knyazik wants a model kit (split thread)

[knyazik]

This might be a little bit off topic, but I remember back in college we had a lot of different colored wooden spheres that had different amount of holes in them that would represent different elements. I remember black was carbon, white was oxygen, etc. And you could connect them with either wooden sticks or springs, which we used in a lot of our chemistry classes. I cannot remember what they are called. If someone knows what they are called, or even better has a link to where I can read up on them/buy them, I would be really grateful.

[hypervalent_iodine]

What you want is a model kit. Any university book shop will have them, or you can find them online.

[Sensei]

This might be a little bit off topic, but I remember back in college we had a lot of different colored wooden spheres that had different amount of holes in them that would represent different elements. I remember black was carbon, white was oxygen, etc. And you could connect them with either wooden sticks or springs, which we used in a lot of our chemistry classes. I cannot remember what they are called. If someone knows what they are called, or even better has a link to where I can read up on them/buy them, I would be really grateful.

 

This 245 balls with 310 connectors set costs $140

http://www.eduvis.pl/oferta/chemia-pomoce-dydaktyczne/chemiczny-zestaw-nr-27-modele-kulkowe-detail

 

Google for "ball and stick model"

http://en.wikipedia.org/wiki/Ball-and-stick_model

or "molymod"

http://www.molymod.com

 

ps. Stop bringing up ancient threads to top.. It's unwelcome behaving on the all forums..

Edited October 23, 2014 by Sensei

[hypervalent_iodine]

!

Moderator Note

Just for the record, I've split this into its own thread.

Knyazik, if you think something is off topic, please start a new thread.

[knyazik]

Thank you for splitting me to a different topic hypervalent. I'm pretty new here, so didn't know how to go about this, so I appreciate you letting me know. And thank you Sensei for the links. That picture is exactly what I wanted!

[hypervalent_iodine]

Not a problem. Glad we could help you with your question.

[knyazik]

Just wanted to update you guys, the kit that I got is awesome. Thank you so much for helping me. Definitely brought back college memories

[Enthalpy]

You may want to have a molecule modeling software as well. Less a hand pleasure than plastic balls, but it has advantages as well, provided you don't overtrust it. This one is decently easy to use and is free:

http://www.arguslab.com/arguslab.com/ArgusLab.html

many more exist.

[Sensei]

Just wanted to update you guys, the kit that I got is awesome. Thank you so much for helping me. Definitely brought back college memories

 

What you bought? Show link to your new toys..

Edited November 13, 2014 by Sensei

[knyazik]

This is the kit I ended up getting:

 

 

http://www.amazon.com/Molymod-Inorganic-Organic-Molecular-Student/dp/B005NWGISS/

 

Now I have plenty of things to try. I haven't downloaded the software yet, but thank your for that suggestion, Enthalpy. I anticipate that using a computer simulation, I can make some crazy biochem molecules. Wonder if you spent enough time on it, if you can actually form a model for the DNA... probably an unrealistic task that would take a lifetime to accomplish.

[Enthalpy]

Wow, nice toy!

 

Dna: the modelling algorithms used by Arguslab (every programme uses the same ones more or less) take a computation time that increases faster than the number of atoms, possibly as the square. 10 atoms run, 30 atoms are slow, 100 atoms are already too much.

 

Faster methods should check first which atoms interfere or not, but most programmes don't, alas. An other option models atoms as mechanical balls that can stick; far less accurate, this has the capacity to model poteins and their interactions.

[Sensei]

Dna: the modelling algorithms used by Arguslab (every programme uses the same ones more or less) take a computation time that increases faster than the number of atoms, possibly as the square. 10 atoms run, 30 atoms are slow, 100 atoms are already too much.

 

Programmer who wrote it didn't know how to program OpenGL efficiently. He probably used OpenGL v1.0 API..

OpenGL v1.0 requires sending polygon data every frame.

But newer OpenGL have many features that allow uploading data to GPU memory once, and stay there, so spinning,zooming viewport (replaying them from GPU memory) should be extremely fast.

 

On my ancient gfx card (it's maybe 4 years) my game is rendering at 50 millions triangles per second. Which is 1.666 mln triangles per frame at 30 FPS, so it's pretty smooth.

When we will have high quality atoms-spheres like 24x12 segments = 528 triangles per sphere. 1.666 mln triangles / 528 triangles = 3155 atoms at full quality, at full FPS..

 

With this Fullerene, which has 476 carbons, it goes quite smooth (probably something like 15 FPS) at ball resolution set to normal (~224 triangles per sphere).

It's still too small to say it's efficient. It could run 30 times faster than that.

Edited November 15, 2014 by hypervalent_iodine

[Enthalpy]

It's not a matter of OpenGL... And believe it or not, molecular modelling software does use OpenGL, DirectX and the like, efficiently.

 

Molecular modelling tries to compute the structure of a molecule (bond distances, angles, and with some models, electron distribution) from the atoms and bonds given by the user. This is the difficult part of the task, not image rendering.

[Sensei]

It's not a matter of OpenGL... And believe it or not, molecular modelling software does use OpenGL, DirectX and the like, efficiently.

 

If I would show two C/C++ source codes for OpenGL v1.0 API, and OpenGL v2.0+ API, would you be able to tell me from memory which is which?

 

I am specialist from optimizations...

So I can judge by myself whether code is running efficiently and when it's not.

I was surprised by your comment that's so slow working with 100 atoms, and especially downloaded software to check it out by myself. Don't take it personally, you didn't write it after all.

On my Core i7 CPU it's working quite fine with ~500 atoms Fullerene (deleting atoms and adding new one is quick).

 

You can quickly calculate quantity of triangles per second rendered by Arguslab while spinning/rotating viewport (while NO MODELING).

Calculate quantity of triangles per single sphere, bond, multiply by quantity of atoms, multiply by frames per second, and you have total quantity of triangles code is rendering per second.

I am receiving approximately 1.5 mln triangles per second.

From other experiments with my own code, I know that's approximately rate when OpenGL v1.0 API is used on my gfx card.

 

After using OpenGL v2.0 API or higher, we can have 50 mln triangles rendered per second. 30 times faster than with OpenGL v1.0 API.

 

See by yourself - load the same FullereneTube.agl, it's in C:\Program Files\ArgusLab\Samples

by default it's loaded with no atoms-spheres, just bonds are displayed.

Spin viewport, rotate, zoom it. It's going with 30 FPS on my machine.

Then press Settings > Display Options,

and pick up Rendering Style > Ball Cylinder > High.

And then spin, rotate, zoom viewport. It's going with 1 FPS or so.. Jumping.

 

(don't click Fast Render While Moving, it'd be cheating)

 

Try editing this Fullerene object (adding new atoms, removing existing) when Rendering Style is set to Wireframe, and when it's Ball Cylinder High.

This is *all* due to using OpenGL v1.0

 

Molecular modelling tries to compute the structure of a molecule (bond distances, angles, and with some models, electron distribution) from the atoms and bonds given by the user. This is the difficult part of the task, not image rendering.

Such calculation has to be done ONCE. Literally ONCE. When user clicked f.e. "delete atom", or added new atom.

It's not infinitely going process.

With Arguslab we can see significant slow down while spinning/rotating viewport, when no recalculation of bonds, angles etc is required.

 

ps. Analogy for engineer: you have two engines, one is consuming 1 L of fuel per 100 km, and second one is consuming 30 L per 100 km. Same weight of car. Same velocity. Which one is more efficient.. ? You don't need to be rocket scientist to answer that question...

Edited November 15, 2014 by Sensei

[knyazik]

 

If I would show two C/C++ source codes for OpenGL v1.0 API, and OpenGL v2.0+ API, would you be able to tell me from memory which is which?

 

I am specialist from optimizations...

So I can judge by myself whether code is running efficiently and when it's not.

I was surprised by your comment that's so slow working with 100 atoms, and especially downloaded software to check it out by myself. Don't take it personally, you didn't write it after all.

On my Core i7 CPU it's working quite fine with ~500 atoms Fullerene (deleting atoms and adding new one is quick).

 

You can quickly calculate quantity of triangles per second rendered by Arguslab while spinning/rotating viewport (while NO MODELING).

Calculate quantity of triangles per single sphere, bond, multiply by quantity of atoms, multiply by frames per second, and you have total quantity of triangles code is rendering per second.

I am receiving approximately 1.5 mln triangles per second.

From other experiments with my own code, I know that's approximately rate when OpenGL v1.0 API is used on my gfx card.

 

After using OpenGL v2.0 API or higher, we can have 50 mln triangles rendered per second. 30 times faster than with OpenGL v1.0 API.

 

See by yourself - load the same FullereneTube.agl, it's in C:\Program Files\ArgusLab\Samples

by default it's loaded with no atoms-spheres, just bonds are displayed.

Spin viewport, rotate, zoom it. It's going with 30 FPS on my machine.

Then press Settings > Display Options,

and pick up Rendering Style > Ball Cylinder > High.

And then spin, rotate, zoom viewport. It's going with 1 FPS or so.. Jumping.

 

(don't click Fast Render While Moving, it'd be cheating)

 

Try editing this Fullerene object (adding new atoms, removing existing) when Rendering Style is set to Wireframe, and when it's Ball Cylinder High.

This is *all* due to using OpenGL v1.0

 

Such calculation has to be done ONCE. Literally ONCE. When user clicked f.e. "delete atom", or added new atom.

It's not infinitely going process.

With Arguslab we can see significant slow down while spinning/rotating viewport, when no recalculation of bonds, angles etc is required.

 

ps. Analogy for engineer: you have two engines, one is consuming 1 L of fuel per 100 km, and second one is consuming 30 L per 100 km. Same weight of car. Same velocity. Which one is more efficient.. ? You don't need to be rocket scientist to answer that question...

 

Thank you for that detailed explanation. I am not that good at programming, so unfortunately did not understand everything that you said. I was wondering if you had a different recommendation for simulation software that would not be buggy, would be easy to setup, would be relatively easy to use, and would be free. Thank you for your suggestion.

[Sensei]

 

Thank you for that detailed explanation. I am not that good at programming, so unfortunately did not understand everything that you said. I was wondering if you had a different recommendation for simulation software that would not be buggy, would be easy to setup, would be relatively easy to use, and would be free. Thank you for your suggestion.

 

For your purposes Arguslab should be fine.

Remember we're talking about situation with hundreds or thousands atoms.

I wouldn't call it buggy - it's simply not using higher version API, but lower.

Bug is behaving abnormally, or crashing application, disallowing any usage.

Edited November 17, 2014 by Sensei

[Enthalpy]

 

If I would show two C/C++ source codes for OpenGL v1.0 API, and OpenGL v2.0+ API, would you be able to tell me from memory which is which?

 

I am specialist from optimizations...

So I can judge by myself whether code is running efficiently and when it's not.

I was surprised by your comment that's so slow

 

[...]

Such calculation has to be done ONCE. Literally ONCE. When user clicked f.e. "delete atom", or added new atom.

It's not infinitely going process.

With Arguslab we can see significant slow down while spinning/rotating viewport, when no recalculation of bonds, angles etc is required.

 

You have not understood.

 

The difficulty is not 3D rendering. Sorry I have to repeat.

 

Molecular modelling software computes atom positions (or tries to) according to the atoms' interactions. This cannot be done once: in ring or cage molecules, the bonds don't have their usual length nor angles. Steric hindrance prevents it as well. Even the links to other atoms influence the length to one atom. This is exactly what users expect from such a software. For constant bond length and directions they'd take any general 3D drawing software.

 

Obviously, you ignore what molecule modelling software does, and much about molecules as well. Why make definitive claims on the topic then?

 

Molecular modelling is the difficult task. 3D display is negligible in comparison, and no-one cares whether this trivial part of the job is optimized or not.