Sign in with
Sign up | Sign in

Power, Performance Scaling, And Efficiency

System Builder Marathon, December 2010: Value, Compared

Power consumption is the one place where a low-cost single-GPU machine usually shines, and today’s $500 build is no exception. Three things that hurt the $2000 machine’s power consumption in comparison to the $1000 PC are its CPU manufactured on a larger lithography process, extra CPU cores, and an old-school GF100-based GPU.

Since most of our benchmarks get zero direct benefit from SSD drives, money spent there was formerly unrealized in our value comparison. Our previous resistance to SSD requests ended with the compromise that drive performance would count towards ¼ of the final results in today’s comparison. Let’s take a look at what that does to our overall performance charts.

Ouch. So large a difference might make today the first time that our high-end machine is fully-competitive with the lowest-cost build in both efficiency and value!

Accounting for only ¼ the overall score, vastly superior hard drive performance roughly doubles the high-end machine’s average performance rating. The result is a giant lead in efficiency, since efficiency is a measure of work per unit of energy. Some of our readers disagree about this being an unfair advantage, since program launch time is a big part of what helps them to define a “high-end” build.

React To This Article