-
Which is to say that the biggest perfomances differences come from:
Faster local storage
Faster networking locally and to the internet
More physical cores on the CPU
More total memoryThat list is very dependent on application. Currently using a 24 core 3960X with 128GB RAM for work (3D modelling) and it's about half as fast as my i9/32GB. Even a new i5 would be significantly quicker. Network and storage speeds also irrelevant in this case, even loading and saving is CPU-bound.
Brun-
Yeah, I am noticing that now, I am converting thousands of images in Lightroom and thought I move them onto the local SSD instead of the network drive, doesn't make a bit of a difference.
-
Network and storage speeds also irrelevant in this case, even loading and saving is CPU-bound.
Since 2005 CPU and memory speeds haven't improved: https://colin-scott.github.io/personal_website/research/interactive_latency.html
What's improved have been storage speeds (SSDs got faster, SSDs replaced HDDs) and networking speeds.
The only big difference on a CPU has been the number of physical cores provided.*
We've gone from Moore's law to Amdahl's law.
* And heterogeneous CPUs with mixed types of cores making "a core" (almost all recent Intel, AMD, and ARM designed chips like the M1) and also shared virtual memory which allows a single piece of on-processor memory to be exposed to multiple processing units (i.e. certain Intels, AMD Ryzen, Apple M1).
sohi
Velocio
You're into such minutaie that it's all irrelevant.
The CPU has 2 memory channels, meaning the 4 DIMMs work in pairs for the CPU. If you only have 2 DIMMs to put into the machine then they need to go into specific DIMM slots such that each CPU channel gets 1 DIMM. If you have 4 DIMMs then 2 supply the 2 CPU memory channels.
But there is no optimal here... it's all good enough.
The more pressing question is whether you want to purchase 4 x 16GB or 2 x 32GB. The latter gives you an upgrade path in future for a small premium now, the former is maxing out the upgrade path for those components... slightly cheaper now but way more expensive to change in future.
The other arguments seem excessively pedantic... memory speed hasn't been a bottle neck for years, the bottlenecks are predominantly storage and network, and then the number of physical cores. You would not notice mild memory differences, only big jumps in total memory volume.
Which is to say that the biggest perfomances differences come from:
Exact configuration of RAM and RAM timings and speed isn't as noticeable as anything above, perhaps a +/- 1% difference that you couldn't perceive unless you're running a system at full capacity for a sustained period of time which no-one is (even tasks people think of like that don't use all the things all the time).