After the Internet lit up last week over stories that the iPhone 6S has different levels of performance due to the different CPUs, a whole slew of new tests and benchmarks are being run to find a more definitive conclusion and calm some of the frayed nerves over this issue.
Apple, of course, has said the benchmarks are no big deal and there is very little difference between the TSMC-manufactured A9 processor and the one made by Samsung. But the fact that Apple is even commenting should say something, because Apple usually doesn't say anything unless it's backed into a corner.
Now, consumer electronics companies use parts sourced from multiple manufacturers all the time, usually to meet demand. In Apple's case, it wants to move away from Samsung and entirely to TSMC due to their legal fight.
Different manufacturers will have different methodologies, so their products will not be identical. But there should not be a notable variance like Apple users first found in the iPhone 6S. But that's what happened. Individuals determined what make of CPU their iPhones 6S had thanks to a utility and began testing the TSMC vs Samsung processors and found the Samsung chips had significantly lower battery life than the phones with TSMC chips in certain tests.
Apple sent a statement to several sites (including ExtremeTech) claiming that the tests being used were not reflective of "real-world usage" and that the difference between iPhone models was no more than two to three percent.
ExtremeTech, Tom's Hardware, and AnandTech have all done their own sets of tests and the gap between Samsung and TSMC phones did indeed close. The biggest variance was with GeekBench, which was used in the initial tests that started everything. GeekBench uses a constant workload, which explains Apple's comment about not reflecting "real-world usage."
I'll let the Mac sites and propellerheads do their own benchmarks and tests. I returned the 6S because of poor performance that I noted days before this issue blew up and the fact it offered me no real advantage over the 6. The extra application memory was nice but beyond that I got no real benefit. So I went for the cheaper iPhone 6 with more storage (16GB really is a bad idea).
Let's talk how this kind of variance can happen. With TSMC and Samsung using different manufacturing processes, the chips are bound to perform differently even though they are using the same design.
"Assuming they are using the same geometries doesn’t mean they will have the same performance. Each foundry has its own secret herbs and spices," said Will Strauss, principal analyst with Forward Concepts. He follows the mobile and embedded chip market.
The foundries have to do something called implanation, where they put impurities in certain places on the wafer to get the performance they want by altering the conductivity of electricity. This is done literally at the atomic level. Layers can be only a few hundred atoms thick.
Every company uses a different method of diffusions and other processes in the fab, each one has its own recipe. Ions are blasted at the wafer like an aerosol spray. "The impurities define how good the transistors will be. It simply makes a conductor a semiconductor or an insulator into a conductor," said Strauss.
Plus, both chips are made using the FinFET process, which is a 3D transistor design, and there is great variance between different FinFET manufacturing processes. Intel also uses FinFET, and has since it went to 20nm. All of the vendors supporting FinFET use different chemicals in the manufacturing process and different lithography sizes, which will affect the geometry size of the transistors.
The bottom line is there is of room for variance in manufacturing but it should not be too great, because it's still the same CPU core, cache system and other SoC elements. Will it hurt Apple? Are you kidding? It shipped a defective phone five years ago (Antennagate, anyone?) and didn't suffer a thing for it.