the time I have done several tests with Raspi 4 (4GB) and Samsung 860 EVO 250 GB (e.g. #5367 on 2019-12-18 and #32570 on 2020-09-28).
    I have done firmware and bootloader updates as soon as they where in stable state. I believe my updates have made the system more stable and efficient due to less heat dissipation.
    On the other hand during time the score went worse.
    I have done software updates of your software too in order to stay comparable.
    What do you think about? What may be the root cause of this slow down?
    Regards, H.

    1. Avatar for Jim "JR"

      Assuming you’re using an SD card, mounted within the Pi’s SD slot, have you tried a TRIM of the device?

      Viz.: fstrim -v /

      Part of the problem may be that you are continually reading and writing from the device and solid-state devices like SSD’s or SD cards do not necessarily handle TRIM very well, especially if you are running something repeatedly.

      You can find more information here:
      https://www.jeffgeerling.com/blog/2020/enabling-trim-on-external-ssd-on-raspberry-pi

      1. Avatar for jamesachambers

        Sorry I’m late to the party, but I second what Jim said. When I’ve seen this on my own devices it has been because the SD card needed a trim. Once I did that and rebooted typically performance would be 90-95% restored if not 100%!

      2. Avatar for DerFritz

        Hi Jim,
        thanks for the hint.
        This is not a SD card. The Samsung 860 EVO is a mSATA device on a Lycom Adapter (in a USB3-version with controller from Renesas/NEC). Anyway, the provided link worked.
        However, after trimming the device is not faster.

Leave a Comment

Your email address will not be published. Required fields are marked *

Type here..