Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

How does the simulation speed scale with number of CPUs?

Status
Not open for further replies.

irfan1

Full Member level 5
Joined
Oct 26, 2003
Messages
294
Helped
25
Reputation
50
Reaction score
10
Trophy points
1,298
Activity points
2,872
I wonder how does the simulation speed scale with the number of CPUs for MWS?
I noticed something very odd. I have a intel dual-core CPU machine. Solver is set to use both CPUs. But when I check the CPU usage I noticed that almost only one CPU is active.
 

Hi irfan1


I noticed it to. I think that CST has a problem with multi processor
I have one HT CPU. And the CPU usage not passing the 40% with 1CPU its 100%

Can someone tell us what the CPU usage in HFSS 10 multi processor?

Maybe it depends on the algorithm of Time Domain.

PL
 

I think it depends on the complexity of the model. Most of the stuff I simulate only uses one processor on MWS once in the transient stage, yet the same models on HFSS v9+ hammer both processors. This doesn't however equate to a quicker simulation time (time vs freq domain).

Matt
 

irfan1 said:
I wonder how does the simulation speed scale with the number of CPUs for MWS?
I noticed something very odd. I have a intel dual-core CPU machine. Solver is set to use both CPUs. But when I check the CPU usage I noticed that almost only one CPU is active.
I used a dual-Xeon workstation for CST MWS programme simulation.
Simulation time was one and a half times less than that with one CPU.
 

It seems that for complex structures FIT algorithm can not take full advantage of multiprocessors. This seems to be a major disadvantage. And kind of makes me to beleive that due to this reason CST did not incorporate distributed computing (parallel computing). On the other hand FDTD can take full advantage of multi-processor. In fact, to my experience, the speed of FDTD simulations scale almost linearly with the number of cpus.
 

In my opinion the efficiency of having a compunter with more than one precessor depends only on how the simulator was implemented. If a simulator was not written to share the amount of calculation on more processors there will be little advantage in having a dual or more processor system. In my opinion it is better to have a fast processor with a large amount of memory.
Sonnet allows the sharing of calculation on more than one computer, you can see details in an article published on Microwave and RF.

bye
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top