The first reviews of Apple’s iPhone 5S are in, and the news is unilaterally positive. Unlike the iPhone 5 launch, which was badly overshadowed by the Apple Maps debacle, the iPhone 5S has no such problems. General coverage of everything from the new camera to iOS 7 is positive. The issues that people were concerned might be gimmicky, like the new fingerprint sensor, are receiving high praise. The camera, with its dual flash LED technology, is noted as taking better low-light photos than its predecessor.

As for the relative strength of the upgrade, David Pogue over at the New York Times writes in his review that the 5S is one of the few smartphones that’ll keep you delighted over two years of your contract.AllThingsDigital echoes that it’s a big step up for anyone who doesn’t already own an iPhone 5, but says that it may not be a huge case for upgrading if you’re already on that device. Full reviews and coverage of iOS 7 is less a focus than the overall phone, but there don’t seem to be any obvious design flaws or issues the way there were a year ago. Daring Fireball confirms the positive sentiments.

Relative performance and 64-bit

If you’re curious about performance, Anandtech has the best coverage of any site that’s released a review. There’s a great deal of information there, including performance data for the CPU, GPU, 64-bit evaluations, and battery life. To hit the highlights first, before the mea culpa — battery life is on par with the iPhone 5 when browsing over 4G LTE, but over an hour worse when running on WiFi (10.2 hours vs. 8.97 hours). In their most power-hungry state, the iPhone 5S takes a 12% battery life hit, even though its battery is actually ~12% larger than its predecessor. Cellular, video, and wireless capabilities are all largely unchanged.

So let’s talk about the 64-bit/32-bit transition. Many of you were adamant last week when I maintained that this shift was mere marketing fluff. A great number of people swore that thanks to various advances, the iPhone 5S would offer 2x the performance in 64-bit mode vs. 32-bit mode.

Who was right, and who was wrong? All of us it turns out.

Many of the performance benefits that people claimed were a product of the 32-bit-to-64-bit conversion are actually the result of a superior CPU architecture. The new chip, codenamed Cyclone, has 20% lower latency to main memory compared to the old Swift core. The L1 data and instruction caches are twice the size — 64K, up from 32K. Memory bandwidth in 32-bit mode is increased by 6-58% depending on the operation. The new chip is 42% faster in integer code and 67% faster in FPU code.

It’s surprising to see Apple leveraging a new architecture this aggressive, but not impossibly so. The A7′s Cyclone processor is clearly a well-balanced piece of work. But the point I wanted to make that got lost within my 64-bit story is that huge performance jumps across a CPU generation don’t necessarily have anything to do with whether or not the chip is 32-bit or 64-bit. So how much does that 64-bit jump matter?

Excluding the order-of-magnitude leaps delivered by hardware AES blocks, the new chip is 245% faster in SHA-1 cryptographic testing. In more mainstream tests around file compression, JPG handling, PNG manipulation, and other various mathematical tests, the gain ranges from 25% slower to 28% faster. The floating point gains are stronger — none of the Geekbench tests go backwards, and the performance ranges from 0% (no change) to 195% for DGEM (double precision matrix multiply). DGEMM, it should be noted, is the sort of function that we’d expect to run much faster on the 64-bit chip thanks to the additional registers and NEON support.