First, the chip in the MBPro is fabricated and packaged for a much higher TDP. The memory and storage subsystems in an MBPro are also designed for a higher power envelope and battery size. How much better an A12x-workalike or later would perform when designed for that higher TDB is unknown (to us, not necessarily to Apple).
Second, you cherry picked one set of benchmark tasks. If you pick another set of tasks, for instance LLVM, the lower TDP A12X already beats the higher TDP i7. Improved real world compilation speed would make a lot of developers happy. Especially if they have to spend less time per annum hunting for wall plugs.
And the one task where the i7 performed significantly better, SGEMM, runs the type of algorithms that Apple has been offloading to the Neural Engine(s) and Metal GPUs in its newer SOCs, rather than the ARM cores.
you didn't read anything really posted did you? just immediately decided you had to defend the ARM cause of Apple?
first of all. if you want to try and be pendantic without reading throughouly, most of what you say can be pretty easily refuted.
First on the TDP. Any claims that Apple could just up the TDP and suddnely they're faster is not true. TDP to power doesn't scale linearly on silicon and we have existing examples of ARM servers CPU's that show that once you bring the CPU TDP Upwards, you still run into the same ceilings that x86 does. Does Apple potentially have a 30 or 95 or even 115w version of ARM scaled up? possibly, but we have never seen one either in benchmarks, leaks or anything else. Right now we can only look at their fastest max offering. And that's the A12x in the iPad Pro.
Hell of a CPU. But as my second set of points pointed out, which you ignored for some offshoot about SGEMM, for day to day user tasks (the ones i nitpicked, and left out the ones that won't affect your average user, such as you know, the average facebook user), Intel still performed faster in almost every single one.
ARM does better in certain tasks. But those tasks tend to be very specific. Usually places that are low power requirements, or highly parallel workloads (such as webserving encrpytion).
the jproblem you, and the person I originally responded to keep ignoring is that the "total" score in geekbench is a WEIGHTED AVERAGE. WHich means geekbench itself outright puts more points towards specific tasks and does NOT weigh them all equally.
This sets up a secnario where the benchmark number may very well NOT be truly indicative of real world performance based on the specific task at hand.
for example, the A12x might have a better score than the i7 overall. But if my primary use case of the computer focuses on tasks that the i7 does bette,r than the i7 is going to be the faster chip. this does also apply vice versa.
So this constant "BUT GEEKBENCH SAYS!!!!" rhetoric is just that. Rhetoric and isn't applicable in the real world.
I would love to see what Apple's designer chops can do with higher thermal envelop. It COULD be truly inspiring. I feel like out of all the problems Apple has currently, their silicon design team is probably the best and smoothest running division in the company, constantly performing and resulting in better and better results.
But given all that, no matter what we have seen, They still have not released a single A series CPU that manages more than 15w. And that's just not going to cut it for users who have focused tasks needing a lot more power. Remember, these benchmark results that we're talking about are also only intel's 28w offering. Apple would also need chips to compete against intel's 95w offerings and 115w offerings before they can even consider a full switch over. And those offerings need to be at least closely comparable in a lot of tasks.
if you're a DB Admin for example, and suddenly Apple's newest laptops are 20% slower for your tasks, even though the geekbench says the A series should be faster. who cares what geekbench says? my tasks are now 20% slower and Apple is no longer competitive. This is what they have to avoid before switching.
