PBP benchmarks

[Der Geist der Maschine]

Ok, I agree that we can expect that the performance will be up to 4000 points, so about 3-4 times faster than first PB

I had a very quick look at the geek-benchmark. It's a kind of number-crunching benchmark such as SPEC. I believe that does not say much about its desktop use ... and most here are interested in using the Pinebook Pro as a desktop machine.

Back in the day I was subscribing to printed computer magazines. I vaguely remember ct using a benchmark consisting of macros to torture Word, Excel and so forth. These macros typed text, reformatting text etc etc. Such benchmarks are probably more valuable for the the typical Pinebook Pro user.

Feel free to disagree. Let's start an interesting discussion. All these "where are my coupons" discussions are not

[Wizzard]

You must at least agree if some machine has a score of 1000 and another score of 4000, it is almost sure that the second is much faster This is what the benchmarks are all about.

[CampGareth]

You must at least agree if some machine has a score of 1000 and another score of 4000, it is almost sure that the second is much faster This is what the benchmarks are all about.

I disagree. It all depends on what the test does. I could make a benchmark that rewards slower computers with better scores. A small tweak to a benchmark or compiler could result in 10x better results on one machine or 10x worse results on another. Geekbench don't want to tell us what their tests actually do so their results should be taken with a KFC family bucket of salt.

Unfortunately there's no gold standard in benchmarking. Phoronix is a reasonable bet since you can examine the source and compiled code but it's still prone to huge performance differences depending on compiler tweaks.

[Der Geist der Maschine]

You must at least agree if some machine has a score of 1000 and another score of 4000, it is almost sure that the second is much faster This is what the benchmarks are all about.

The Pinebook Pro is memory constrained and killer apps such as Firefox and Libre Office will swap from time to time.

The more swapping, the marginaller the difference between a Greekbench score of 1000 and 4000. Actually, anyone still remembering Belady's anomaly?

I think most of us would be interested in learning about benchmark results of Firefox automatically opening tabs, scrolling etc and Libreoffice automatically typing text, reformatting it etc. and comparing that to an x86 system with sufficient memory (the latter is what we are used at and are familiar with).

[Wizzard]

Dont forget that original Pinebook has only 2 GB RAM, so Firefox is not a good example, because the difference between PB and PB Pro would be even bigger, I think.

Anyway, I like Geekbench because I can compare my phone with tablet with a computer and that is great, I dont know any other benchmark that runs on so many different systems and architectures.

[Arwen]

If the original Pinebooks have 2GB of RAM, then it might be nice to try and make a kernel for the PBP that only sees 2GB. Then run the same tests on both PB & PBP.

And since the Allwinner is 4 cores, we could also limit the PBP to the 4 A53 cores as a test of lower power usage. This can help determine if somethings like LibreOffice or Firefox are usable at lower power settings.  Like when people are running off battery.

For example, my old Asus EeePC 900 had a single core Celeron @ 900Mhz. While booting was slugish, (even from it's flash drive), it was quite usable in it's day.

[Luke]

Let me know what Linux benchmarks you want me to run and I'll run them -- we can also check how fast the PBP will thermal throttle

[CJMax]

Let me know what Linux benchmarks you want me to run and I'll run them -- we can also check how fast the PBP will thermal throttle

That would be cool! I can think of:

• sysbench (CPU, RAM and I/O)
  
• geekbench
  
• linpack
  
• browserbench https://browserbench.org/Speedometer2.0/ , this one heavily depends on the browser that's being used
  

There are also benchmark scripts for gimp (e.g.: https//openbenchmarking.org/test/system/gimp-1.1.2 ) that look interesting in terms of real life performance, but I haven't tried that on my own PC, yet. 

Phoronix test suite also appears to have a bunch of interesting benchmark tests: https://linuxconfig.org/how-to-benchmark...hics-tests

[Luke]

geekbench on ARM Linux? - there better be a deb in that case, I'm not compiling it from source

[CJMax]

geekbench on ARM Linux? - there better be a deb in that case, I'm not compiling it from source

You're right, seems ARM is only supported via Android.