Whenever there is an upgrade, there are always mixed report regarding performance. Some said good, some said mediocor. I can understand that, a piece of software eval cannot be done in a few hours, to post it great or horrible in a few hours of use is premature to say the least. However, whether it is fast or not should be a YES or NO answer. Still, look at every single update, whether it is faster, the answer is ALL over the place. I think it is time for the community or some gurus to decide on a set parameter for the benchmark tests, be it web browsing, spreadsheet loading, txf back and forth a large document via BT/wifi/whatnots, W/R millions times on SD card, things like that and keep a record, so one can always test a new upgrade performance and compare and report, be it faster or bloater software.

This is very important for users that are not living at the bleeding edge. I am one. I dont care whether I am using OS2005 as long as it serves me well. So, to decide to upgrade or not, I would have to balance between pros and cons in order to make a decision. And benchmark performance is a very important factor. As a matter of fact, benchmark performance also useful in comparing different kernels, mmc boot, rotation install etc. It just helps to streamline performance. Yes, I know benchmark measurement has its limitation, but hey, OS2005, OS2006, OS2007 and OS2008 and now look into Fremantle and we still aint got any.

I am also aware some xterm/command/speed measurement regarding SD cards W/R, it is limited,
1) measure only SD card
2) intimidating, not an average joe can do it

For more benchmark report/use, it has to be a piece of .deb that an ave joe can install and use.

TIA,

bun

awesome, let us know when you have finished writing it!

We could do that; or rather we could start. But the flamewar over which editor to write it in would only be exceeded by the one over what tests and how to weight them. Unless you have a definite purpose in mind, you have to make a synthetic benchmark, which is way more art than science. (Unless, of course, you know which system is to win; then it's all science.)

I think this is a start.

The tablet is a wifi/audio/video device. I think benchmark in these 3 areas would be a good start
1) 10,000 looping of drawing curves on screen (video)
2) R/W repeatedly for a large file to SD card (media), hdparm or this would be just fine, http://www.internettablettalk.com/fo...d.php?p=137282
3) dl/upload a large dummy files to maemobox.org (wifi/web whatever)

a dumb graph showing test results side by side with 'established' system, OS2005, OS2006, OS2007, OS2008, Fremantle, 770/N800/N810/internal flash vs MMC boot.

Really, this took me less than 3 min...........only if bun can write codes I know cobol and Fortran and Basic. Oh well, Write a small app that automate the above steps, package into a .deb, and keep it in wiki called it BBB test, and anybody interested could just dl and install and BBB it 1st before making any system change. The key is automation. Most ave joes are either ignorant/lazy like me, if it is easy, people is going to bench it. simple.
Oh, remember KISS, keep it simple and stupid, so it would run on any old/present/future OS.

BBB? Better Business Bureau or Benson something....
Come on, lets get this done before next weekend........ we can do it. Maybe that will help me to decide to stay in OS2005 or upgrade....

If not, I will stay in the stone age, like this, do it by hand....http://www.internettablettalk.com/fo...ad.php?t=23168

bun

But you didn't even cover JS performance in the browser.

And unless this also provides a good way to quickly dual-boot, how much use is the comparison of wifi speeds? Different times, locations, other network traffic, etc., will totally drown out the differences between typical updates.

You got to start from somewhere......so does Einstein or chairman Mao.

The most valuable info about benchmark is NOT a single measured data point, the absolute value weighs much less than its relative value in a trend. With the same system, same wifi, same operator, BBB over several system or several tablet or several kernels, the info is still valuable. Yes, location, wifi traffic, whatnots also play a part, thats why one observe a trend instead of a gawking at a single datum.

I remains, the tablet is a wifi/AV device, benchmark in these areas are important.

bun

Wow, um, that's an interesting set of examples.