smallpt is a 99 line OpenMP path tracer (an embarrassingly parallel problem, so scales well) written in c++. Simple to compile (read first couple of lines of source for instructions, leave out the -fopenmp for single threaded version), no dependencies, its raw number crunching (no other optimisations applied other than OpenMP, not even SSE), resultant PPM can be viewed in gimp. You could even unplug your graphics card since its not needed (after, you would perhaps need to plug it back in so you can view the image 😵).

http://www.kevinbeason.com/smallpt

And you get a nice stereotypical cornell box scene for your desktop.

^^ Oh neat! I'll definitely give that a try. Always liked oldschool CPU-driven raytracing. Thanks for putting me on to that.

There's a big thread of 7zip benchmarking across lots of CPUs. Very threaded.
7-zip benchmarks for over 60 CPUs

I have a primes number searching program that I made that can theoretically handle up to 4,294,967,295 threads. It is called Fastest Primes.

You can set number of threads, work unit size, number to calculate to or number to calculate from-to.

The work size unit should be sized to stay in the CPU cache for the best times. If you try to calculate too large of a range it will use your swap file and slow waaaay down.

It is a pretty good for testing systems against each other as far as CPU and RAM speed/latency goes. If you have 2 similar systems, but one CPU is clocked a bit higher and one has the RAM clocked higher, the one with the faster RAM will win.

I've never tried it on an older system so not sure it will even run on a P4, but it should.

It is only compiled for Windows as getting it to work on both Linux and Windows took too much time because of the differences in the calls, threading libraries and compiler settings.

There are both 32-bit and 64-bit binaries available. 64-bit is faster.

https://sourceforge.net/projects/fastestprime … ource=directory