Improving Simulation Speed (ISE and ISim)

[Pack34]

Does anyone have any tips on improving simulation speed? I remember National Instruments having a compiler farm option when I worked with LabView in college, is there some way I can network together a couple desktop computers together to compile and simulate faster? Or is that just not possible without coughing up a significant amount of money? If I have to put some money down for a more powerful computer, does anyone know how limited the software is to 4+ core processors? Is it intelligent enough to be able to divide up the workload over 8 cores, or is it only really optimized for a dual core setup? I know that ISE has been depreciated and it's moving to Vivado so I'm suspecting that compilation and simulation speed isn't going to be impacted by newer processors.

The ideal setup I'd like to run with would be to keep using a laptop to do all of my design and have a second machine handle the simulations and compiling.

Thanks for any help!

[rstofer]

You can do a Google search for this and you will find that even a 4 core 2 thread/core is overkill.  The software won't use it and, in particular, it won't use it during synthesis.  So, buying that 4 chip Xeon processor with 16 cores and 32 threads won't help at all.  Anyway, that's my summary of what I have read but you can run the search yourself.

About the only thing you can do is get as fast a CPU chip as possible.  I use a 2.8 GHz I7 (4 core dual thread) and it runs ok.  It's not speedy but it's the best I can do right now.

[Kilrah]

Yep, look for a processor that has the highest possible single-core performance, multiple cores are pretty useless for that particular use (you might still want a quad for other things or just to do something else while a build runs without significant slowdowns).

A full synth/implementation/bitstream run in Vivado that takes 11:45 on my decent desktop machine (i7-970, 6-core OC to 4GHz) only takes 2 more minutes on my wimpy core m7-6y75 ultraportable. Windows actually notices it's better off shutting down one core and pushing the other from 1.3 to 2.8-3GHz, it's the first time I've really seen the "Turbo Boost" in action for a significant amount of time.

[joelby]

Are you using the free Webpack licence? If so, ISim is artificially slowed down.

[6thimage]

For synthesis/implementation, you can enable multi-threading support (up to 4 threads for certain parts of the process), but as others have said single core performance is the main thing to look for.

For simulation, multi-threading is not used. As joelby said, isim on a webpack license does restrict the simulation speed - this is for larger designs and you will have a message appear in the log saying the speed is being limited. If you want to use a webpack license, but don't want this slow down, you can always use vivado - but be careful in case Xilinx have changed parts of their IP between versions.

[Pack34]

Are you using the free Webpack licence? If so, ISim is artificially slowed down.

That would explain it.