iPhone 12 Pro and iPhone 12 Pro Max about 20% faster than iPhone 11 Pro

AppleInsider

Preliminary benchmark scores for the iPhone 12 Pro and the iPhone 12 Pro Max have shown up on Geekbench with about a 20% speed improvement over the previous generation.

Benchmark scores show 20% improvement from A13 to A14

The iPhone 12 uses the A14 processor which boasts a 6-core CPU and 4-core GPU in a 5 nanometer process. Apple gave an inscrutable benchmark during its "High, Speed" event of "50% faster than other smartphone processors."

The early Geekbench benchmarks show that Apple improved the overall processing by at least 20% year-over-year, which was in line with the iPad Air 4 early benchmarks.

Scores vary depending on when the benchmark test was run, battery charge, and background operations. Scores will trend lower during initial setup when the iPhone is cataloging a photo library vs a day later on a full charge.

One score for the iPhone 12 Pro showed a 1613 single-core score and a 4076 multi-core score. This compares well to the iPhone 11 Pro at 1343 single-core and 3478 multi-core.

Pre-orders for the iPhone 12 and iPhone 12 Pro start on Friday October 16 at 5 a.m. Pacific time and will begin shipping on October 23. The iPhone 12 mini and iPhone 12 Pro Max will be available to pre-order on November 6, and ship on November 13.

ericthehalfbee

Of course it’s faster. Apple never lies about benchmarks, and typically uses conservative estimates so people can’t claim they were lying.

So sad for the haters who were drooling at the early score of 3100 for multi. They were so happy at the prospect the A14 was somehow slower than the A13 only to have those hopes shattered. 

blastdoor

It looks like about half of that improvement is due to a higher clock speed and about half due to higher IPC. 

That's a nice, but not huge, improvement. Seems like where they really spent their transistor budget was in the Neural Engine. 

tht

blastdoor said:

It looks like about half of that improvement is due to a higher clock speed and about half due to higher IPC. 

That's a nice, but not huge, improvement. Seems like where they really spent their transistor budget was in the Neural Engine. 

Have to wait for the X-rays. It looks like Metal improved by about 20% to 70%. Note sure if some of the Metal subtasks are being executed by the NPU or the MLU, but it looks like a lot of transistors were devoted to low precision SIMD performance everywhere: CPU, GPU and NPU. Then, the SLC or L3 size will be interesting to see. The A13 had 16 MB. The A14 could be 24 MB? No change?

The Mac silicon is going to be really interesting to see as a lot of things have to be added. They may have SVE/SVE2 units. Still think there is a chance of that. The memory architecture or memory performance will have to be improved by 2x to 4x, possibly 8x, depending on which Mac. The GPU resources would have to be increased 2x, 4x, 8x as well.

KITA

blastdoor said:

It looks like about half of that improvement is due to a higher clock speed and about half due to higher IPC. 

That's a nice, but not huge, improvement. Seems like where they really spent their transistor budget was in the Neural Engine. 

Something like that. The move from N7P (A13) to N5 (A14) brings an approximate 7.5% performance increase with the same power consumption (according to TSMC).

We don't know the exact consumption and sustained performance just yet, but I'd expect it to be around the same as the A13 given little to no change in battery life of the new iPhones.

KITA

tht said:

blastdoor said:

It looks like about half of that improvement is due to a higher clock speed and about half due to higher IPC. 

That's a nice, but not huge, improvement. Seems like where they really spent their transistor budget was in the Neural Engine. 

Have to wait for the X-rays. It looks like Metal improved by about 20% to 70%. Note sure if some of the Metal subtasks are being executed by the NPU or the MLU, but it looks like a lot of transistors were devoted to low precision SIMD performance everywhere: CPU, GPU and NPU. Then, the SLC or L3 size will be interesting to see. The A13 had 16 MB. The A14 could be 24 MB? No change?

The Mac silicon is going to be really interesting to see as a lot of things have to be added. They may have SVE/SVE2 units. Still think there is a chance of that. The memory architecture or memory performance will have to be improved by 2x to 4x, possibly 8x, depending on which Mac. The GPU resources would have to be increased 2x, 4x, 8x as well.

Arm has still been pretty quiet on v9, so I'm not sure if we'll see things like SVE2 off the line.

MplsP

Nice for a phone but this will be far more beneficial in an iPad (or a MacBook?) Very few people actually need this kind of speed in a phone.

CheeseFreeze

I bet it’s even faster in real world usage. The neural processing isn’t captured in these tests from what I can see, and the improvements there according to Apple look impressive. And it’s a smart optimization strategy considering their ‘privacy friendly’ services relying a lot on on-device calculations.