Click the images for full-size renditions.
1FJF.pdb is a large 52,000 atom PDB file -- I thought it would be fun to see how each of the various programs copes with it, and so here is what I found:
DISCLAIMER: I am NOT an impartial judge. Also, the following comparisons were made rather quickly, so I don't guarantee thoroughness or perfect accuracy. Treat what you read here with some skepticism until you've verified the results yourself. Please let me know if you find any substantial errors on this page, since comparing programs can be a touchy subject, and I wouldn't want any factual errors to persist. Hardware: 1 Ghz Athlon, 256 MB RAM, nVidia GeForce2 GTS.
![]() |
PyMOL (0.63) loads 1FJF.PDB file in under 4 seconds and delivers smooth rotations (wire-frame). It can compute a high-quality surface in about 2.5 minutes, even on a machine with insufficient RAM and still give 2-3 frames/sec rotation.
![]() |
OpenRasMol (2.7.1.1) loads the structure even faster than PyMOL and can draw wire-frame almost as fast, but the quality is about what you would expect from a software renderer. As always, RasMol does take first place with VDW sphere (CPK) rendering (3-5X PyMOL). As far as I know, there are no surfaces in RasMol and no Python interface. In my opinion, the age of software-only renderers has ended now that most computers ship with hardware accerelated graphics built-in. With its free status, broad capabilities, and open architecture, PyMOL has a good chance of becoming the "Super-RasMOL" of the future.
![]() |
MOLMOL (2.6, for Windows) loads the structure very quickly and can draw wire-frame as fast as anyone. Update speeds for the other representations lagged behind PyMOL and VMD. However, the main problem with MOLMOL under Windows with a GeForce card is that the binary distribution only works right if you are logged in as Adminstrator (otherwise the renderer is crippled). MOLMOL wasn't able to compute a surface for this molecule after 15 minutes of churning. Other than PyMOL, MOLMOL has one of the best open-source licenses out there, since you can distribute modified versions of MOLMOL! Unfortunately however, there are a whole bunch of GPL-like conditions applied to the code which would keep me from ever working on it.
![]() |
DeepView/SwissPDB Viewer (3.7b2) is free, but closed-source. It loads 1FJF in a flash and computes a crude molecular surface in just a few seconds. The problem with SPDBV is that the quality isn't great, and you can't force it to draw all atoms during object rotation, so some of the 3D effect is lost. The slab mode is a bit cumbersome. With all of its other amazing capabilities, this is one program that would really make a difference if it were open-source so that the rest of us could work on it!
![]() |
Visual Molecular Dynamics (VMD) (1.7) can be scripted from Python and it matches PyMOL peformance for wire-frame. This isn't surprising since VMD was used as a performance standard during PyMOL development. PyMOL and VMD remain neck-and-neck when it comes to performance for more complex representations, including VDW spheres. However, VMD's surface module crashes (under Windows) if you try to use it on this molecule
![]() |
Chimera (ver. 1, build 1298) is UCSF's Python-enhanced viewer and shares much of PyMOL's vision. NOTE: this viewer is not free for companies, and it can't be redistributed! Chimera takes several-fold longer to load the file than PyMOL. Performance is hard to judge: some operations are much slower (anything involving selections), but a few representations are equal or slightly faster in Chimera. Chimera crashes when you try to create a surface for this structure.
![]() |
Python Molecular Viewer (PMV) (1.0) matches PyMOL's drawing performance once the molecule is loaded, but it takes OVER TWO MINUTES to load it on a 1 Ghz Athlon PC. Performance optimization will need to be a big priority if this package is to succeed. Like Chimera and VMD, MSMS (PMV's surface routine) crashes PMV under Windows for this structure.
![]() |
gOpenMol (2.0.1) can load the structure, but is noticably slower to render wire-frame than many of the others. gOpenMol has a very interesting license. This package is so feature-rich that I wasn't able to determine whether or not I could get a surface calculated! This massive 20+ MB collection of interesting functionality merits a closer look. perhaps some if it can be exposed from Python.
![]() |
QMol (1.91) is free, but not open-source or Python enabled. I'm including it because it is the fastest no-frills 3D viewer I've come across. The default ribbon in QMol renders with liquid motion (25+ frames/sec) with 1FJF and looks pretty cool. No surfaces, but if all you need to do is grab a PDB (of any size) and look at the overall structure, then QMOL is your program. With a few more features (front/back clipping planes, adjustable depth-cue for instance), QMol will surely dominate the free-but-closed-source category.
![]() |
Biodesigner (0.60) is free to academics only, is closed-source, and is Windows-only. It takes OVER THREE MINUTES to load this file on a 1 Ghz machine! Wireframe rendering performance was noticably behind the leaders. Nevertheless, Biodesigner presents one of the most beautiful and "slickest" molecular graphics environments I've seen, complete with a molecular browser and sequence editor. Unfortunately there are a couple of other major problems this version: 1. poor stability (many functions don't work) and 2. the software hogs the CPU through "busy-idling", whichs makes it quite difficult to do anything else on your machine when running Biodesigner. There doesn't seem to be any surface representation or any kind of extensibility.
![]() |
WebLabViewerLite (4.0 - a closed-source, commercial loss-leader) was totally incapacitated by 1FJF. I put it out of its misery after it wedged for 15 minutes. Nicely done MSI/Accelrys!
The Conclusion? PyMOL meets or beats every other free viewer for looking at large structures, but it is the only program you can modify or extend from Python and then share with your friends! I think the choice is obvious...