Download this project!
Download the one sprite and one script of "speed test" and open it in Scratch Project Notes
PLEASE POST YOUR RESULTS!
Include:
Online or Offline
Browser type & version
CPU type and speed
Time elapsed
Execution speed
Whether anything else was running.
Do 2 or 3 runs if you have the patience - I got fairly variable results (+/- a good 3 Hz) on the same CPU & browser.
Particularly I want to see if the CPU affects it much, or if it's purposely limited inside the interpreter engine. A 3.4Ghz quad-core i7 running at 100Hz while an old 1.1 Celeron can only muster 10Hz will have a definite impact on how widely usable more complex programs will be (and how much we can expand in future)
trying to find out how fast the different interpreters run at. the offline editor appears to go at a rather strange 38.7Hz on my office PC (not sure if that's a CPU limit, or something built-in) both for loops and for sprite movements (which on that evidence I guess are "cost free").... what speed does it go at on the website?
(this shifts an invisible sprite by 9600 pixels - i.e. back and forth across the whole width of 20 lines. each time it hits the side it plays a drum note, when it's finished it plays a longer piano note and displays the time taken. divide 9600 into the displayed time to work out the movement / cycle speed / framerate)
This version now gives a better instruction lead in and a fully loquatious readout of both elapsed time and processing speed at the end, and both times and calculates more accurately.
Takes a while to get going and not exactly graphically lush, but you'll only have to run it once per PC and it's supposed to be as sparse as possible to get the best possible score - it's supposed to be a guide to the HIGHEST speed you can reasonably expect from the Scratch processor. I'd be very lucky to see 30+Hz for some of the other projects I've done here...
More Projects by tahrey
-
crapsteroids …
11 views -
crapsteroids …
7 views -
crapsteroids …
6 views -
crapsteroids …
4 views -
chordguitar_v…
12 views
1 | 2
Comments
You need to be logged in to post comments
Add a Comment
Goddamn it comment thinger. Work with me here, stop deleting what I write. <p> Celeron (III) 2.2Ghz, 128k cache, Online, IE7, 75Hz refresh: Processing speed 30.2 Hz, took 320 seconds.<p> I just don't know any more. Presumably it's aiming for 30fps on basic hardware but has very poor rate control? Bear that in mind if you're making anything time-critical. I'd make heavy use of the Timer instead of delay loops.
And though I've switched to 60Hz refresh (with an option on 70), I can't test what the effect of that would be because it's just showing "ERROR" in place of the app, even when I reload repeatedly. Gah. I love the Scratch concept but the flakiness of this website lets it down badly :( For starters, I can't seem to keep logged in for more than about 2 minutes at a time. Comments go missing. And its handling of formatting in the comment boxes is useless :( :(
Core 2 Duo (Hyperthreading), 2.5Ghz, Online, IE 7.0.5730.13: 31.9 Hz, 302 sec, hardly touching any of the four virtual cores. How strange. I may have to see if choice of browser, and screen refresh rate (this and the P4M are at 60Hz, the CelM was 75Hz) have much influence.
New Result: Office PC --- Celeron-M 1.86Ghz, Online with Firefox 3.0.7, second tab open in background (writing comment elsewhere), Folding@Home and a couple other apps (Outlook, on-access virus scanner) also running: 29.8 Hz (323.3 sec) Got an older 2.4Ghz Celeron (overall, noticably slower at most tasks) and a stonking 3Ghz Quad-core monster I intend to test it on later. So far: Cel-M 1.86 Offline: 38.7, Online 29.8 (77.0%) Pnt-M 1.73 Offline: 32.6, Online 24.0 (73.6%)
ok let's try again. Nothing much to "figure out" or go read up on. Simple idea - find out internal speed of Scratch system by running a set number of instructions, then dividing elapsed time into that number. Voila, execution speed :) I went for 9600 just because it was convenient (back & forth for 10 repeats) and reasonably large, though I could probably get away with just 960 for a 1/10th Hz accuracy level.
And the result this time: Very disappointing. It was already slow enough that a 1981 Sinclair running BASIC would embarrass it, but online is even slower. In Firefox 3.0.5, on the P-M 1.73, with about 5 secs extra delay for sound library loading: 23.5 instructions/second (411 secs taken). Assuming it was only 400 sec w/o the delay, that's still a rubbish 24.0Hz. Stick to conceptually simple but graphically rich stuff, I guess.
why does it erase my comment and say "comments cannot exceed 500 characters" when I'd only gone up to 470 by it's own length checker?
the comments system on here is so unreliable. i was trying to reply to annoyingboy but it just won't go. maybe i can trick it into working by posting here first then deleting after.
I'm gonna have to read the math book now. How did you figure this out?!
BTW if you meant getting it to display the results ... well, download the code and have a look. Everything I've made so far has been purely experimental, based off other peoples / example scripts, or just throwing blocks together and seeing what sticks - I haven't looked in any instruction guide yet.
(view all replies)