Apple A10 Geekbench 4 Score

[JoeRambo]

Just shows how unreliable cross OS comparisons are. For all we know Apple could optimize their CPUs for benchmarks and no one would ever know.

I haven't seen other benchmarks yet, but I will take a look when AT publishes their review. For sure I want to know what they have done with the architecture.

Reliable or not, some scores are hard to cheat. My work 4.6Ghz 2500K has ST score for some subtests that is barely above A10. Apple surely has some secret sauce in their CPU if they can "cheat" LLVM, SGEMM or SFFT tests.

 

[asendra]

I hope AT can run SPEC06 on the 7+ with 3GB of ram. Not sure if the 4GB was a minimum requirement, or it needed more than 2GB only.

I feel like Intel is playing with fire with things like this though
https://twitter.com/AndrewWrites/status/778340105496981504

> "To summarize: this slide shows MacBook Pro-class processors basically jogging in place until *late 2018*"

 

[IntelUser2000]

I hope AT can run SPEC06 on the 7+ with 3GB of ram. Not sure if the 4GB was a minimum requirement, or it needed more than 2GB only.

I feel like Intel is playing with fire with things like this though
https://twitter.com/AndrewWrites/status/778340105496981504

"To summarize: this slide shows MacBook Pro-class processors basically jogging in place until *late 2018*"

I do not believe they are doing this deliberately. They are doing this because they have no choice. If they could have CPUs that perform 2x faster per clock with 50% higher clocks in average across entire SKUs, they would. The ENTIRE CPU world is converging towards what's mostly the same philosophy in design. The high end chips will soon have very similar IPC. If you strip out fancy techniques that benefit server, and multi-threading, IPC of the 200W+ IBM Power chips are somewhat behind top of the line Intel chips.

So clock speed of the chip is determined nearly entirely by TDP headroom(irregardless of design team, company, process), and partially by number of pipeline stages. The IPC differences exist, but by a small amount separated by specialization(whether its PC focus, server, or mobile). If that doesn't mean Physics is telling you to "STOP!!", and end of measurable gains, I don't know what it is.

The interesting thing is even LN2 setups with CPUs consuming 3x the power(300W+) or more hasn't breached the 10GHz barrier. Engineers at Intel 1.5 decades ago wanted 20GHz processors in the near future. That tells me regarding the future, they don't know what the heck is coming. And overclock numbers simply tells us what Intel concluded after failure of Netburst. If we could have 10kW processors we may be able to have our 20GHz CPUs.

 

[AtenRa]

Has this been posted before ??

Apple iPhone 7 Teardown
http://www.chipworks.com/about-chipworks/overview/blog/apple-iphone-7-teardown

 

[witeken]

The small core isn't actually all that small.

The modem die size is ~66.4 sq. mm.

How big is Qualcomm's modem?

 

[KTE]

Not particularly interesting for DT/Server.

Excellent news for Mobile!

Plenty of older CPUs have high IPC but did not scale frequency, power and cost wise for mass production.

FO4s limit the clockspeed band.

A10 might be good for mobile workloads at mobile clocks. Even for Entry DT. But anything more, and they just won't get the clocks, even with the best of process.


Sent from HTC 10
(Opinions are own)