Ascalaph Liquid - supercomputing on PC

Ascalaph Liquid outdated.
Please use the Abalone instead.



In 2007, NVIDIA introduced video cards that could be used not only to show graphics but also for scientific calculations. These cards include many arithmetic units (currently up to 240) working in parallel. Long before this event, the computational power of video cards was used to accelerate calculations. What was new is that NVIDIA made it possible to develop parallel programs in a high-level language. Actually, it is the C language supplemented with several key words. This technology called CUDA substantially simplified programming by introducing the CUDA code in programs written in C/C++.

Ascalaph Liquid GPU is a hardware accelerated version of Ascalaph Designer, identical in all features except performance. It requires special hardware and software. The program uses an NVIDIA card to speed up the computation of non-valent interactions.

Before purchasing the program, please pay attention to the following

  1. You must have a video card supporting CUDA 2.0 technology.
  2. To achieve substantial acceleration, the video card must have at least 12 multiprocessors.
  3. Such cards consume much energy, so your computer may need a new power supply unit.
  4. You must install a driver supporting CUDA 2.0 technology.
  5. No acceleration will be observed for very small models comprising 100-200 atoms, the computation may even slow down. Substantial acceleration is achieved for models consisting of thousands of atoms.
  6. Currently the program handles medium-sized models, limited by computer configuration. Thus with 2 GB RAM and 300 MB video card memory, you can work with models including about 5000 atoms; these specifications are regarded as minimal.


Benchmark analysis

0.1 ps dynamics of water box.

Water molecules 216512 7291728
Atoms 6481536 21875184
CPU Intel Core2 Quad Q6600 @ 2.40GHz, sec
GPU GeForce 8800 GTS (320 MB), sec
1.35 (x3.0) 2.46 (x7.8) 4.20 (x8.9) 10.7 (x19.0)
GPU Tesla C1060, sec
0.99 (x4.1) 1.95 (x9.9) 2.73 (x13.7) 6.98 (x29.1)