JSmol/Rotation Speeds

From Proteopedia

Jump to: navigation, search

JSmol is the default form of Jmol used in Proteopedia and FirstGlance in Jmol. JSmol is used here to mean the HTML5/Javascript (non Java) application working within a web browser (see Jmol has four forms). The "JS" in JSmol refers to JavaScript. As mentioned in JSmol Notes, JSmol is substantially slower than the Jmol Java applet. With the exception of performance speed, both forms of Jmol have identical capabilities.

Performance of JSmol depends on the underlying performance of the javascript in the browser being used. With smaller macromolecules (under about 2,000 atoms) performance is generally good in all browsers except Internet Explorer. With larger macromolecules, the choice of browser becomes more important for optimal performance.

Browser javascript performance varies as new browser versions are released. Below are rotation speed results obtained at various times.
See also Browser popularity worldwide.

Contents

Conclusions

Update April, 2021

Versions current in April, 2021, of the Chrome, Edge, Firefox and Opera web browsers all seem equally good for JSmol. Safari is acceptable, but rotation of solid (spacefilled) renderings is substantially slower/choppier than the others, especially at high zoom levels.

Update January 2019

Differences from July, 2017:

  • Chrome is now faster than Firefox; slightly faster on Mac but 2-fold faster on Windows.
  • Opera now performs nearly as well as Chrome (in both Windows and macOS).
  • Safari is now 5X slower than Chrome for spacefilled rendering. This is true in OS 10.10 and 10.14.

As before:

  • Waterfox is comparable to Firefox for JSmol, and very fast with Java.
  • Edge is about 5X slower than Chrome.
  • Internet Explorer is 10X to 30X slower than Chrome.
TEST RESULTS
Frames/Second for spinning in cartoon, spacefill.
Operating System Chrome Edge Firefox Internet Explorer 11 Opera Safari Waterfox
Windows 10 3.4, 3.0 0.5, 0.7 1.5, 3.4 0.35, 0.1 2.5, 4.2 n/a n/d
Mac OS 10.14 or 10.10 3.6, 3.6 n/a 2.3, 4.0 n/a 3.5, 3.7 3.8, 0.7 2.3, ~5.0
~7.0*, ~7.0*

"n/a": not available. "n/d": not determined.
JSmol without Java. Data for December 30, 2018. Bad performance. Poor performance.
* Running Jmol in Java, not JSmol. Tested only in OS 10.14. Installing and enabling Java.
Methods same as March, 2017, except zoom for spacefill was 260%.
Rendering quality was low (antialiasdisplay false).

FirstGlance in Jmol

Previously, this section reported that FirstGlance in Jmol version 2.74 (and several earlier versions) was very slow to load and display large RNA structures. This was due to a bug in FirstGlance that was fixed in version 2.8 released May 29, 2019.

July, 2017

Differences from March, 2017: none. From May, 2016: minor.

  • Windows (JSmol, no Java):
    • Firefox and Chrome perform similarly. (Cartoon rotation: Chrome outpeforms Firefox, about 2x more frames/second. Spacefill rotation: Firefox outperforms Chrome also about 2x.)
    • Edge: NOT RECOMMENDED. Very sluggish. Spinning and rotation by mouse are very jerky.
    • Internet Explorer: UNACCEPTABLE, extremely sluggish, with spining/rotation extremely slow and jerky.
  • Mac OS X (JSmol, no Java):
    • Firefox performs best.
    • Chrome: Performance very close to that of Firefox. Both spinning and rotation by mouse are a slightly jerkier than in Firefox (less than two-fold difference).
    • Safari: Overall performance is good. Spinning and rotation by mouse are a little smoother than Firefox for cartoon rendering. For spacefill rendering, spinning was jerkier than Firefox by about three-fold.

March 27, 2017

JSmol (Javascript)

All browsers were updated to the versions current at the time of testing (March, 2017). Tests below were done in FirstGlance in Jmol in the default HTML5/javascript mode. In order to slow down rotation to enable manual counting of frames/second, a large molecule was used: 1g3i, a single model with 45,528 atoms (which puts it in the largest ~2% of entries in the PDB). Tests were done in a browser window approximately 1,900 x 1000 pixels on a late 2014 MacBook Pro (2.2 GHz Intel Core i7 with 1600 MHz DDR3 RAM). Windows 10 was a virtual machine (VMware) on the Mac, running at native speed on the Mac's Intel processor. For spinning, frames/second were determined at zoom 170%. For rotation by mouse, frames/second were determined at zoom 100%, cartoon rendering.

Frames/Second for spinning in cartoon, spacefill; for rotation by mouse in cartoon.
JSmol without Java. Data for March 27, 2017.
Operating System Firefox Safari Chrome Internet Explorer 11 Edge
Windows 10 3, 6; 3.4 n/a 2, 2; 3.3 0.4, 0.14; 0.5 1.0, 1.3; 1.5
Mac OS 10.10.5 3, 6; 3.6 4, 1.7; 5 2, 4; 3 n/a n/a

"n/a": not available.


Older Results

Everything below is partially obsolete. It remains here only for historical reference. In December, 2015:

  • Firefox and Safari outperform all other popular browsers for smoothness and speed of rotation, and therefore are the recommended browsers for websites that use JSmol such as Proteopedia and FirstGlance in Jmol.
  • Chrome and Opera rotate molecules in JSmol about 4-fold slower than Firefox (or Safari) for cartoon rendering, and almost 2-fold slower for spacefilled rendering. That is, they generate fewer frames/second, so rotation or spinning is less smooth or more jerky.
  • Edge (the new browser in Windows 10) is almost 4-fold slower than Firefox for both cartoon and spacefilled renderings.
  • Internet Explorer 11 gives unacceptably slow rotation of molecules in JSmol, with a performance about 9-fold slower than that of Firefox.


Historical note: In 2014, Chrome outperformed Firefox. Changes in Chrome in 2015 reduced its performance for rotating molecules in JSmol. In December, 2015, a change in Chrome (first in version 47) made it unable to rotate molecules in JSmol satisfactorily. As the mouse moved, the molecule remained fixed until the mouse movement stopped, whereupon the molecule jumped to the final position in one step. Bob Hanson was able to work around the change in Chrome with a modification to JSmol, but frames/second are still poorer than they were in Chrome in 2014.
  • The Jmol Java applet is only 1.4-1.8 fold faster in Firefox, but 20-fold faster in Internet Explorer.

Data (December, 2015)

JSmol (Javascript)

Tests below were done in FirstGlance in Jmol in the default HTML5/javascript mode. In order to slow down rotation to enable manual counting of frames/second, a large molecule was used: 1g3i, a single model with 45,528 atoms (which puts it in the largest ~2% of entries in the PDB). Tests were done in a browser window approximately 1,900 x 1000 pixels on a late 2014 MacBook Pro (2.2 GHz Intel Core i7 with 1600 MHz DDR3 RAM). Windows 10 was a virtual machine (VMware) on the Mac, running at native speed on the Mac's Intel processor. With spinning on, the number of frames in 10 or 20 sec was counted and used to calculate frames/second. Browsers were the versions current in December, 2015.

Frames/Second in Cartoon, Spacefill
JSmol without Java
Operating System Firefox Safari Chrome Internet Explorer 11 Edge Opera Maxthon Torch*
Windows 10 3.9, 4.3 n/a 1.05, 2.7 0.45, 0.45 1.1, 1.25 1.0, 2.4 1.5, 2.5 1.6, 3.0
Mac OS 10.10.5 4.2, 4.1 4.0, 3.3 1.05, 2.95 n/a n/a (failed) 4.2, 3.5 n/a

"n/a": not available.
* Despite a high review, Torch seems a rather iffy browser. It is accused of hijacking and does not identify itself in navigator.userAgent.

Jmol (Java applet)

Because Java is so much faster than Javascript, the spinning was further slowed by clicking the "Zoom Larger" button in FirstGlance twice.
(Torch does operate Java in Dec. 2015.)

Frames/Second in Cartoon
JSmol without Java - Jmol Java applet
Operating System Firefox Safari Chrome Internet Explorer 11 Edge Opera Maxthon
Windows 10 3.2, 4.7 n/a No Java 0.3, ~7 no Java n/d 0.2,
Mac OS 10.10.5 3.3, 6 3.0, ~7 no Java n/a n/a (failed) No Java

"n/a": not available. "n/d": not determined.

See Also

Proteopedia Page Contributors and Editors (what is this?)

Eric Martz

Retrieved from "http://proteopedia.org/wiki/index.php/JSmol/Rotation_Speeds"
Personal tools