Raspberry Pi 4 vs ODROID-C4 Features Comparison

Yesterday, Hardkernel launched ODROID-C4 Amlogic S905X3 SBC as an update to ODROID-C2 SBC launched in 2016. We’ve seen the board can compete with Raspberry Pi 4 4GB board thanks to benchmarks released by Hardkernel, but since those boards are so versatile, meaning they can be used in a variety of applications, it’s impossible to benchmark all use cases.

One way to find out which board might be right for your application without going through benchmarks is to look at a list of features for each, and that’s exactly what we’ve done in this post.

ODROID-C4 (left) vs Raspberry Pi 4 (Right)

Let’s get straight to the Raspberry Pi 4 (4GB RAM) vs ODROID-C4 comparison table.

Features/Specs Raspberry Pi 4B (4GB)
ODROID-C4
Release date 24th June 2019 23rd April 2020
SoC Broadcom BCM2711
quad-core Cortex-A72 @ 1.5 GHz (overclockable to 2.0+ GHz) with NEON, FPU
Amlogic S905X3 quad-core Cortex-A55 processor @ 2.0 GHz with NEON, FPU, Armv8 Crypto extensions
GPU VideoCore VI with OpenGL ES 1.1/2.0/3.0, Vulkan 1.x (WiP) Mali-G31MP GPU with OpenGL ES 1.1/2.0/3.0, Vulkan 1.0 and OpenCL 2.0
Video Decode H.265 4Kp60, H.264 1080p60 H.265 4Kp60, VP9 4Kp60, H.264 4Kp30
Video Encode H.264 1080p30 H.265/H.264 1080p60
Memory 4GB LPDDR4 4GB DDR4
Storage microSD card microSD card + eMMC module connector
Video & Audio Output 2x micro HDMI ports up to 4Kp60 with CEC, HDR (later)
3.5mm AV port (composite + audio)
MIPI DSI connector
1x HDMI 2.0a port up to 4Kp60 with CEC, HDR (now)
Audio header with I2C & S/PDIF
Camera MIPI CSI connector N/A – USB camera only
Ethernet Gigabit Ethernet
WiFi Dual-band 802.11 b/g/n/ac N/A, USB dongle only
Bluetooth Bluetooth 5.0 + BLE N/A, USB dongle only
USB 2x USB 3.0 + 2x USB 2.0
4x USB 3.0
Expansion 40-pin GPIO header
Power Supply 5V via USB type-C up to 3A
5V via GPIO header up to 3A
Power over Ethernet via PoE HAT
5.5V to 17V via 5.5/2.1mm DC jack
Power Consumption with minimal OS Idle: 2.10 Watts (Source)
CPU stress: 4.36 Watts (Source)
Idle: 1.89 Watts
CPU Stress: 3.11 Watts
Cooling N/A, 3rd-party heatsink or fan recommended Ships with heatsink
Dimensions 85×56 mm
Official OS(es) Raspbian Buster
(Debian 10 based)
Ubuntu 20.04
Android 9.0 Pie
Support Documentation & Forums Wiki & Forums
Price $55 $50+

As you can see from the table above each board has its strengths and weaknesses. For example, if you need to connect two displays to the board, Raspberry Pi 4 is the obvious choice, but for the best multimedia experience with 4K HDR you may want to use ODROID-C4 with Android 9.0 Pie, although it’s also working with CoreELEC on C4.

Note that price comparisons are always challenging because the final price depends on taxes and shipping as well as whether you already own some of the accessories (power adapter, MicroSD card, WiFi dongle…).

The features may also evolve over time as software support improves (e.g. HDR), and it’s possible some hardware features do not work perfectly depending on the selected operating system. Both boards have active communities, and good software support including Linux mainline (WiP), although the Raspberry Pi community is much larger.

Share this:

Support CNX Software! Donate via cryptocurrencies or become a Patron on Patreon

68 Replies to “Raspberry Pi 4 vs ODROID-C4 Features Comparison”

  1. As a Linux noob, what matters to me most is the quality of documentation and forums. Raspberry’s are both rather good (though a bit messy especially with Pi 3 stuff not always being applicable to the Pi 4).

      1. Over there you never know who teaches whom between users and engineers, certain responses and suggestions are frightening!

      2. Try to be fair. Millions (note the plural) of Pi-4 s are in the field. The pi forums are community forums and are as good as their users – some of them are trolls. That said, nearly every question is multiple times answered by different users (with very different skills) some of them are experts.

        1. Yes but it’s like everywhere in real life: the most vocal ones are often the least informed ones and the ones spreading the most inappropriate solutions. Those coming with too good solutions are sometimes asked to shut up by the forum moderators.

    1. Other problem is no RTC module. It’s a big no for me. In places when your internet connection is abysmal and electricity goes on and off in days, what to expect with that RPi. Has to use RTC shield like ds3231 is a PITA. Why not use something like XU4 or N2 that you can just put the battery?

      BTW, that C4 too must use RTC shield (no module RTC onboard). A major PITA for me (I’m generally using them for web server)

      1. I second that, for having added RTC by soldering stuff into about all of my SBCs in production at home!

  2. The problem with odroids are ancient kernels like 4.9. (still supported but ancient – december 2016 + fixes only until today).

    1. I’ve seen BayLibre guys getting ODROID-C4 boards and the plan is to have Vanilla kernel 5.7 run on the board. That does not mean all features will be supported (thinking about GPU/VPU here).

    1. No, that’s what is mentioned on their site, 5.5 to 17V DC input. They sell a 12V/2A adapter for it however.

  3. raspberry pi 4 SOC has crypto extensions
    but you will need to use proper 64bit distro for example ubuntu 19.10 or newer
    raspbian is still 32bit and they don’t use crypt extensions I think

    1. I can assure you that RPi4 does NOT have crypto extensions, which still makes it lag very far behind about all competitors. That might even be the main reason RPi is not that much interested in migrating to 64b:

      1. > That might even be the main reason RPi is not that much interested in migrating to 64b

        Booting an RPi 4 with an upstream aarch64 mainline kernel lscpu also shows ugly stuff like this:

  4. one more note comparision
    raspbery pi has open source working drivers in mesa an
    odroid c4 has panfrost mesa drivers in heavy development
    both has close source arm/broadcom drivers

  5. I’d rather get Odroid N2. When you factoring in case, emmc, power supply, and shipping cost, the total cost is not that great. Sadly, N2 still have usb problem.

    Hopefully, with the C4 launch, will be helping other project like balbes armbian for s905x3 stb .

    1. > N2 still have usb problem

      What is different to C4 other than the GL3523 SuperSpeed hub on the N2 being replaced with the much better VL817 on the C4?

      1. You can upgrade VL817 by native tool, while GL3523 needs some Win environment (binfmt-support for .exe), AFAIK.
        Taiwan’s Via Labs 817 is ~US$0.3 below New Taipei City GenesysLogics. GL provides better insight (compared to FPG’s subsidiary) into power consumption within their datasheet.
        info-tools: https://www.usb.org/usb32tools

        1. ODROID N2 is based on Amlogic S922X and C4 is based on Amlogic S905X3. Both SoCs are pin compatible and I would assume share the same USB3 IP block.

          Quoting @tonny: ‘N2 still have usb problem’. Why should the C4 not be affected by the same issue?

          1. Depends on usb problem? Signaling quality degradation, powering failures, electrical interference, attenuation, timing or interrupt delays, ahb-apb bridge related, shared caching, bandwidth, peripherals, firmware or temperature induced problems?
            Cortex-A55 has 50% improved latencies on core integrated L2-cache compared to A53’s shared L2-cache (and 1/3 faster L1d cache). No concurrency switching between core clusters on S905X3 compared to S922X’s bigLittle and comparable top cpu frequencies and memory speeds. C4-S905X3 is newer gen3 AmLogic device, too.

          1. For 2 days, I cannot access their forum. Will go there when connection restored. Thanks.

  6. A big difference is the (claimed) Secure boot support for the Odroid. Getting Secure Boot on the Raspi is impossible, there is always a gap between power-on and U-Boot where SW can simply be swapped.

    1. Ya its almost as if Raspi are just going thru motions as they know Asians are cutthroat and aRaspi dont wanna go there, they prefer to milk it and promote use in schools. Its nice to see someone up the ante, make Raspi sweat a little so the next rev wont suck as bad.

      1. Quite frankly, I find the RPi4 the first *respectable* board in this family. I’m not saying it’s good, it’s just respectable, which is already a big step forward compared to the previous ones. And I really don’t think such benchmarks can have *any* effect on future RPis design, nor on their fanboys’ choices since they always buy the early batches for the time needed to debug them, and are usually happy after hacking with fans or soldering wires to get them to run more or less stable.

        No, it just seems that RPi makes boards around unsold surplus of outdated Broadcom chips. Having managed to catch the only A72 on earth lacking crypto extensions deserves a prize!

        1. @Willy Speaking of Broadcom.

          ” Broadcom bows to EU pressure

          Broadcom has made an offer to the EU to get it to drop its anti-trust investigation, reports Reuters.

          In return for ending the investigation into its sales practices, Broadcom has said it is willing to drop ‘exclusivity deals’ whereby TV and modem customers who buy more than 50% of their requirement from Broadcom get rebates and special treatment.

          “In these uncertain times, we welcome the opportunity to avoid protracted litigation and to resolve the investigation without recognition of liability or the imposition of a fine,” said Broadcom.

          The EU issued an interim order last June telling Broadcom to drop the practice. The fine for the breach could be 10% of Broadcom’s turnover. “

        2. > it just seems that RPi makes boards around unsold surplus of outdated Broadcom chips

          I don’t think so. Back in 2011 they started not with an ARM solution but with a VideoCore SoC with one crappily integrated ARM core. Then there was the ‘backwards compatibility’ mantra and so they ended up even in 2019 not being able to use a true ARM design but again a VideoCore with some ARM cores tacked onto.

          I would believe they simply have to take the VideoCore SoC that best fits their needs and deal with the limitations, see for example: https://github.com/raspberrypi/documentation/issues/1217#issuecomment-538081361

          And I agree: for me personally the RPi 4 is the first Raspberry that is not a total disaster. But of course it always depends on the use case.

  7. IMO the comparison misses that the C4’s choice of powering will result in providing stable 5V to USB3 connected consumers like external disks. A lot of the annoying ‘USB storage hassles with SBC’ are the result of voltage drops on the USB ports.

    1. Absolutely! I really like it when boards support a wide range of input voltages like this. You can connect whatever PSU you have in your junk box and expect it to work reliably. Also it allows to build a very cheap UPS using those blue “12V lithium batteries” that are found everywhere for $8-15.

    2. Being an engineer, I really appreciate the new power design that HK has used on the N2 and now the C4. They don’t try to use the power supply as a reference voltage like the Rpi boards and many other SBCs do. They also use a connector that is much more reliable and can handle higher currents safely–5.5mm barrel jack vs micro-USB. But, best of all, they use a buck power supply to generate their 5V and that allows them to be able to accept a wide range of input voltages–5.5V to 17V.

      All of that is exactly what you want from a good power design. The only issue I have with it is that they didn’t use a chip that would allow 20V input. That would have allowed these boards to use cheap/surplus/leftover 19V laptop power bricks. But that’s not much of a complaint considering how well they did on all of the other aspects.

      To underscore why this matters–going back to the Rpi3, the micro-USB connector *could not* supply a high enough voltage to keep the processor from throttling if you put a high CPU load on it. To be clear, the same 5V supply when connected via micro-USB throttled while when connected over the GPIO pins, did not throttle.

      1. You seem to know what you are talking about. What is your opinion on power via USB Type-C on the Raspberry Pi 4? I rather like the aesthetics of that plug and the shape and size of the official Raspberry Pi 4 power supply. Unfortunately I find most other SBC and NUC PSUs so ugly that they keep me from buying those devices.

        1. > USB Type-C on the Raspberry Pi 4?

          RPi 4 is not USB PD compliant so it’s just a ‘dumb’ 5V input there (with USB PD for higher load scenarios PSU and device can negotiate 12V or even 20V and the device uses then ‘a buck power supply to generate their 5V’). So if the 5V input voltage drops then the 5V available on USB ports will drop even more.

          Asides that USB-C is a huge improvement over crappy Micro USB (8 instead of 2 contacts inside the jack) and RPi Trading Ltd.’s USB-C PSU is a solid choice for a dumb 5V PSU (I bought several of them to be used with NanoPi M4 and NEO4).

          When David talked about RPi 3 and the usual ‘powered by Micro USB sh*t show’ he mentioned throttling but I would assume he was talking about ‘frequency capping’ instead. The primary OS on every RPi (an RTOS called ThreadX) monitors input voltage and if this drops below 4.65V or something like that a GPIO is triggered and various clocks are instantly limited (cpufreq down to 600MHz for example while a secondary OS like Linux still reports it would run at 1.5GHz).

          1. Thomas is correct, the ThreadX monitor setting the CPU frequency to 600MHz when detecting a ‘brown out’ via micro-USB delivered power is what I observed.

            To Horst’s question about USB-C: it is a better connector as Thomas says. I do find it dissapointing that they chose not to make it a USB-PD compliant device. That would have opened up the option for it to have a much more stable power delivery system. Sadly, they not only chose to not take that option, but they messed up their USB-C jack wiring.

      2. I imagine the 5.5V to 17V range might be intended to allow operation directly from the auxiliary circuit in most cars.

        1. That may be so. I still wish they had upped the specs of the parts just a tiny bit so that we could use cheap 19V laptop bricks. Still, what they have moved to is a vast improvement over what the industry has been doing for way too long.

          1. Very likely the regulator chip’s margin would allow this to work anyway. Just put two 1N4004 diodes in parallel to slightly lower the voltage, maybe double them to drop 1.2V and that would be fine. But usually DC-DC buck regulators with a wide input voltage range do have quite some margin, at least to absorb transient spikes. Ah I just found on the schematic, the chip is a sy8120b1abc whose standard input voltage is 4.5 to 18V, and the abs max is set to 19V. So at least the 18V adapters are riskless, only the 19V are a bit border-line.

          2. It’s not indicated but anyway, it cannot be 16V, so it must be at least 20 or 25.

  8. I’m already sick of waiting for HDR on my almost 1 year old RPi4. The only reason I bought it was to be a TV box with HDR and atmos and 4k. HDR is nowhere near (and I always ask and check forums). Gonna buy myself the latest Shield and that’s all.

    1. If you just want to use C4 for it’s encoder (so with android I guess), You’ve best served with A95X F3 or others s905x3 stb. With ~$44, you get same processor, good casing, 4GB RAM (albeit DDR3), 64GB ROM, wifi 2.4 and 5 GHz.

      But if you planned to dev with C4, then yes, get it.

        1. There’s other models that have DDR4 and/or gigabit eth (like X96max). Just browse gearbest or aliexpress. There’s few stb that suit your need.

  9. The Vulkan on RPI 4 is pretty much a WIP and given that there are no info on the progress since January then a slow one to it. I wouldn’t put that in the table as of yet, as some devs could be disappointed about it when they buy the RPI 4 and can’t use Vulkan.

    1. I noticed as well. Maybe he’s too busy reproducing Odroid’s embarrassing benchmarks which show how slow an RPi4 can be compared to lower-spec’d competition, and trying to optimize it thinking “it’s impossible, there must be something wrong with the benchmark”. At least I hope it’s just this an he’s OK.

  10. I have a C4 arriving in two days. I also have a 2GB Pi4. If anyone wants any tests run on them, please let me know. The Pi4 is stock–no extra cooling, etc. I don’t intend to add any. If they wanted it to have a better thermal solution, they should have built it in.

      1. I will try. I’ll have to hook them both to a display. I’ll only have a 1080p available to test with, I hope that’s representative. Come to think of it, I could try them on a 4K as well. Might be interesting to see the difference the scanout BW makes on these and other tests.

  11. So I did my common x264 benchmark
    60s video in PAL(720×576)
    Pi4 on 2.1GHz 24.2 fps
    C4 on 2,0GHz 11.45 fps
    N2 on 2.0Ghz 26.5 fps

    so out of order is 2x times faster in this case

Leave a Reply

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