Nice one Brian... though this includes only CFD part and not FEA...
[David Roccaforte from Fisher has done a great work with Bench-marking]
I Have the same problem. It seems to me that Solidworks is managing the hardware resources VERY badly. I have a BOXX workstation with 32 Gb RAM, 3.3GHz and 12 cores which allegedly is the best Solidworks workstation outthere specially built for speed and memory. But the reality is different than advertised. In Flow Simulation it works fine, the whole 12 cores are used at maximum (100%) so I can't complain here. But in Non-Linear Simulation and Motion Simulation, Solidworks barely is using 8% of CPU.When I complained here about that Simulation is using only 6 cores out of 12, the reply was that it should work with all available hardware resources. It should but it doesn't. Even those 6 cores are not used at maximum, but barely 20% Of each core. So I had a feeling that in reality Solidworks is using only one core or just a little more which is the reason why is dividing the task on several cores. 100% divided by 12 = 8.3% which is exactly the usage in my system when solving simulations.
I understand that for rebuilding the parts is a linear process which does not allow to use multithreading, but Simulation and Motion Simulation can use multithreading, just as Flow Simulation is using.
The management is asking me why it takes so long to solve a simulation?
My answer is :because Solidworks is crashing all the time, in the middle of a simulation or when saving the file and I need to do all over again, and because it is not using the hardware at the maximum capacity.
So now the management is considering buying HP workstations 2.9 GHz dual 6 cores which are also advertised as best opyimised for simulation, but still this will not solve the problem. The problem is in the software, the Solidworks memory management, so please Solidworks programmers/developpers please fix this problem urgently!
To be honest I love SolidWorks Simulation & Flow (Cosmos product's) but would not disagree with you... For CFD, it's fine and also we can control allocation of core in hardware. As you said, developers must look into including more features to control the hardware for FEA Simulation through Simulation Premium license.
I would like to add here, graphics card also plays major role when we work on mesh refinements...
for both you and andrei, your answers are in the solidworks kb. solidworks did testing with all products and wrote an article about which benefit from more ram, more processors..etc.
from my experience though, you need to look at this in a different way.
yes, you can throw hardware at the problem. ssd, make sure you have enough ram to fit your whole problem in ram, the fastest core speed (vs number of cores), and the number of cores that you can afford, and then tune your system to prioritize the solve, not use page file, not have anything else running...etc.
when you go deeper than that, anything that benchmarks faster from a hardware perspective will make simulation go faster.
BUT, if you're looking for orders of magnitude improvements, you're not going to get that. that only comes from simplifying the problem and solving what needs to be solved. we spend a LOT of time with customers on this. for example we had a customer that his simulations were taking weeks, we brought that down to hours. he was considering spending 10k on a computer to solve it faster and my guess is that it would have only been a little faster.