Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Odroid-H3

hardkernel.com

Post reply on HN
Print view View original

71–80 of 120 posts

Re: Odroid-H3

#71

Post by sliken »

ThrowawayR2 wrote: post #2

Nice to see a successor to the H2/H2+ finally. > Power consumption: IDLE : ≃1.9W Better idle power than a Raspberry Pi 4 plus it has functioning suspend modes.

As someone else mentioned it's 1-9W not 1.9W for the H3 and H3+.

I don't particularly trust the numbers, generally the N6005 runs hotter than the N5105, which makes sense since it's the same process and largely the same chip with more GPU EUs.

ServeTheHome mentions VERY similar boxes (same CPUs) "At idle, we saw between 10-12W depending on single versus double SODIMM configurations. Maximum power consumption hit just over 30W. Both at idle and at maximum, this is significantly higher than the J4125, but one also gets more performance, so there is a clear trade-off. At the top end, adding ~5W over the N5105 was unexpected based on similar TDPs."

On the N5105 (they reviewed at least two) "Power was not a winning point of this solution. At idle, we saw between 10-12W depending on single versus double SODIMM configurations. Maximum power consumption hit 24.5W even with the Topton configured (512GB/16GB) unit. There is clearly room to go up from there. This is a lot higher on both idle and maximum power consumption than the Celeron J4125 that we looked at previously."

Re: Odroid-H3

#73

Post by flatiron »

StephenSmith wrote: post #43

Does anyone know how this compares to something like the NVIDIA Jetson Nano? https://developer.nvidia.com/embedded/jetson-nano-developer-...

Different use cases. Jetson is for open CL. This is a SBC for general server/light desktop usage.

Re: Odroid-H3

#74

Post by toast0 »

FloatArtifact wrote: post #72

I wish they would have up to eight at 6 to 8 SATA ports. Then it could be a proper nass.

I've seen some pci-e m.2 5x sata boards. That would get you to 7 ports. You can boot from emmc (i assume) if you don't want to boot from your array.

Re: Odroid-H3

#75

Post by eneumann »

Word of caution to anyone thinking of using this for pfsense: the realtek 2.5G drivers are not included in the pfsense image and have to be installed manually. A real pain if you don't have easy access to another freebsd machine.

Re: Odroid-H3

#76

Post by toast0 »

devonkim wrote: post #13
capableweb wrote: post #5

Earlier quoted context omitted.

Do those come with two ethernet ports, two DDR4 slots (up to 64GB), eMMC, M.2 and two SATA ports? For me, this almost looks like an instant buy.

Not just gigabit or 100 mbit Ethernet ports either but 2.5 gbps and Realtek at that (ironically people have been complaining more about Intel NICs than Realtek recently but that’s anecdotal still IMO). There’s also other factors like GPIO pins as well. This is a board better suited for relatively higher performance and power efficient embedded projects by far than what most OEMs will produce.

> ironically people have been complaining more about Intel NICs than Realtek recently but that’s anecdotal still IMO

I'm not sure if Realtek is getting better, I've heard a lot of issues with Intel 2.5G nics; people say some revs of the hardware are unusable at 2.5G (but do work fine at 1G). Personally, I've only got good things to say about Intel 1 and 10G nics I've worked with.

For me, Realtek 1G works ok until it doesn't, it's possible to overwelm the hardware with traffic and then you get bad behavior --- queues get stuck or seem to get stuck and the queue processing doesn't always fully reset when you tell it to --- this can lead to wild writes, following descriptors that the OS thought were dead. Sometimes better results with Realtek's drivers than open source drivers, but they frob the hardware with crypticly named defines, and I've seen them send pause frames (which nobody wants) with no way to disable. On one machine, using Realtek's drivers results in immediate kernel panics, so it's not universally great. The interrupt design is poor anyway --- MSI-X dates from 2002, they could have used a separate irq for rx, tx, and misc, rather than a single interrupt and a status register. But that might be ok if the status register worked properly. It's too easy to have timing issues when poking the status register and then you don't get interrupts you should. I haven't touched their 2.5g hardware, hopefully they fixed some stuff.

Re: Odroid-H3

#77

Post by moffkalast »

eqvinox wrote: post #66
moffkalast wrote: post #20

Earlier quoted context omitted.

Not only that, but after some digging it seems that the GPIO header is actually not half bad either despite it being somewhat small: https://wiki.odroid.com/_detail/odroid-h3/hardware/h3_24pinm... Supports i2c, UART, USB, and hopefully manual pin control too. That's already miles ahead of a NUC in terms of interfacing.

> Supports i2c, UART, USB, and hopefully manual pin control too. From the set of pins provided I wouldn't get any hopes up for manual pin control. All pins are single-function, USB pins are almost never switchable (because they're high-speed analog-ish PHY functions), and UART & I2C are not GPIO-capable functions on most x86 platforms.

Well that sucks if true, but at least an i2c GPIO expander/servo board is not out of the question I guess.

Re: Odroid-H3

#78

Post by Klasiaster »

Svenstaro wrote: post #59
Klasiaster wrote: post #46

Who cares just about plain HW specs? Nowadays I also expect the info on whether it supports UEFI booting, whether the firmware gets updates (and if fwupd or not), and if mainline Linux can boot and from what version. Too often these products have a limited lifetime (and are too hard to use) due to lack of the above.

I'm running the H3+ right now and I can tell you that I'm booting plain UEFI with a plain Arch Linux and no special stuff required at all. The firmware received 2 updates so far in 3 months.

Sounds better than the other boards, they could have written this stuff as selling point.

Re: Odroid-H3

#79

Post by tbyehl »

sliken wrote: post #71
ThrowawayR2 wrote: post #2

Nice to see a successor to the H2/H2+ finally. > Power consumption: IDLE : ≃1.9W Better idle power than a Raspberry Pi 4 plus it has functioning suspend modes.

As someone else mentioned it's 1-9W not 1.9W for the H3 and H3+. I don't particularly trust the numbers, generally the N6005 runs hotter than the N5105, which makes sense since it's the same process and largely the same chip with more GPU EUs. ServeTheHome mentions VERY similar boxes (same CPUs) "At idle, we saw between 10-12W depending on single versus double SODIMM configurations. Maximum power consumption hit just…

I suspect that H3/H3+ numbers are on the DC side with an eMMC module, a single DDR4 SO-DIMM of the lowest capacity, with the most aggressive power management settings, and nothing else attached. 1.9w idle is plausible but not what people are going to see with a realistic running configuration measured from the AC end.

Re: Odroid-H3

#80

Post by sliken »

jesperwe wrote: post #7

THANK YOU a thousand times for providing actual power consumption specs! (Sorley missing from most SBC product pages...)

As mentioned elsewhere, it's 1-9W not 1.9W idle. Other similar products idle at 10-12 watts with the same CPU.
Post reply on HN