On Pentium Up using only 5A power, what is maxium cpufreq?

kveroneau

Hello, so when I originally purchased my Up Gateway box, I completely forgot to also purchase a PSU with it... Locally I could only find a PSU at 5V 5A, however the recommended Amps is 6A, while most people say it does work with 5A as long as you do not connect too many devices to Up board. Since this is only going to be an Internet gateway/firewall device, it won't require many devices besides two Ethernet cables. It has been working perfectly with 5A, until today when I decided to install VirtualBox, and it rebooted itself once I began compiling the kernel module required for it. Compiling always drives up the CPUs to their max due to their nature... After figuring out the low amperage was the issue during the compile, I decided to look into cpufreq to see if I can keep the CPU's max frequency low enough so that it can still function perfectly with only 5A. Here is what I did with cpufreq, and tested that it indeed does not go higher than the max I provided:

root@gateway:~# cpufreq-set -c 0 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 1 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 2 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 3 -d 800Mhz -u 1.7Ghz

Using "cpufreq-info" during the module compile process again showed that it remained stable, and the CPU did not go over 1.7Ghz. However, I am wondering if anyone else has tested this out, and figured out a sweet spot frequency to use here. I should note that even when not doing a compile, the CPU frequency of a single core was able to go up to 2.5GHz with no problems, issues seem to arise when all the cores are pinned at 2.5Ghz for too long. It can technically survive at 2.5Ghz for a few seconds without dying, although this is probably not good for the processor itself... I am wondering if something else might have been taking additional power during the compile when the cores were pinned at 2.5Ghz, as I only experienced this compile-based reboot of the unit just once. To be safe, I'll be keeping all the cores at 1.7Ghz until I can figure out a good sweet spot for 5A power.

So in case you did not want to read my entire store, the question is, does anyone know of the maximum core frequency for the Up Pentium board to run stable without using more than 5A with only 2 Ethernet cables being used. No display or USB devices are plugged in. Cooling is done passively using a mega head sink provided by the Up Gateway system kit.

ccalde

Hi @kveroneau ,

As 2.5GHz is the hardware limit for the frequency, I think it should be good enough for your application to set the freq at 1.7GHz.

But, after all, the best way is to test your setup during some time to be sure your system could work properly.

Cheers!

eduncan911

@kveroneau said:
Hello, so when I originally purchased my Up Gateway box, I completely forgot to also purchase a PSU with it... Locally I could only find a PSU at 5V 5A, however the recommended Amps is 6A, while most people say it does work with 5A as long as you do not connect too many devices to Up board. Since this is only going to be an Internet gateway/firewall device, it won't require many devices besides two Ethernet cables. It has been working perfectly with 5A, until today when I decided to install VirtualBox, and it rebooted itself once I began compiling the kernel module required for it. Compiling always drives up the CPUs to their max due to their nature... After figuring out the low amperage was the issue during the compile, I decided to look into cpufreq to see if I can keep the CPU's max frequency low enough so that it can still function perfectly with only 5A. Here is what I did with cpufreq, and tested that it indeed does not go higher than the max I provided:

root@gateway:~# cpufreq-set -c 0 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 1 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 2 -d 800Mhz -u 1.7Ghz
root@gateway:~# cpufreq-set -c 3 -d 800Mhz -u 1.7Ghz

Using "cpufreq-info" during the module compile process again showed that it remained stable, and the CPU did not go over 1.7Ghz. However, I am wondering if anyone else has tested this out, and figured out a sweet spot frequency to use here. I should note that even when not doing a compile, the CPU frequency of a single core was able to go up to 2.5GHz with no problems, issues seem to arise when all the cores are pinned at 2.5Ghz for too long. It can technically survive at 2.5Ghz for a few seconds without dying, although this is probably not good for the processor itself... I am wondering if something else might have been taking additional power during the compile when the cores were pinned at 2.5Ghz, as I only experienced this compile-based reboot of the unit just once. To be safe, I'll be keeping all the cores at 1.7Ghz until I can figure out a good sweet spot for 5A power.

So in case you did not want to read my entire store, the question is, does anyone know of the maximum core frequency for the Up Pentium board to run stable without using more than 5A with only 2 Ethernet cables being used. No display or USB devices are plugged in. Cooling is done passively using a mega head sink provided by the Up Gateway system kit.

I'll do another power draw test with the current 3.3 bios; but, my previous tests with 1.8 showed a UP Squared w/N4200 Pentium only pulling 11.5 Watts or so @ 5.1V (2.4 AMPs converted) under 100% CPU and GPU usage. The N4200 is rated at a max of 6W by itself.

I don't see how an UP^2 with no GPIO Hat nor M.2 nor mPCIe device can pull more than 2.5A, much less 5A. It's those add-ons that start to tack on the extra amps big time.

In addition, most good PSUs will always provide a bit of over current past their rating.

But what I can confirm is a whole bunch of crappy 4, 5 and 6 AMP @ 5V PSUs out there. I have tossed many that would shut down after just a few amps draw. It's funny how none of them showed problems on RPis since the RPis pull much less wattage. you can also see and feel the different between a very cheap PSU and a high quality one.

Edit: I remember now, I did have PSU issues on my RPi 3s with a 4A PSU when I was running a couple with hand jester HATs. Pulled my hair out trying to debug the reboots until I swapped in a higher quality PSU, at just 3A. No problem ever since.

The ones I get from the UP shop are always of very high quality - their supplier is dead on. Especially their big 6A one for the UP Squared.

I even run 2x UP^2 on a single small UP 4A PSU with a Y-splitter that is meant only for a single UP Board, not the UP Squared - much less two (don't do this at home! I am only testing stuff). As long as I don't add any devices onto the UP Squared boards, I run two without issue for multiple installs on the single PSU. If tells you how good the 4A power supply is.

I've been slowly picking a few from AliExpress as I order other parts and have yet to find a good quality one. As a matter of fact, most of these cheap PSUs I get all have a buzzing in them - which is an automatic toss into the garbage for them.

ccalde

Hi @kveroneau,

Is that discussion solved?

Thanks