Hacker News new | ask | show | jobs
by kostya-kow 4839 days ago
One can argue that not everyone has an i7, and offloading work to the server enables it to run on weaker computers.
3 comments

One can also argue that not everyone has google fiber, and running locally enables it to run when internet access is slow or unreliable
What reasonable company would justify spending tens of thousands an hour on server costs, when they could optimise their code a little more and run it for free?
You do know what c-suite we are discussing, correct? ;)
The richest games company makes decisions which benefit them...
So let's assume that the average person wanting to play SimCity is running a Core2Duo (released 7 years ago). It's hard to find benchmarks directly comparing a Core2 E6600 to something like an Ivy Bridge Xeon that you'd expect to find in a modern dual-socket 1U server, but even looking at a TomsHardware chart of x86 core performance can tell you that an i7-2600k is only about twice as powerful as a Pentium 4 HT660 (core-for-core) http://www.tomshardware.com/charts/x86-core-performance-comp...

Bottom line, the total cost of ownership doesn't at all make business sense to do right now. In 10 years, it very well might.

See also: https://gist.github.com/jboner/2841832 ("Latency Numbers Every Programmer Should Know"). TL;DR

Main memory reference 100 ns 20x L2 cache, 200x L1 cache vs Send packet CA->Netherlands->CA 150,000,000 ns 150 ms

The core-for-core thing makes a huge difference in the real world that doesn't show up on single-core benchmarks. The HT660 was a good chip in its day, but it's at a four-to-one disadvantage for code that's multithreaded and/or running on a busy PC.

I think Sandy Bridge is my favorite CPU of all time. It does a truly massive amount of work without consuming significantly more power than the part it replaced.