4 ms·

It also uses a lot more power all of the time for something that will be on 24/7. There is a trade off between these devices, the pi is more expensive to kit eq

It also uses a lot more power all of the time for something that will be on 24/7. There is a trade off between these devices, the pi is more expensive to kit equivalent to the N100 machines (which are a bit quicker) but it uses less power all the time especially under load.

magicalhippo 2y ago
Of course if you're powering this off solar in some remote location then we're into the niches where an RPi 5 might make sense, but otherwise it's just down to electricity cost. How that works out in practice depends on workload though, while the N100 consumes more at idle, it can also finish workloads much faster so can potentially spend more time at idle. While the RPi 5 idles at around 4W (including a NVMe) and the N100 at around 15W, the RPi 5 uses 12W at full tilt[1], ie close to N100 idle power draw. Alternatively the N100 with 32GB of RAM can replace two or more RPi 5's in terms of performance, so in that regard might even come out ahead. [1]: https://github.com/geerlingguy/sbc-reviews/issues/21#issuecomment-2579400396 https://github.com/geerlingguy/sbc-reviews/issues/21#issueco...
fy20 2y ago
The cost difference in power probably works out to much less than $50/year. Modern PCs are very power efficient when idle.
- addicted 2y ago
  $50/yr means you make up the entire cost of the RPi in a little more than 2 years if you’re running it as a home server of some sort, for example.
- PaulKeeble 2y ago
  Based on my current unit prices the difference in idle performance alone would come out to about £30 a year, so you would get the difference in price back within the life of the device. Europe has much higher KW/h prices for power especially in the past few years and is more advanced in transition to green power technologies and as such we tend to care more about the cost of power than those in the USA.
formerly_proven 2y ago
This probably depends more on what hardware is being used rather than the SoC itself. A bare RPi5 seems to idle around 3 W, which a bare N100 can absolutely (out)do (these are designed with S0ix/s2idle after all). If you add gigabit ethernet, you'll add 0.5 - 2 W (depending on the controller). If you add NVMe, you'll add 0-4 W - NVMe SSDs vary wildly in both their own power consumption and how they interact with CPU C-states and ASPM. Some SSDs prevent low C-states and thereby increase CPU power consumption a lot, for example (even in S0ix). This is generally true for every PCIe peripheral (including network controllers), but NVMe SSDs are popular troublemakers in this area. I honestly don't believe any of the "N100 idles at 15 W" numbers. First of all, that substantially exceeds the power limits of the N100, by like a factor of three. So clearly the vast majority of that power isn't being dissipated by the N100. Seeing how N100 boards generally have only one heatsinked component, the power is dissipated elsewhere. Second, people rarely post their exact hardware and how they measured this. "Idles at 15W [because there's a spinning hard drive attached]" is not very interesting. Third, many of the N100 boards have ATX power connectors, and if you use any old ATX power supply, that alone can cause such a number. Fourth, if you're using a cheap power meter, many of them are still wildly inaccurate at the low end. And, fifth, as a reality check, even much older 1L PCs using actual desktop platforms, even with separate chipsets and all that jazz, don't idle at 15 W. Unless you're using Windows, then, maybe. But Windows can't and shouldn't be the yard stick for power efficiency.
- poulpy123 2y ago
  My minipc idle (with nothing added) idle at 6W and works at 22w
- geerlingguy 2y ago
  From the 5 N100 systems I've run, they all idle between 5-8W, PSU inclusive (measured at the wall). The SoC might idle slightly lower, but I compare total system power draw when I test SBCs since that's the only fair comparison.
  - magicalhippo 2y ago
    Now you made me want to hook up my N100 box to my lab PSU so I can get some accurate numbers of what it does and how it compares to the wall.

magicalhippo 2y ago

Of course if you're powering this off solar in some remote location then we're into the niches where an RPi 5 might make sense, but otherwise it's just down to electricity cost. How that works out in practice depends on workload though, while the N100 consumes more at idle, it can also finish workloads much faster so can potentially spend more time at idle. While the RPi 5 idles at around 4W (including a NVMe) and the N100 at around 15W, the RPi 5 uses 12W at full tilt[1], ie close to N100 idle power draw. Alternatively the N100 with 32GB of RAM can replace two or more RPi 5's in terms of performance, so in that regard might even come out ahead. [1]: https://github.com/geerlingguy/sbc-reviews/issues/21#issuecomment-2579400396 https://github.com/geerlingguy/sbc-reviews/issues/21#issueco...

fy20 2y ago

The cost difference in power probably works out to much less than $50/year. Modern PCs are very power efficient when idle.

addicted 2y ago
$50/yr means you make up the entire cost of the RPi in a little more than 2 years if you’re running it as a home server of some sort, for example.
PaulKeeble 2y ago
Based on my current unit prices the difference in idle performance alone would come out to about £30 a year, so you would get the difference in price back within the life of the device. Europe has much higher KW/h prices for power especially in the past few years and is more advanced in transition to green power technologies and as such we tend to care more about the cost of power than those in the USA.

formerly_proven 2y ago

This probably depends more on what hardware is being used rather than the SoC itself. A bare RPi5 seems to idle around 3 W, which a bare N100 can absolutely (out)do (these are designed with S0ix/s2idle after all). If you add gigabit ethernet, you'll add 0.5 - 2 W (depending on the controller). If you add NVMe, you'll add 0-4 W - NVMe SSDs vary wildly in both their own power consumption and how they interact with CPU C-states and ASPM. Some SSDs prevent low C-states and thereby increase CPU power consumption a lot, for example (even in S0ix). This is generally true for every PCIe peripheral (including network controllers), but NVMe SSDs are popular troublemakers in this area. I honestly don't believe any of the "N100 idles at 15 W" numbers. First of all, that substantially exceeds the power limits of the N100, by like a factor of three. So clearly the vast majority of that power isn't being dissipated by the N100. Seeing how N100 boards generally have only one heatsinked component, the power is dissipated elsewhere. Second, people rarely post their exact hardware and how they measured this. "Idles at 15W [because there's a spinning hard drive attached]" is not very interesting. Third, many of the N100 boards have ATX power connectors, and if you use any old ATX power supply, that alone can cause such a number. Fourth, if you're using a cheap power meter, many of them are still wildly inaccurate at the low end. And, fifth, as a reality check, even much older 1L PCs using actual desktop platforms, even with separate chipsets and all that jazz, don't idle at 15 W. Unless you're using Windows, then, maybe. But Windows can't and shouldn't be the yard stick for power efficiency.

poulpy123 2y ago
My minipc idle (with nothing added) idle at 6W and works at 22w
geerlingguy 2y ago
From the 5 N100 systems I've run, they all idle between 5-8W, PSU inclusive (measured at the wall). The SoC might idle slightly lower, but I compare total system power draw when I test SBCs since that's the only fair comparison.
- magicalhippo 2y ago
  Now you made me want to hook up my N100 box to my lab PSU so I can get some accurate numbers of what it does and how it compares to the wall.