Geekbench benchmarks

Surprising result here - i.e. the RPi4B with 4 GB running Stretch came out slightly better than the RPi4B with 8 GB of RAM (Ubuntu 21.04 arm64), and was surpised (and somewhat disappointed) to see them look dismal compared to an ancient Dell 7470 laptop… Not surprised to see the Macbook Pro M1 wipe the floor with my relatively new Ryzen 7 3700x, and the Lenovo Laptop with a Ryzen 5 3500u (but kinda disappointed the Lenovo Ryzen isn’t that much faster than the Dell running Windows).

Note : the arm64 linux geekbench version is still in beta (and 5.4.0 - everything else is 5.4.1)
I’m starting to wonder if the lower result for the 8 GB Pi4B (Ubuntu 21.04) could be Ubuntu VS Debian?.
I might have to setup my other 8 GB Pi4 with arm64 stretch…

Kernels :
Pi4B - Stretch : 5.10.17-v8+
Pi4B - Ubuntu 21.04 : 5.11.0-1015-raspi
Ryzen 7 - Ubuntu 20.04 : 5.4.0-80-generic
Ryzen 5 - Ubuntu 21.04 : 5.11.0-25-generic
M1-arm - MacOS Big Sur 11.4 : Darwin Kernel Version 20.6.0

I might see if I can grab geekbench results off my daughter’s 2020 Intel Macbook Air, my other daughter’s 2020 Intel Mac Mini, and my wife’s circa 2012/2013 Intel Macbook Air…

And I guess I could see if I can upload results for my 2018 iPad Pro 12.9"…

Considering things - that Lenovo was only about $50 more than Dell E7470’s are going for on e-bay - so much better value…

I don’t know how much trust to pay in those results (and it cost me $15 for a geekbench license!) - given my “anecdotal” observation that the Lenovo felt quite “snappy”… I may boot up my Dell 7270 (Ubuntu 20.04 i7 with 8 GB DDR4 RAM [the Dell running Windows only has DDR3 RAM).

I mean, yes, people compare any things in the world like this all the time, while I prefer a more holistic approach most of the time. For example, if you count in how much size you have to pay to get a certain amount of power, then the Raspberry Pi is by a vast magnitude more powerful than this laptop. Or, count in power consumption and again you’ll see the difference between Mount Everest and a little village on sea level.
So, yes, if you have the space, power and need to just have the best power to price ratio, perhaps you might as well use an old laptop, instead of a Raspi. But, if you count in more things, like space and power consumption the way described above, then Raspi is pretty much certainly the undeniable winner.

This is one reason I don’t like buying laptops. Beside all the other downsides to laptops, like lack of repairability or part replacement at a value price, this issue makes it really hard to get a “good” deal on a laptop. You really have to know your stuff, to get a “good” deal for a “good” laptop, as too many times you might get a cheaper one and it still may be one that is by far better than some stupid more expensive one.

I personally do not see much value in such benchmarks. First of all, it seems to me like they are mostly pretty inaccurate for several different reasons. Secondly, they are pure lab benchmarks!
If you want to deduct real life performance from such benchmarks, I doubt you will be able to translate them one by one to what you will actually get in reality.
When I want something benchmarky to look at, I most of the time watch Gamer’s Nexus on YouTube. They are doing professional accurate benchmarks plus real life benchmarks and explain everything about it. Such a 15 minute YouTube video tells you Mount Everest amount more about a certain piece of hardware, than looking at a couple of stupid numbers from Geekbench. Well, that’s my opinion, which arose from experience.

Next bunch of benchmarks I’m going to attempt are blender… not planning on running blender, just their benchmarking suite…

Haven’t gotten around to figuring out blender benchmarks - but - decided to try that same Pi4 with 8 GB running Raspbian/Raspios : arm64 (aarch64) - and - it did slightly better than running Ubuntu 21.04 - but - still marginally slower than a Pi4 with only 4GB of RAM… I can only put that down to the fact the Pi4 4 GB model is essentially “headless” - i.e. it’s not even running lightdm…

In both cases, they’re booting via USB 3 off USB 3 SSD …

Pi4 4GB (slightly better result) : Samsung T5 512 GB SSD
Pi4 8 GB (slightler worse result) : Hynex 256 GB SSD (came out of the Lenovo Thinkpad E495)

I guess I should try the same test again with all the GUI stuff disabled from autostarting?

Here’s something that’s a tad annoying about the Pi Foundation - that 64bit “aarch64” Raspbian build is still marked as “Beta” but it’s from May 2020!!! Talk about “glacial”…

This chart probably describes their “agile methology” :

1 Like