Raspberry Pi 3 Model B+ Benchmarks

[willmore]

The method was the same for testing across all systems and the stock/used distros is indicated on the system table chart in the article.

That table (on page 3) is both incomplete and inaccurate. For all the SBCs, the Disk entries are not meaningful. I assume those are the name reported in some fashon by the uSD cards. But, they're lacking brand and model info in a meaningful format. The Orange Pi PC is listed has having three cores. The version of Armbian is missing as well--and appears to be very old based on the kernel version. Older versions of Armbian are based on the BSP kernel which has signifigant problems with DVFS and is inappropriate for most use.

[noduck]

Thanks, looks like a reasonable incremental hardware upgrade: real improvement in networking speed (better usage of USB2) and thermal improvement, and a little more processing speed.

[caligula]

That table (on page 3) is both incomplete and inaccurate. For all the SBCs, the Disk entries are not meaningful. I assume those are the name reported in some fashon by the uSD cards. But, they're lacking brand and model info in a meaningful format. The Orange Pi PC is listed has having three cores. The version of Armbian is missing as well--and appears to be very old based on the kernel version. Older versions of Armbian are based on the BSP kernel which has signifigant problems with DVFS and is inappropriate for most use.

Well, if the test is designed to make RPI 3b+ look good and opi look like crap, that's how you do it. Instead of using new, mainline kernel, use a buggy legacy kernel. While at it, use wrong voltage and freq to make it throttle and fail. Also don't compare $35 RPI with $25 opi, use the years old $15 version to make them look worse. The new version only has twice the ram and 10x IO bandwidth. No biggies.

[t.s.]

The method was the same for testing across all systems and the stock/used distros is indicated on the system table chart in the article.

As the others stated, could you use the newer armbian distro for testing orangepi pc? People says that it has quite impact.

[LinuxID10T]

Could you try some graphics benchmarks with the VC4 driver and gl4es on the other ARM SBCs? I'd you decide to do so I can definitely help with the setup for gl4es

[ResponseWriter]

Thanks for the tests, Michael! Any chance you'll include more encoding (flac, opus) and compression benchmarks in your next tests? A test of AV1 encoding (perhaps a second or 2) might be interesting when encoders are more widely available.

Looks like nothing to warrant an upgrade from my mostly idle Pi 3 until they at least double the RAM, or I could look to one of the alternatives when I have some more demanding workloads to run on it.

[wizard69]

Thanks for the continued interest in ARM boards! Sadly this doesn't look like the new board i was hoping for. The years old Odroid C2 remains a better board, this is especially frustrating know how rapid the world of ARM application processors is improving.

[pcxmac]

Yeah, the Exynos ARM processor is a lot more interesting than the GPU-CPU Broadcom part.

[chimpy]

Michael Hopefully next time you do an ARM sbc round up you can get your hands on a XU4; it's still the best bang for your buck board (based on raw power).

[jacob]

Michael, which OS did you run the benchmarks on? If Raspbian, don't forget that AFAIK it's still strictly ARMv6 so the CPU-intensive benchmarks don't do justice to the Pi.