Raspberry PI 4 – initial thoughts

I’ve had a Pi4 (4GB) for a few weeks now and overall I have to say I am really quite impressed! and for a curmudgeon like me that’s saying something… (bonus video at the end of the post…!)

The first thing to discuss is temperature, while for the first week I was using the Pi with a simple heat sink I soon noticed especially with the unseasonable weather we’re having this summer (30°C+) that heat is a definite issue. Its very easy to overheat (make -j4 for a few minutes) even a decent sized heat sink is going to become saturated. Active cooling (and you really don’t need to go overboard like some have) is in my view essential, especially in hot weather or extended heavy load. While the fan on Pimoroni’s “fanshim” looks small, it is certainly effective, even on a hot day several minutes of continuous heavy load the temperature stayed at the solidly in the mid 50’s in contrast to an additional 20°C or more and hitting thermal throttling even with a heat sink….

Conclusion. Active cooling is a must (but you don’t need to go overboard!) if you really need maximum performance. You’ll probably get away with just a good decent heat sink, but only if you’re just using it for simple browsing etc… its likely that active cooling should prolong the boards lifespan.

Having not done much hacking on a Pi for a while, I was pleasantly surprised that there is a move away from the original Broadcom graphic stack, towards a more usual DRM/Mesa implementation. My initial investigations centred around a simple lighting shader, using 4 point sources. While with a small window, performance wasn’t terrible I wasn’t surprised that a full HD screen rendering speed wasn’t especially stellar, clearly you need to keep your shaders simple and your fullscreen resolution at a maximum of 720p, if you want plenty of 3d objects on screen. Its still early days yet and as I understand it there are still optimisations to have from the new GPU, so it will be very interesting to see the full power of the GPU unleashed in the future.

Conclusion: Great strides have been made and there is more room for improvement, it is possible with care to code 3d that doesn’t suck!

As a departure from previous boards, the Pi4 has not one but two HDMI outputs, I can maybe see for some applications this might be useful, but you’re probably going to be really pushing the GPU to drive two HDMI outputs at the same time. I have had some issues finding a decent converter cable for my monitor as it has HDMI mini sockets, while one converter would work with one monitor it wouldn’t with another…. much confusion getting it all working!

Conclusion: I’d far rather see just a single full sized HDMI socket on the Pi 4 and I expect I’m not alone in this…

Its difficult to find a Linux distribution for the Pi 4 mostly you’ll find systemdOS clones, running a 64 bit OS isn’t currently possible but things are in the works, Gentoo64 on Pi is likely to be the first distro out of the door with 64bit support, as there is already a proof of concept, (can only use 1GB of ram but does work!)

Conclusion: When more work has been done on 64bit for the Pi4 there is likely to be more choice and speed! Its still early days but it will come. At the very least its going to be needed for the 8GB model that’s mentioned in the leaflet that comes with the Pi4…..

Even running a 32bit OS suffering systemd, raspbian (a systemdOS clone) has sufficient performance to be a “daily driver” the ability to put your root file system on a USB3.0 to SATA bridge means you can run on an SSD which is significantly faster than running from the SD card. (You can pick up modest sized SSD for surprisingly cheap now a days)

I did find one of my USB3.0 to SATA bridges was misbehaving when used for a root, adding to the start of cmdline.txt:

usb-storage.quirks=aaaa:bbbb:u 

where aaaa is usb vendor and bbbb is usb product id enabled me to use one of my misbehaving usb bridges…. (see)

Overall conclusion: While there is still work to be done, the Pi4 is by far the most usable Pi when it comes to a traditional “desktop” scenario, many of the internal bottle necks have been completely removed and it shows. While this power makes the desktop a usable prospect rather than feeling that you’re walking through treacle, it remains to see if there will be any real benefit for scenarios like headless sensor monitoring for example.

GPU seems to be the limiting factor at the moment, the CPU is barely bothered by the physics..

Leave a Reply

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