Just How Fast Is the Apple Inc. A9 CPU?

AnandTech recently published its review of Apple's (AAPL 0.04%) iPhone 6s. As part of its review, the site conducted extensive performance tests of the Apple-designed A9 CPU inside of the device. What is particularly interesting, though, is that AnandTech ran a test known as SPEC CPU2000, a retired but respected test of CPU performance.

Apple routinely calls its A-series processors "desktop class." This claim certainly holds up in some mobile benchmarks, such as Geekbench, but does it hold true in what is arguably a more robust general purpose CPU performance test like SPEC2000?

Let's take a closer look.

Apple A9 versus the Intel Core 2 Duo E6700
According to the popular mobile CPU performance test, Geekbench 3, here is how Intel's (INTC 3.03%) Core 2 Duo E6700 -- which was a top-of-the-line desktop processor in 2006 -- with the A9 processor found inside of the Apple iPhone 6s Plus in integer performance:

Data source: Author, Primate Labs website.

Per this test, the A9 offers around 48% better CPU performance than the Core 2 Duo E6700, an impressive result given that the A9 can power a smartphone that lasts a whole day while the E6700 is a desktop processor.

Let's see how the A9 compares to the Core 2 Duo E6700 in the SPECint2000 test suite:

Source: AnandTech, SPEC.org

In SPECint2000, the Core 2 Duo E6700 wins more than it loses, although the A9 puts up a very good fight. I think it would not be a stretch to say that iPhone buyers today are getting high performance desktop CPU performance from 2006 in their pockets today.

That's just really cool.

Some caveats to these performance numbers
It's important to keep in mind a number of caveats with these numbers. Firstly, these tests are only a measure of the integer capabilities of the device; modern CPUs dedicate a significant amount of hardware to perform floating point computations (useful in applications such as games). AnandTech did not run the corresponding SPEC2000 floating point test, so we can't compare the two devices in floating point at this time.

Additionally, I suspect that Intel's code compiler (used to generate the benchmark on the Intel platform) generates far more optimized code for its processors than whatever compiler was used to generate the code for the iOS platform. From what I have heard, the only sub-test that isn't so easily "broken" with smart compiler tricks is the 176.gcc subtest.

In the 176.gcc subtest, it looks as though the Apple A9 CPU and the Core 2 Duo give around the same score, which leads me to believe that these two CPUs are roughly comparable.

It would be interesting to see how A9X compares with Intel's Core m7
One comparison that I would really like to see is a comparison of the A9X against Intel's top Core m7 processor, the Core m7-6y75. This should help give investors insight into just "how close" Apple is to Intel in terms of CPU performance in tablet-like workloads.

It would be good to see results in SPECint2000, although it would be even more interesting to see how the A9X and the Core m7-6Y75 compare in the newer SPEC CPU2006 test-suite in terms of both integer and floating point performance.

AnandTech indicated that it couldn't run SPEC CPU2006 on the iPhone 6s/6s Plus because one of the tests required more memory than what is available on the devices, but this shouldn't be a problem on the iPad Pro which is expected to feature four gigabytes of memory.

At any rate, I look forward to seeing Intel and Apple push each other to try to make the best mobile processors on the planet.