Increasing System RAM

I have started with the simplest possible PC upgrade in an attempt to start improving benchmark scores - since mypy is taking a ridiculously long time to run and I have previously observed it consuming large amounts of memory on my Mac, I hypothesized that increasing the PC's RAM beyond its current 8GB (6GB usable) should help improve mypy's performance.

I didn't want to invest too heavily in DDR4 memory since that will probably not be part of the PC's Mac-killing final form, so I purchased 32GB of Silicon Power DDR4 RAM and threw it in the PC.

Benchmark Results

mypy

The increased memory has indeed made a difference, although it is quite modest: with 29GB of system memory now available the mypy median execution time is has dropped from 502s to 475s, a 5.8% improvement.

Other Benchmarks

Unfortunately I failed to take into consideration the fact that even though the new memory was clocked faster (3200mhz compared to 2400mhz on the old Crucial DIMMS), the new memory also had much higher latency settings than the decent-quality Crucial DIMMS I removed (CL22 vs CL17). I also forgot to reset the motherboard memory clock to "Auto" so the new memory was running at 2400mhz like the old memory. This means that for benchmarks that weren't hitting the 6GB memory limit, they were likely to run slightly slower instead of faster. For example, the "pytest single" benchmark is now 1.4% slower on the new RAM:

The performance loss was smaller on other benchmarks - enough that I'm willing to consider the possibility that the latency settings have no effect and the performance loss is actually just variation within the normal margin of error.

Summary

Increasing total system RAM seemed to improve a couple of benchmarks by 5-6%, but other benchmarks were degraded, likely due to inferior memory speed/latency settings.

As a follow-up it could be worthwhile comparing different memory latency settings and/or memory speeds to determine whether they make a significant difference to overall benchmark scores.