MATLAB benchmarks for i7-12700K vs i9-12900

Hello,

I’m planning on purchasing a Dell XPS 8950 desktop with an i7-12700K CPU, RTX 3060 Ti GPU, 16 GB RAM DDR5. I will be using it mainly for medium-heavy computations on MATLAB (with the max clock time of around 3-4 weeks), so was wondering if there are any MATLAB bench scores available for this configuration or similar.

Also, how do/Will they compare with an i9-12900 that has 8P + 8E cores (4E cores more than the i7-12700K)? Will there be any significant difference in processing speed between the two?

Thanks!

[Walter Roberson]

Single precision or double? https://www.phoronix.com/review/nvidia-rtx-3060ti/7

[Vikram Sudarshan]

Thanks! Yes, these are great benchmarks for the GPU. Any idea regarding the benchmarks for the 2 CPUs, especially on the MATLAB bench? I've seen various cinebench and geekbench scores and there doesn't seem much considerable improvement for the i9 over the i7 (~ 10% max).

Answer 1

[Rik]

That will depend on the exact code profile you're going to use. Some code benefits from multiple cores, some does not. Some computations are faster when transferring the computation to the GPU, others are not.

The only way to know for sure is to create a benchmark for your own code and run it on both machines.

That said, there is a function called bench that will benchmark your computer, so you may consider searching for its results.

The YouTube channel Hardware Unboxed sometimes includes Matlab in their benchmarks. You might have some luck there. I doubt 4 efficiency cores will have much impact on any Matlab code except the most parallelized code (perhaps only functions like conv).

[Vikram Sudarshan]

Thanks! I usually do medium-large-scale simulation of stochastic ODEs. Lately, I’ve been trying find simple ways for parallelizing my code (I.e, trying ways to use the parallel computing toolbox), but haven’t been effective (mainly due to loop dependency, as is the case with numerical integrations)

Any idea how can one use Parallel Computing Toolbox for numerical integration of ODEs/SDEs?

[Rik]

I would look in a different thread for that question. I don't solve ODEs myself, so I don't know if or how to parallelize that.

[Vikram Sudarshan]

Thanks! I figured it out. For ODEs, direct parallelization is possible only when running multiple, independent dynamic simulations; for solving each ODE, direct parallelization doesn't seem to be possible.