In June, ABI Research made waves with a report that claimed Intel’s “Clover Trail” Atom chip consumed just 60 percent of the power of a competing ARM chip, with comparable performance. Now, another analyst has claimed that’s simply not true, especially after the author of the benchmark revised its code.
After a re-test, Jim McGregor of Tirias Research claimed that the 2 GHz Clover Trail chip used by the Lenovo K900 smartphone now scores 20 percent less in the AnTuTu CPU benchmark, and 50 percent less in the RAM benchmark. That benchmark, now revised to version 3.2.2, was reportedly revised after tests showed that Intel’s chip gained an unfair advantage by “skipping” some of the benchmark code.
What does this all mean? Originally, ABI claimed that the Clover Trail chip in the Lenovo K800 smartphone consumed just 47 percent of the power of the Krait 300 chip found within the US version of the Samsung Galaxy S4. If the performance between the two chips had been comparable, this would have given a significant advantage to the more power-efficient Intel architecture.
Rejiggered tests show more predictable power-performance tradeoffs
Now, the distribution between the two chips appears to be more traditional: Intel’s chips still appear to consume less power, but the new benchmarks make them appear slower, too. McGregor has released a slew of benchmarks to show that the S4’s Krait 300 chip (as well as the Exynos Octa found in some European versions of the S4) outperforms the Clover Trail by a significant margin – up to 2.5 times in the case of some benchmarks, such as the Linpack Multi-Thread benchmark.
There’s one other wrinkle. ABI’s own power estimates were based upon current draw, with the assumption that the operating voltages would be roughly comparable. ABI didn’t measure the overall power (probably because the devices tested were housed inside of a phone, with variables including the screen resolution to consider). So there’s that.