benchmarks

I had to pause some of my work getting a current-gen AMD graphics card running on the Pi 5 and testing a 192-core AmpereOne server to quickly post on the M4's efficiency.

I expected M4 to be better than M1/M2 (I haven't personally tested M3), and I hoped it would at least match the previous total-system-power efficiency king, a tiny arm SBC with an RK3588 SoC... but I didn't expect it to jump forward 32%. Efficiency gains on the Arm systems I test typically look like 2-5% year over year.

The M4 mini I just bought reaches 6.74 Gflops/W on the HPL benchmark.

UPDATE 2024-11-20: I re-ran all my tests without my 10 Gbps Ethernet plugged in, and the efficiency is even better. I also re-tested LLM performance, and made this quick video:

I've recently been debugging some NVMe / PCIe bus errors on a Raspberry Pi, and I wanted a quick way to test NVMe devices without needing to create a filesystem and use a tool like iozone. I don't care about benchmarks, I just want to quickly push the drive and read and write some data to it.

fio is the tool for the job, and after a quick install sudo apt install -y fio, I create a configuration file named nvme-read.fio:

[global]
name=nvme-seq-read
time_based
ramp_time=5
runtime=30
readwrite=read
bs=256k
ioengine=libaio
direct=1
numjobs=1
iodepth=32
group_reporting=1
[nvme0]
filename=/dev/nvme0n1

Then run it with:

sudo fio nvme-read.fio

Easy way to put some stress on the drive, and test your PCIe setup and the drive itself.

For many years, I've maintained some scripts to do basic disk benchmarking for SBCs, to test 1M and 4K sequential and random access speeds, since those are the two most relevant tests for the Linux workloads I run on my Pis.

I've been using this script for years, and it uses fio and iozone to get the metrics I need.

And from time to time, I would test a number of microSD cards on the Pi, or run tests on NVMe SSDs on the Pi, Rock 5 model B, or other SBCs. But my results were usually geared towards a single blog post or a video project.

In 2021 James Chambers set up PiBenchmarks to move to a more community-driven testing dataset.

You can run the following command on your SBC to test the boot storage and upload results directly to PiBenchmarks.com:

Sometimes life has a funny way of lining up opportunities, and one presented itself when Patrick from ServeTheHome reached out and said, "Jeff, I have an Ampere Altra Max server. You wanna come see it?"

Of course I did.

But seeing as Patrick is more than 800 miles away, I had to come up with a reason to go see it, so I pulled out my 6-node Raspberry Pi cluster—with it's 24 ARM Cortex A72 CPU cores—and decided to have a little competition.

And of course that competition is documented in a YouTube video:

Last month, I built an all-AMD PC to try out Linux Gaming with Steam and Proton, and so I'd have a faster native Linux build machine for my various compilation tasks.

This month, Apple introduced the Mac Studio, and as a now full-time video producer, it was a no-brainer for me to upgrade from an M1 Mac mini.

My Mac Studio arrived Friday, and over the weekend, I spent some time benchmarking it against not only my M1 mini, but also my new AMD Ryzen 5 5600x PC build.

My Mac Studio's specs:

Earlier this year, I pitted the $549 ASUSTOR Lockerstor 4 NAS against a homebrew $350 Raspberry Pi CM4 NAS, and came to the (rather obvious) conclusion that the Lockerstor was better in almost every regard.

Well, ASUSTOR introduced a new lower-cost NAS, the $329 Drivestor 4 Pro (model AS3304T—pictured above), and sent me one to review against the Raspberry Pi, since it make for a better matchup—both have 4-core ARM CPUs and a more limited PCI Express Gen 2 bus at their heart.

Around the same time, Radxa also sent me their new Taco—a less-than-$100 Raspberry Pi Compute Module 4 carrier board with 5x SATA ports, 1 Gbps and 2.5 Gbps Ethernet, an M.2 NVMe slot, and an M.2 A+E key slot. (The Taco will soon be available as part of a kit with a CM4 and case for around $200.)

The specs evenly matched, at least on paper:

If you read the title of this blog post and are thinking, "10 Gbps on a Pi? You're nuts!," well, check out my video on using the ASUS XG-C100C 10G NIC on the Raspberry Pi CM4. Back? Good.

To be clear: it's impossible to route 10 gigabits of total network throughput through any Raspberry Pi on the market today.

But it is possible to connect to a 10 gigabit network at 10GBase-T speeds using a Raspberry Pi Compute Module 4 and an appropriate PCI Express 10G NIC. And on my Pi PCI Express site, I documented exactly how I got an ASUS XG-C100C working on the Raspberry Pi. All it takes is a quick recompile of the kernel, and away it goes!

(With a caveat: I'm compiling the ARMv8 64-bit Pi OS kernel.)

It seems every week or so on Hacker News, a story hits the front page showing some new benchmark and how one of the new M1-based Macs matches or beats the higher-priced competition in some specific benchmark—be it GeekBench, X86-specific code, or building Emacs.

Well, here's my quick story.

I've been doing a lot of work with Raspberry Pis lately—more specifically, work which often requires recompiling the Pi OS Linux kernel for the aarch64 architecture. I recompile the kernel enough I made my own shirt for it!

Last week, I posted about the Pi NAS hardware build, and compared setting up an off-the-shelf ASUSTOR Lockerstor 4 to the same thing, but with a Raspberry Pi Compute Module 4 an IO Board.

Some people wondered why not just use a Raspberry Pi 4 model B with USB hard drives, and there are a few reasons:

I'm a fan of Raspberry Pi cases that keep my Pi cool. And the cases made by Argon Forty have great cooling, which is one reason they're a popular choice. Their latest Argon One M.2 case also adds a built-in high speed SSD drive slot!

A few months ago, someone from Argon Forty reached out and asked if I'd like to review the Argon ONE M.2, after they watched my video on booting a Pi 4 from an external SSD, and I accepted.

Unlike most Pi cases, this one actually adds features through it's design, like putting all the ports on the back, and adding a mostly-internal SSD, and so I decided to put it through its paces and see what I liked, and what I didn't like.

Video Review

I also posted a video version of this blog post on YouTube: