Slashdot Mirror
Raspberry Pi 3 Model B+ Benchmarks Show Significantly Improved Performance (phoronix.com)
fstack writes: Pi Day was marked this year by the launch of the Raspberry Pi 3 Model B+ as the next evolution to this $35 ARM single-board computer. Phoronix has now put out Raspberry Pi 3 B+ benchmarks showing that the Ethernet performance is indeed much faster now but still doesn't stack up to other high performance boards, the SoC temperature is noticeably lower than the very warm Raspberry Pi 3, and the overall performance is a nice upgrade while retaining the same price point as its predecessors. Follow up tests looking at the Wi-Fi performance also show the new 802.11ac dual-band wireless to be much faster as well.
What I found the most interesting in these benchmarks however is how much faster the Asus Tinkerboard is.
It also has four ARM cores and clocked at 1.8 GHz (a third faster) but is several times faster than the Raspberry Pi B+ in some CPU benchmarks. The difference is that the Tinkerboard's CPU cores are running out-of-order while the Raspberry Pi B+'s A53 cores run in-order.
Other than that, the A53 is capable of running 64-bit ARM code which is supposed to be faster than the corresponding 32-bit code.
These tests were run on Raspbian however, which does still not have support for 64-bit code.
"We mustn't be caught by surprise by our own advancing technology" -- Aldous Huxley
>> I don't think you understand what OnShape [onshape.com] is.
OnShape as well as many other Web-based CADs (did you try the TinkerCad?) use client-side rendering (projecting a 3D object into 2D screen surface). That is, the client's (here the PI or whatever your students are using locally) CPU/GPU is processing a 3D instructions of WebGl and plotting it to the 2D screen - a lot of floating point math involved.
WebGl (see: https://en.wikipedia.org/wiki/...) is client-side rendering functionality.
RPi has a 'decent' GPU (measured by cost and power) but it is limited, esp with respect to the memory size and number of FLOPs and likely memory bandwidth. Usually, CAD models are complicated enough to demand more resources. All in all you are going to make the students suffer through a sluggish process.
>> It runs almost entirely in the cloud.
If it did it would not need WebGl. The cloud would render the 3D into a 'picture' surface and serve it back to client. I do not think this is how it works. All in all it would make the UI rather sluggish as the round-trip latency would be remarkable.
4wdloop