Ask HN: Does RAM speed matter for programming?
It is a well accepted fact that lots of RAM does improve performance. But does RAM speed matter in computation intensive operations? Is there a difference between 1600MHz RAM and a 2400MHz RAM in performance, everything else being equal? If so, how much?
I have googled around and came across a few links and this excellent review http://www.anandtech.com/show/6372/memory-performance-16gb-ddr31333-to-ddr32400-on-ivy-bridge-igp-with-gskill . But all of these seem to be in the context of gaming? Does it apply to programming too?
(
9 comments
| 2 points
| on HN
)
Additionally, Mechanical CAD (Solid Modeling) isn't multi-threaded in the vast majority of cases right now. There are some niches where it can help, but really we are still identifying how this might work.
For CAD, you want a very fast CPU and fast being single thread performance maximized, and then fast and a lot of RAM.
The order depends. Small, but complicated data sets favor speed over size.
Larger ones demand both.
On the software I work with, that big of a jump in RAM speed, coupled with a good CPU and cache can reach 30 percent faster update / build times. When those are seconds, nobody cares. When they are an hour, yeah. Big deal.
CAD is also compute intensive, meaning it's a great system stress case.
As with most non-gaming activities, your bottleneck would most likely be your storage device. Get a great SSD.
Depending on what you're actually programming, your CPU, GPU, and RAM speed might matter a little. What are you programming?
For neural nets, it depends heavily and how the code is optimized and what your inputs are. I know that isn't very helpful.
In general, the people I know who have trouble with high-end, consumer-grade hardware (~$1,000 in parts for a single machine) are usually researchers. Then you're getting to the point where you'd want to look into distributed processing.
I'm a web developer, so when I'm programming, I often have a database GUI, full IDE, multiple browsers, multiple Docker containers, and RDBMS running.
That ends up at about 4GB of RAM, depending on what's in my database.