Initial Raspberry Pi 4 Performance Benchmarks

[phoronix]

Phoronix: Initial Raspberry Pi 4 Performance Benchmarks

It's been (and still is) a particularly busy few weeks for benchmarking. For those curious about the Raspberry Pi 4 performance that was announced at the end of June along with Raspbian 10, here are our initial performance benchmarks of the Raspberry Pi 4 Model B in 2GB and 4GB variants compared to various other ARM SBCs.

http://www.phoronix.com/vr.php?view=28057

[cen1]

Nice upgrades

[etam]

Those differences where 2GB raspi is faster than 4GB are puzzling. I'd really like to know how if it's really just heating problem and how would benchmarks look with it resolved.

[willmore]

You're going to need to add some thermal monitoring to your testing if you want it to be informative. There are a few tools already available to monitor the clock speed, temperature, and throttling history on Rpi boards. It would be very useful to see the output of those while running these tests.

I'm curious about the heatsinks you used on the 2GB board. Where did those come from? It was my understanding that the boards do not come with them. Was that something your supplier recommended?

[cyclistefou]

Those differences where 2GB raspi is faster than 4GB are puzzling. I'd really like to know how if it's really just heating problem and how would benchmarks look with it resolved.

I concur. I think this is really worth investigating - just letting it go like it is in the article, passes a very unlikely message. I of course understand it is not necessarily Michael's job to do everything but I think there should be stronger warnings, even at the beginning of the article, saying that some results are provided as-is but would require investigation, and should be taken with a grain of salt in the meantime.
Edited for grammar

[AndyChow]

I wonder if the thermal differences are caused by the enclosure, or silicon lottery. It would be cool (no pun intended) if you could run the tests in an environment where no thermal throttling happens at all.

[AndyChow]

I'm curious about the heatsinks you used on the 2GB board. Where did those come from? It was my understanding that the boards do not come with them. Was that something your supplier recommended?

Guessing from the sticker on the bottom left in the first picture, I'm going to guess it's from CanaKit. Their basic kit comes with three heatsinks and a power supply.

[sykobee]

Does the 4GB have the VLI firmware fix installed? That should help the thermals.

But I agree, a small heatsink seems to be a must for the Pi 4 to prevent throttling under sustained load. I will remember to get some when I get around to getting the device.

[blackshard]

I agree about benchmarks being almost trash... If we don't know when thermal throttling engages all these numbers are just raw numbers (that's true for other boards too).

[some_canuck]

but how does it stack up vs pc hardware