LTSPICE CPU/system benchmarks ?

[MathWizard]

I'm looking at the latest intel/AMD CPU's, considering an upgrade, mainly for gaming. I don't do video editing, or any other hard processing, except ltspice simulations.

Does anyone know of a thread or site with benchmarks of running spice sim's on different systems ?

What effect does the GPU have for that matter, when you finish a sim, and it sometimes lags while "drawing" traces ? GPU's cost a fortune so that's not a factor, since I'm a gamer anyways, but I never actually benchmarked a circuit and recorded the vales, on my current PC or other's ones I have/had.

[SiliconWizard]

What effect does the GPU have for that matter, when you finish a sim, and it sometimes lags while "drawing" traces ?

Probably none. It usually "lags" when there is a LOT of data generated by the simulation. And LTSpice's waveform viewer could definitely use a big overhaul. It's kinda buggy and inefficient.

[langwadt]

I suspect disc speed make big difference

[magic]

I think it's some preprocessing of the data for display but disk could be a possibility too if you manage to generate gigabytes of data (I never did, so far).
Check the size of the generated RAW files and check for CPU activity spikes when loading the results. At any rate, it probably is not a GPU bottleneck.

[AndyC_772]

The limiting factor with rendering results in LTSPICE is hard drive speed. The GPU is barely used.

Tip: always save your schematic .asc file to a folder on an SSD, then run the simulation. That way, the results get written to the fastest drive available.

[MathWizard]

Yeah I only have SSD and nmve drive these days.

So hasn't anyone bench marked spice sim on different CPU's ?

[SiliconWizard]

Even with very fast drives, LTSpice will always lag while displaying waveforms when there's a lot of samples. It's in the way the waveform viewer handles data, and we can't say for sure what the culprit exactly is without seeing source code... But what we can say is that some lag will occur even with very fast GPUs, SSDs and CPUs. As you may have noticed, the waveform viewers tends to trigger several screen updates in a row and it's particularly visible when there's a lot of data. The screen updating code is visibly inefficient for some reason and would need a big overhaul IMO, I've also encountered other screen updating issues.

So if you're looking for a fix, either wait until LTSpice gets improvements, or try making your simulations generate less data, which is a whole topic in itself.