xccel
时间：03-24
整理：3721RD
点击：
The following info is from Microwave Journal, September 2008:
-----------------------------------------------------------------------------------
The new 3D solver EMPIRE XCcel 5.2 offers the Perfect Geometry Approximation (PGA) algorithm to yield more accurate results for curved structures, even with coarse meshes. The most complicated designs are modeled quickly and precisely. The unique on-the-fly code generation for each model and processor produces up to 1600 million FDTD cells per second on a conventional PC supporting the complete RAM. Thus very complex structures can be modeled very fast and very accurate.
-----------------------------------------------------------------------------------
I believe if Empire XCcel 5.2 provides PGA, which seems to be something like PBA of CST, with that speed, i.e. 1600 million FDTD cells per second, or even half of that, it?s going to make things real difficult both for CST MWS and its accelerating boards.
-----------------------------------------------------------------------------------
The new 3D solver EMPIRE XCcel 5.2 offers the Perfect Geometry Approximation (PGA) algorithm to yield more accurate results for curved structures, even with coarse meshes. The most complicated designs are modeled quickly and precisely. The unique on-the-fly code generation for each model and processor produces up to 1600 million FDTD cells per second on a conventional PC supporting the complete RAM. Thus very complex structures can be modeled very fast and very accurate.
-----------------------------------------------------------------------------------
I believe if Empire XCcel 5.2 provides PGA, which seems to be something like PBA of CST, with that speed, i.e. 1600 million FDTD cells per second, or even half of that, it?s going to make things real difficult both for CST MWS and its accelerating boards.
Sounds a little (too)fast to me ;)
Dear rfmw,
Although I've not tested EMPIRE XCcel to verify the high speed claims for myself, I have recently found some comments by CST which indirectly referred to EMPIRE and "its high speed" but meanwhile tried to downplay it by emphasizing the lack of conformal FDTD mesh in it which effectively meant "less speed compared to CST for the same accuracy". It made me think that there might be some truth in IMST's claims. Please note that on a dual quad-core Intel XEON X5482, for simulations including PML, CST provides a speed of less than 30 million cells per second and for other scenarios, it could go up to something around 60 million cells per second.
Regards