The Raspberry Pi Foundation has once again upped the bar on its little computer’s capabilities, debuting a Pi 4 with 8GB of RAM, comparable to many desktop and notebook computers that you may be reading this on right now. Since the Pi 4 is touted as a possible desktop replacement, this certainly makes sense, but do you actually need the Pi 4 4GB or even the 8GB model? Or would the (now base) Raspberry Pi 4 2GB version work just as well for your needs?
Ähnliches Produkt:
RASPBERRYPI4 2GB | 4 Model B
Raspberry Pi Foundation Eingebettete Systementwicklungsboards und -kits AnzeigenÄhnliches Produkt:
RASPBERRYPI4 4GB | 4 Model B
Raspberry Pi Foundation Eingebettete Systementwicklungsboards und -kits AnzeigenRaspberry Pi Memory Test: Comparing Configurations
To test different Raspberry Pi memory configurations, I threw a number of programs at the different models. As listed, programs were opened successively and not closed unless noted, with total Raspberry Pi RAM use listed at the end. There are also a few notes on swap memory usage, which generally meant that it was time to try out the next model. Performance was measured remotely via SSH/PuTTY, using the htop command
Pi 4 2GB:
· Startup into Raspbian: 169MB
· Entire LibreOffice suite loaded (Writer, Calc, etc) without opening anything more than what was prompted: 567MB
· Minecraft Pi (creating new world): 610MB
· Blender: 725MB
The numbers above are used as base “weight” for memory usage. From there, I loaded Chromium, with tabs set to Arrow.com:
· (+1 tab: 1 total) 912MB
· (+1 tab: 2 total) 962MB (1MB swap after some time with tab open)
· (+1 tab: 3 total) 1.01GB no change in swap
· (+1 tab: 4 total) 1.06GB, 1.5MB swap
· (4 tabs total) arrow.com + (1 tab) YouTube tab playing random video, 1.18GB memory, 34.2MB swap.
Considering this is the baseline model, I was extremely impressed with its capabilities. As it was using a small amount of swap memory at the end, the next model up would theoretically provide some performance gains.
Pi 4 4GB:
Screencap
· Startup into Raspbian: 197MB
· Base as previously outlined: 707MB
· (+1 tab: 1 total) YT first: 1.05GB
· (+4 tabs: 5 total) Arrow.com tabs: 1.22GB
· (+4 tabs: 9 total) Arrow.com tabs 1.45GB
· (+4 tabs: 13 total) 1.67GB
· (+8 tabe: 21 total) 2.02GB
· (+8 tabs: 29 total) 2.4GB
· (+8 tabs: 37 total) 2.478GB It seem like there are some spikes in CPU usage; ended up closing some tabs. Total tabs open at this point became questionable.
· Open YouTube on 4 tabs on another Chrome instance (2.62GB usage). At this point, YouTube tab said that it’s auto rendered and had an error and that I needed to restart my computer. Whether because of YouTube, my internet connection, or the Pi itself, the system is definitely misbehaving. RAM is still not anywhere near maxed out, and swap isn’t used.
· Started a new game in Minecraft (2.71GB usage).
· Loaded several of the built-in games, (2.86GB usage); swap usage seen at 768KB. CPU usage, however, was quite high.
YouTube did not appreciate being excessively multi-tabbed
Pi 4 8GB:
· Startup into Raspbian: 196MB
· Base as previously outlined: 743MB
· Immediately opened 8 tabs of YouTube, which gave me some errors, and processor seems to spike. Memory usage still under 2GB
· Closed down Chrome, loaded up every game on the menu, and RAM was under a GB, though CPU usage was spiking.
· As it was very difficult to max out 4GB of RAM in previous testing, and CPU usage appears to be the real performance bottleneck, testing was left off here.
2, 4, or 8GB for a Typical Desktop?
Going through this experiment, it seems that the 2GB Pi will be good enough for most desktop usage, with a possible performance boost at the 4GB level. Keep in mind that I didn’t load up any non-browser files, beyond any templates where I was prompted. Your usage might mean it doesn’t take such a ridiculous number of programs open to get to the point where 4GB gives you a performance edge, but I’d guess that in most cases, as it stands today, an extra 2GB won’t make a huge difference.
That being said, upgrading the base model from 1GB to 2GB appears to have been a very worthwhile move, as “eating” a GB is fairly trivial. Unless you’re doing a rather non-RAM intensive headless operation, and can get a good discount, I’d make sure to get at least the 2GB version. If you do specifically need it, for whatever reason, you should be able to find the 1 GB model for some time.
As far as the 8GB device, for typical desktop operation, you may be better off saving a few bucks with one of the lower models. Even at 4GB of RAM, memory doesn’t seem to be a serious performance bottleneck. Why, then, does it exist? Where could it be used? While I haven’t tested all of the following, I’ve listed a few ideas for where the extra GBs might come in handy. As outlined in this Pi 4 hands-on post, there are some minor changes to the hardware, and you can be certain it won’t have the “infamous” (though not actually that big of a deal) USB C issue seen in early Pi 4 versions.
RAM Disk and Cache
With an abundance of extra RAM, you could use this as dedicated disk space to load up important programs for easy access. Linux automatically loads frequently used applications onto RAM, and it’s possible to also set up a dedicated RAM drive if you’d like to handle this in a more hands-on manner. Between potential speed benefits for programs that need to load a lot, and saving wear and tear on your SD card, this could be a good justification for loading up on memory.
Headless Applications
What about so-called “headless” applications, where a Pi is put in a corner somewhere to execute a task without a dedicated screen? As it just so happens, I have multiple Pis around my house for these purposes. One popular application that I run is called “OctoPrint,” which allows for remote control of your 3D printer. Logging into this web interface on both my phone and desktop, I measured the memory used htop at just 168MB. There was little change when actually running a print. Granted, this was a Raspberry Pi 3B, so things may be somewhat different with a Pi 4, but it appears a RAM increase wouldn’t be terribly useful.
Logging on to my Pi Zero W, which runs a plant watering operation (as outlined here), this application uses only 41MB—well under the 432MB listed as being available. No surprise here, however, it’s worth noting that the Pi 4 8GB has roughly 195 times the amount of RAM needed to run this simple headless application—quite impressive. I do have one other headless device, however, that I thought was a bit more likely to see a benefit:
Raspberry Pi NAS (Network Attached Storage) Acceleration?
Switching earlier to a Pi 4 2 GB from a Pi 3B did increase speed, but there was no benefit to 8 GB of RAM
It’s been reported that the Pi 4’s 8GB can be helpful when transferring large files, so I decided to test it out on my (rather sluggish) Raspberry Pi NAS setup. Here are the results. Spoiler: it’s not that impressive:
· 27.2MB file transfer (wireless connection from both computer and Pi to router):
Pi 4, 2GB: 17 seconds to Pi, 14 seconds to computer
Pi 4, 8GB: 17 seconds to Pi, 14 seconds to computer
· 27.2MB file transfer (wireless connection from computer, wired Pi to router):
Pi 4, 2GB: 5 seconds to Pi, 10 seconds to computer
Pi 4, 8GB: 5 seconds to Pi, 9 seconds to computer
· 5.76GB file transfer (wireless connection from computer, wired Pi to router):
Pi 4, 2GB: 13:50 from computer to Pi
Pi 4, 8GB: 15:28 from computer to Pi
There was no increase in performance using the Pi 4 8 GB over the 2GB device, and in one case things were actually worse (whether or not it was due to the Pi itself). That’s not to say that extra RAM would never be beneficial, as there are many factors that affect these numbers (variable internet usage, that USB 2.0 was used for some of the experiments, and more).
Generally speaking, you’ll want to identify your performance bottleneck before throwing more RAM at it. I was hoping that my NAS setup would be the “killer app” here, but such was not the case, for my setup at least.
Multi-Headless Pi?
As noted by Linux aficionado Pat Regan, a lot of services sit idle most of the time, but still consume RAM. With enough RAM, a single Pi could potentially run multiple such services like OpenHAB, OctoPrint, and more without skipping a beat. Whether this actually needs 8GB is an open question. On the other hand, there are some advantages to having things segmented, as the failure of one Pi wouldn’t bring all of your automation crashing down—and your plant will still hopefully get watered!
The 64-Bit Future
Raspberry Pi’s official operating system, Raspberry Pi OS (formerly Raspbian), is 32-bit, which means that single processes can’t use more than 4 GB of RAM. While there can still be some benefit from extra RAM (e.g. caching, multiple programs and tabs open), it does restrict how it can be used. The good news is that this OS is finally going 64-bit, with a beta version available now. Once it becomes more developed, this should mean a significant performance improvement. With software playing catch-up, it’s possible that the extra 4GB will become more valuable, computing-wise, in the future.
8 GB RAM? Less?
Is 8GB worth it right now? That really depends on your use case. If you want the best of the best, and money is no object, then why not? The fact that it may become more attractive in the future is a good selling point, and there are a number of more advanced tasks that could potentially benefit from the RAM boost. On the other hand, at a list price of $75, you could theoretically get two 2GB Pi 4s for less than a single 8GB model (though actual prices do tend to deviate). It’s certainly worth considering your different options, but it’s always good to have choices!