Mini PC vs Raspberry Pi – Which Is The Better Value?

Are you in the market for a small, low-powered server for your homelab? You may have considered the Raspberry Pi 4 Model B, but recent (relative to the time this article was written) supply chain shortages have made it difficult to find at a reasonable price. If you’re open to alternative options, consider used mini desktop PCs, which are available on sites like Ebay. These systems are relatively small, use minimal power, and can provide better computational capabilities than the Raspberry Pi without breaking the bank.

In this article, I’ll compare the performance of two models of inexpensive mini PCs to the 8GB Raspberry Pi 4B. We’ll also look at the power usage and relative costs of these devices to help you determine which option is best for your needs. If your use case is purely computational, like mine, you’ll find you don’t necessarily need a Raspberry Pi at all.

Before we dive into the comparisons, it’s worth noting that there are ARM-based Raspberry Pi alternatives on the market, such as the Inovato Quadra and the Rock 5B from Radxa. The Inovato Quadra is incredibly inexpensive – costing only $35 – while the Rock 5B has incredible performance for an ARM based board.

Additionally, I’ve written a follow up to this article that includes comparisons with a number of Intel i3 based systems that you should absolutely check out if you’re considering a mini PC purchase for your homelab. Where this article only compares performance using dual channel memory configurations, the companion article shows both single and dual channel memory setups for all of the mini PCs tested.

HP ProDesk 600 G2 Mini & Dell Optiplex 3020M
HP ProDesk 600 G2 Mini & Dell Optiplex 3020M

As this is a somewhat lengthy article, I’ve added a Table of Contents here so that, if you want, you can skip ahead to a specific section – however – I do recommend reading the article in its entirety.



Acknowledgements

Before we get into the comparisons, it’s important to note that I don’t actually own a Raspberry Pi myself. In order to conduct these tests, I was able to borrow a friend’s 8GB model. However, there was one catch – I live in Pennsylvania and my friend lives in the United Kingdom, so all of my testing had to be done remotely.

Unfortunately, my friend didn’t have a way to measure the power usage of the Raspberry Pi, so I had to rely on an external source for that information. I was fortunate to come across a Computer Science professor in Iceland who had published some power usage numbers for the Raspberry Pi as part of a comparison with the Intel Celeron J4105 CPU. I want to extend my sincere gratitude to both my friend and the professor for their help in making this review possible.

The Competitors

Let’s take a closer look at details of each device.

Raspberry Pi 4 Model B

The Raspberry Pi has made a significant impact in the world of computing and has earned a place alongside other iconic devices such as the Apple II, Commodore 64, Amiga, and TRS-80. Since its launch, the Pi pioneered the market for affordable single-board computers, making it a popular choice for tinkerers and DIY enthusiasts. Its place in the history of computing is well-deserved and secure.

Device Specifications

  • CPU: Broadcom BCM2711 – Cortex-A72 (4 Cores / 4 Threads ARM v8)
  • CPU Speed: 1.5GHz Base Clock
  • Lithography: 28nm
  • L1 Cache: 4x 48KB Instruction + 4x 32KB Data Cache
  • L2 Cache: 512KB Shared
  • GPU: Broadcom VideoCore VI
  • Display Out: 2x Micro HDMI
  • Max Memory: Not Upgradeable – 1GB / 2GB / 4GB / 8GB LPDDR4-3200 Options Available
  • Tested Memory: 8GB LPDDR4-3200
  • Storage: Micro-SD Card
  • Connectivity: 1Gbps Broadcom Ethernet + 802.11ac Wifi + Bluetooth 4.2
  • Ports: 2x USB 3.0, 2x USB 2.0
  • Other: 40 pin GPIO
  • Footprint: 85.6mm x 56.5mm

The specific Raspberry Pi 4 computer that was tested for this article was equipped with the tower heatsink and fan combination pictured below. While I’m not including temperature data in this article – as temperatures will differ depending on environment, cooling devices, thermal paste application, etc – I did monitor temperatures during testing and this inexpensive case + cooler combination (affiliate) performed extremely well.

The Pi never experienced any thermal throttling – in fact the highest temperature I recorded was just 41.4C. I’m sure that you could overclock a Pi easily with this setup if you chose to, though all tests were performed at stock clock speeds.

Raspberry Pi 4 Model B with Tower Heatsink & Fan
Raspberry Pi 4 Model B with Tower Heatsink & Fan

Dell Optiplex 3020M

Representing Intel’s 4th generation 22nm “Haswell” architecture, is the Dell Optiplex 3020M featuring the i5-4590T 35W 4 core / 4 thread CPU.

Device Specifications

  • CPU: Intel i5-4590T (4 Cores / 4 Threads) 35W TDP
  • CPU Speed: 2GHz Base Clock / 3GHz Max Turbo
  • Lithography: 22nm
  • L1 Cache: 4x 32KB Instruction + 4x 32KB Data Cache
  • L2 Cache: 4x 256KB
  • L3 Cache: 6MB Shared
  • GPU: Intel HD Graphics 4600
  • Display Out: Display Port + VGA
  • Max Memory: 16GB (2x 8GB) DDR3-1600
  • Tested Memory: 8GB (2x 4GB) DDR3-1600 CL 11
  • Storage: 1x 2.5″ SATA Drive
  • Connectivity: 1Gbps Realtek RTL8151GD Ethernet + Intel 7260 802.11ac Wifi + Bluetooth 4.0
  • Front Ports: 1x USB 3.0 Type A, 1x USB 2.0 Type A, headphone, microphone
  • Rear Ports: 2x USB 3.0 Type A, 2x USB 2.0 Type A
  • Other: Internal 1.5W Speaker
  • Footprint: 182mm x 176mm

I bought this unit originally to run as a local web server + Plex media server and it worked extremely well for those purposes. I eventually decided to replace the old desktop PC that we’d given my 7 yr old with this little PC as it uses less power than the old desktop did and is nearly as powerful.

Dell Optiplex 3020M
Dell Optiplex 3020M

At this point, Intel’s 4th gen platform is really starting to show its age – the lack of NVMe drive support, limited memory expansion, and relative to 6th gen and newer devices power hungry 22nm process technology should keep the prices on these devices reasonable in the second hand marketplace.

HP ProDesk 600 G2 Mini

Next, from Intel’s 6th generation 14nm “Skylake” architecture is the HP ProDesk 600 G2 Mini featuring the i5-6500T 35W 4 core / 4 thread CPU.

Device Specifications

  • CPU: Intel Core i5-6500T (4 Cores / 4 Threads) 35W TDP
  • CPU Speed: 2.5GHz Base Clock / 3.1GHz Max Turbo
  • Lithography: 14nm
  • L1 Cache: 4x 32KB Instruction + 4x 32KB Data
  • L2 Cache: 4x 256KB
  • L3 Cache: 6MB Shared
  • GPU: Intel HD Graphics 530
  • Display Out: Display Port + VGA
  • Max Memory: 32GB (2x 16GB) DDR4-2133
  • Tested Memory: 8GB (2x 4GB) DDR4-2133 CL 17
  • Storage: 1x 2.5″ SATA Drive + 1x PCIe 3.0 2280 NVMe Drive
  • Connectivity: 1Gbps Intel I219LM Ethernet + Intel 7265 802.11ac Wifi + Bluetooth 4.2
  • Front Ports: 1x USB 3.0 Type C, 2x USB 3.0 Type A, headphone, microphone
  • Rear Ports: 4x USB 3.0 Type A, 1 Modular Port
  • Other: Internal 1.5W Speaker
  • Footprint: 177mm x 175mm

With as pleased as I was with the purchase of the Dell 3020M, I decided eventually I wanted to try out another inexpensive mini PC from another manufacturer and eventually decided on this HP model. I was so impressed with it that after a few weeks, I decided to sell off an old desktop PC that I was using as a server (an AMD FX-6300 based PC) and my son’s PC (which I mentioned previously) with an Intel i5-6400 and bought 3 more of these tiny systems.

HP ProDesk 600 G2 Mini
HP ProDesk 600 G2 Mini

Replacing these desktop “servers” increased the total compute power available in my homelab while reducing the physical footprint and overall power consumption.

Benchmarks

All benchmark tests were performed using Phoronix Test Suite version 10.8. The operating system used on the Raspberry Pi was Raspberry Pi OS Lite 64bit – based on Debian 11 “Bullseye”. Debian 11 64bit was installed on the HP and Dell PCs.

I attempted to select a mix of tests that might represent typical workloads someone might deploy on in a homelab environment. While the Sysbench CPU test is completely synthetic and doesn’t represent any real-world workload, the bulk of the tests were chosen as proxies for a handful of tasks one might deploy in such an environment.

Sysbench – Memory

I expected the Pi to perform well on the memory test as it’s outfitted with relatively fast LPDDR4-3200 memory. The Pi just edged out the Dell – which I was initially surprised to see beat the HP pretty handily in this test. The Dell device is running DDR3 compared to the HP’s faster DDR4 memory. The Pi scored 8793 MiB/second compared to the Dell’s 8767 MiB/second and 8279 MiB/second for the HP.

Sysbench Memory Results Comparison

It wasn’t until after testing that I checked into what the memory timings were on the modules installed in the mini PCs. The Dell’s DDR3-1600 CL 11 has much better memory timings compared to the HP running DDR4-2133 CL 17. Even though the Dell’s memory is technically slower, the better timings more than makeup the deficit.

If you’re not familiar with how memory timings / latency effect performance, Steve and the guys at GamersNexus did a great series of videos explaining why it’s important.

Worth noting, I did try running the test again on the HP with DDR4-2400 memory to try to improve scores, however, the maximum speed the system would allow is 2133. If I want to improve memory performance on the HP, I’d need RAM with lower timings.

FFmpeg Encoding

Moving into more real-world testing, a popular use case for Raspberry Pi is as a Plex / Emby / Jellyfin media server. When testing H.264 encoding with the FFmpeg test, the Pi took nearly 39 seconds – over 4 times longer than the HP PC at 9.7 seconds which only beat out the Dell by less than a second. at 10.54 seconds.

FFmpeg Encoding Results Comparison

Either of the mini PCs will outperform the Raspberry Pi when it comes to encoding video – giving you the option to allow for more concurrent streams and/or better quality encoding.

GnuPG Encryption

When it comes to encryption related tasks, performance is going to vary depending on the type of encryption. When it comes to OpenPGP as was benchmarked in this test, the Dell came out ahead at 98.52 seconds just beating out the HP at 101.81 seconds. The Pi took roughly twice as long at 200.71 seconds.

GnuPG Encryption Results Comparison

I can’t say for certain – it would take additional testing that I wasn’t able to perform because I don’t have direct access to a Pi – that performance might improve in this test if the system were using a better storage medium than micro-sd cards. With physical access to the device, I would have preferred to run the system using a SSD inside a USB enclosure.

While this storage difference puts the Raspberry Pi at a distinct disadvantage compared to the mini PCs, I still view it as being a fair comparison simply because this is one of the limitations you would need to consider when deciding between devices.

PHPBench

PHPBench should give us some idea of relative performance with self hosted web based applications like Nextcloud or WordPress. Here, the HP scores over 2.8 times higher than the Pi at 524,188 points with the Dell coming second at 494,195 points and the Pi managing only 184,704 points.

PHPBench Results Comparison

In addition to the superior performance, for these kinds of tasks, the superior storage capabilities of the mini PCs can’t be ignored – especially for applications like NextCloud where you’ll want the superior speed and storage longevity afforded by a SSD.

Linux Kernel Compilation

I decided to test the Linux kernel compilation time to get an idea of how a computationally demanding task would perform. I did not expect the Pi to do well in this test – given the other test results – but I didn’t expect it to score as poorly as it did.

As with the other tests, the HP and Dell devices scored similarly – 345 seconds and 374 seconds respectively. The Raspberry Pi, however, took 4624 seconds to complete – over 10 times as long as the mini PCs. There was no way to create a reasonable scale for the chart below for these scores, so instead I decided to just have the Pi blow out the chart.

Linux Kernel Compilation Results Comparison

While testing, the mini PCs had no problems keeping the CPU threads at 100% during the test until completion. The Pi, though, could only keep all 4 threads at 100% for 5 or 6 seconds at a time before one of them would back off to around 30% for a few seconds.

If I wouldn’t have been monitoring temperatures during testing, I would have thought this behavior was related to thermal throttling. However, knowing that this wasn’t the case, my suspicion is that the Pi’s CPU cache architecture is to blame here.

If you compare the cache on the Pi with either of the Intel based systems, the Pi has more L1 cache than either of the others, however the Intel chips have twice the L2 cache of the Pi and an additional 6MB of L3 cache available.

System Power

Except for the Raspberry Pi – as I noted earlier in this article I had to source power numbers for this device – power measurements were taken using a Kill A Watt meter (affiliate) and represent the power use of the entire computer, not just the CPU.

Raspberry Pi devices are well renown for their extremely low power use. At idle, the device uses a scant 3.8W of power. The HP computer with it’s i5-6500T isn’t that far far behind with idle power use of just 6.9W. The Dell with it’s i5-4590T uses nearly double the idle power of the HP at 13W. The 14nm process node of the 6th gen i5 CPU as opposed to the 22nm node of the 4th gen part plays a large part in the efficiency gains seen by the HP device – even with a 500MHz higher base clock speed.

Power Use Comparison

Looking at maximum power use, again the Raspberry Pi impresses – using only 6W of power, whereas the Dell and HP units use 38W and 39W of power, respectively, when at full load. To be fair, compared to the Pi’s power use these numbers seem enormous but compared to your typical desktop computer these numbers are quite low.

Overall, the mini PCs aren’t going to dethrone the Pi in terms of power use, but they don’t do badly.

Cost

As the saying goes, there are no bad products, only bad prices – and this is generally true. When examining the costs of these devices, I’m assuming that the user doesn’t have any preexisting equipment that they can reuse. I’m only including the cost of the device, not any additional accessories like display cables, keyboards, or SD card readers.

Raspberry Pi

Let’s start by determining the bare minimum cost you’d pay before taxes and shipping to purchase the 8GB Raspberry Pi 4 Model B like I’ve tested here.

The $75 price is the manufacturer’s suggested retail price of the 8GB Raspberry Pi board. If you don’t need 8GB of RAM, you can opt for one of the lower-end models to save some money. However, to make the Raspberry Pi work, you’ll also need a few additional pieces of equipment, such as a power supply, a micro SD card for storage, and a heatsink to keep the CPU cool. All told, the minimum you’ll spend to get the 8GB Raspberry Pi up and running is $101 before tax and shipping.

Unfortunately, the supply chain problems have made it difficult to find the Raspberry Pi at its MSRP. If you’re still determined to buy one and are willing to pay a premium, you may be able to find them for sale by third-party sellers on Amazon (affiliate) for $185 or more. If you factor in the cost of the minimum required equipment, you’re now looking at spending $211 or more before tax and shipping.

Now, let’s compare these costs to purchasing a used mini PC off Ebay.

Used Mini PCs

It’s worth noting that when shopping on Ebay for a used mini PC, there are many more options available beyond the two models I’ve tested in this article. If you have a specific part in mind – for example, the i5-6500T – it’s best to search for that specific component rather than a device name.

Keep in mind that prices on Ebay are constantly fluctuating. The prices mentioned in this article are current only as of the time of writing. I’m not providing direct links to listings as they will eventually expire.

  • $142 – Dell Optiplex 3020M Micro 8GB DDR4 + 500GB HDD – Ebay (Includes Shipping)
  • $150 – Dell Optiplex 3020M Micro 8GB DDR4 + 240GB SSD – Ebay (Includes Shipping)
  • $49 – 16GB (2x 8GB) DDR3-1600 CL 11 – Amazon (affiliate)

The prices for the Dell 3020M on eBay seem a bit high to me at the moment, but as I mentioned earlier, prices can vary greatly on the platform. If you’re specifically interested in a system with the i5-4590T, be sure to search eBay specifically for that part to find the best deals.

I was able to purchase my Dell 3020M with 4GB of RAM in December 2021 for $80, including shipping. I also bought a RAM upgrade for $17 on eBay. With some diligence, you may be able to find even better prices. For the purposes of this article, I’m using the prices that are currently available to me.

The better mini PC deal, however, is the HP ProDesk 600 G2 Mini.

Prices for the HP ProDesk 600 G2 Mini can vary significantly depending on the seller and the specific listing. Some include better storage options than others, and some don’t come with an AC adapter. However, if you keep an eye out, you can often find listings like the ones I’ve detailed here. As I mentioned earlier in the article, I’ve purchased four of these mini PCs so far – one with 4GB of RAM and a 64GB SSD, and three with 8GB of RAM but no disk. In each case, I paid only $100, including shipping.

Stack of 3 Mini PCs
A stack of mini PCs makes a great small server cluster

If you don’t need much storage and plan to upgrade the RAM, the $100 price point is hard to beat. On the other hand, if you’re fine with a hard drive instead of an SSD – or if you already have an SSD that you can use – the $115 option with 8GB of RAM would be a great choice. Even if you don’t have an SSD, you can usually find a brand new one for around $20.

For comparison against the Raspberry Pi, I’ve used eBay listings that include SSDs and paired them with memory upgrades, as storage needs can vary widely. Memory upgrades will be priced with new parts, like the ones I’ve linked above. If you’re willing to shop around, you may be able to save a few dollars by purchasing used memory on Ebay as well.

Here’s how the costs compare.

Cost Comparison

As I mentioned earlier, the minimal Raspberry Pi setup will cost you either $101 or $211, depending on whether you pay the MSRP or not. However, the next cheapest option, the HP ProDesk 600 G2 Mini, will only set you back an extra $21 if you upgrade the memory to 8GB. Even if you upgrade the memory of the HP to its maximum 32GB, it will still cost you $203 – still cheaper than the “scalper price” of the Raspberry Pi. The Dell system at this cost isn’t worth considering as the value isn’t as strong as the HP.

Even if you manage to get a Raspberry Pi at MSRP, It’s my assertion that a used mini PC is still the better value. Spending the extra $21 gets you a lot of advantages.

  • SATA SSD support
  • NVMe SSD support
  • 2 to 4 times the CPU performance
  • Memory upgrades

Even if the CPU performance was identical between devices, having the built in support for better storage devices would alone be worth the $21.

Where Raspberry Pi Makes Sense

To be completely fair, there are times when you might want to choose a Raspberry Pi or similar device over the alternatives.

For example, if you’re looking to embed a computing device into something like a robot or drone, the Raspberry Pi is likely the better choice due to its much lighter weight, smaller footprint, and ability to be powered via battery / without a heavy power adapter.

While you could add GPIO capability to a mini PC through something like an Arduino, the Raspberry Pi is still likely to be the better choice, unless you really need the improved compute of the mini PC.

Finally, if you’re trying to develop code for ARM-based systems, there’s no used PC that can help you there. In that case, the Raspberry Pi is always going to be the better choice.

Conclusion

Ultimately, only you will know what your project’s needs are when choosing a compute device. However, if your use case for the Raspberry Pi is computational and you don’t require any of the unique characteristics of the Pi, then you’re probably better off buying an inexpensive used mini PC.

Tell me what you think in the comments below. Where would you make use of a Raspberry Pi? Would you still choose the Pi over a mini PC anyway? I’d love to hear your perspective.

If you found this article helpful and would like to support our efforts to create additional resources like this, please consider making a donation. Your support is greatly appreciated!

If you can’t make a donation, please consider sharing this review with others who may be interested. If you have questions about anything regarding this review, please be sure to leave them in the comments below. Thanks for reading, and I hope you visit again soon!

7 thoughts on “Mini PC vs Raspberry Pi – Which Is The Better Value?”

  1. I just got done building a porch, so I’m very tired. But I did find your article interesting to comment.
    I’ve decided about two years ago to purchase a Pi. In fact today, I have two of them. One of them I use as a NAS server. I love it. It stays on 24/7 (even in storms), it doesn’t eat up my electricity bill, and the cost when I purchased it was only $100.00. Now under the desk, I also have two high-power PC computers. One is an I9-11900K and the other is an I7-9700K. I use the I9 for daily web stuff and to play my two games (Risk and World Of Warships). My I7 I use to host my website. Why I know I could do it on a Pi, but I wanted speed and power, which I don’t think my Pi can do (maybe). But on the 2nd Pi, I will be putting Pi-Hole on it. Is there a forum for Pi-hole to where I can ask certain questions? As far as mini PC vs Pi, I would take the Pi.
    Thanks For Reading.
    Dan

    1. Hey Dan … I don’t know a specific forum for Pi-hole but I have an entire series of tutorials on Pi-hole. You can always feel free to ask questions on one of those where appropriate.

  2. Hi! great article. I was wondering if you used the hardware encoder on the pi for the ffmpeg test? (h264_omx)

  3. Your results are somewhat surprising to me, because the passmark website shows the CPU in the pi as being substantially lower in performance than the ones in the mini pcs.

  4. InnocentBystander

    I wish I could find this blog article before. Found the link on the right bar of your series about PiHole. https://www.techaddressed.com/reviews/what-is-pi-hole-why-would-you-use-it/
    I plan to setup a PiHole using Raspberry Pi because it make sense to use a low power device. Expecting $50 max in spending, I am shocked that a RPi will cost $200+. Even on used markets such as ebay or local ads, the price is absurdly high for such a “low tech” device. Then the idea of just using a 2nd hand low power computer came naturally. Searching around I discovered the series of Dell Optilex micro or Lenovo tiny. There were plenty, less expensive, more powerful and easier in maintenance.

    In my use case, PiHole and File server, all I need is a Debian server which can have SATA and NVMe. The PCs is much more convenient than the RPi. The wattage of those “micro” PCs is indeed x3 compared to the RPi, but stilll very reasonable compared to regular desktops. But they are more robust, durable and easy to repair.

    In terms of price, those used micro PCs costs less than half of a RPi. A saving of $100 means they can run 10 years to break even with the operational cost of running a RPi. Sorry for the environment that I put out a few extra KW hours per year running those micro PCs instead of the RPi. But the economics of the Raspberry universe is irrational.

  5. I grab cheap micros when I stumble across them. I have Lenovo 73s, HP, and dell.. I load them up with Debian, mergefs and install docker. They can join swarm or kubernetes clusters. This works well with docker.

    I did this when the 4Bs were stupid expensive. 4 gb were over 100, and 8 gb were 130 or more.

    I paid 60 to 70 for micros with 8 GB ram, I3 or I5, and a small ssd 256 to 512. I only buy 7th gen or later. I was getting a swarm worker for half the price of a pi.

    the only issue us ensuring one uses 86 and aarch64 images. this has been an issue a few times. you have to build your own,

Leave a Comment

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

I accept the Privacy Policy


Scroll to Top