Hacker News Reader

Comment by ogre_codes | Hacker News Reader

> It is not out of the question that Apple, within a year or two, has by far the best performing laptops and desktop computers on the market, just as they do in mobile.

This quote from the article is the big takeaway. Apple isn't going to make a lateral move here. A processor transition that nets Apple merely comparable performance would destroy the Mac brand. Apple needs a CPU that runs native code significantly faster than Intel on every device and it needs to be able to emulate Intel chips at reasonable speeds.

I suspect Apple knows this and that their chips are going to hit the mark. But we'll see.

View on HN

My money is not only on the performance, but also an adjustable thermal envelope, which means perhaps even 5 times the battery life of normal windows laptops.

With some Ryzen Renoir machines hitting 9-10hs of battery (when paired with a decent 80+ watt-hour battery, that is), would you expect a mac hitting 50hs of battery life?

I mean I could see that happening when pairing a current A13 with a much bigger battery like a notebook has, but as others have said, the performance will have to be superb, so maybe we'll be seeing half the power efficiency? Were you thinkg 50+ hours or more in the lines of 20+ hours?

Considering a laptop can be at most 100Wh for US air travel - I doubt you'll see 50 hours anytime soon. That would mean the display is drawing less than 2 watts of energy while the rest of the system is near 0. Typical displays apparently use 2 watts - https://www.digitaltrends.com/computing/intel-low-power-disp... If Intel released a 1-watt display then we might see the 50-hour level reached but that's going to only be on really light workloads. (Light text editing? Playing back a special low-power-codec video?)

Right on point with your considerations, both on maximum battery capacity and the display being a large factor un notebook battery life. Battery capacity can't be expanded, since it's regulation and we're tied to it. Don't really know what drives display power consumption and how it could be improved to get better battery life.

I was thinking about 20 hours of actual battery life. Seeing as the A13 has the performance of 9900K in many cases (as mentioned in the article), I'm assuming it's pretty easy to achieve around 20 hours of battery life with an A13-based chip.

That being said, you could argue that the sporadic and discrete nature of phone usage has nothing to do with continuous "strain" on the CPU in a Desktop usage which could change the consumption.

There's one thing that I've noticed and it's the memmory subsystem. SPECINT and SPECFP dabble on the "raw power" of the chip, and it'd seem in that area the A13 would be king, but as most "micro"-benchmarks go, we no from real world performance that they don't quite tell the whole story.

So I went looking for performance numbers related to memmory, and found some anandtech tests[0], the summary being:

* L1, L2 and L3 are comparable to AMDs 4900 processors (which I'll take as baseline for "best x86 mobile", with a grain of salt).

* LPDDR4 1600 MHz vs AMD's LPDDR4 4266Mhz (dual channel, can't find if A13 is DC too, but I'd expect it to be).

* RAM latencies in excess of 340ns, vs "65+"ns for AMDs. Let's take that as "sub 100ns" since they're not giving straight numbers[1].

* RAM bandwidth to memory in the order of 20-40GB/s depending access patterns. That places it around the x86 "counterpart" which is capable of "around" 25-50GB/s

* Bandwidth in the caches is different too, but I didn't want to risk a bad interpretation on my side. It'd seem the Ryzen hits larger numbers (so better), but I admit I'm not used to reading these charts and feel a bit "out of the water" there.

Seeing all that... I'm not sure the memmory system is on par to what x86 has to offer with its "best contender" right now. And let's also remember that Intel is still beating AMD in memory latency (which helps them drive better numbers in some "real world" benchmarks as we see in certain games, for example, where Intel still leads despite the "raw power" Zen2 has shown).

My take is that a "slow memory system" could be a big drawback for a CPU in real world tasks, resulting in a chip that, on paper should be great, but once you sit at a computer to use it "feels slow".

Of course this is complex to test and try, and far more than I can analyze sitting at home driving searches over the internet.

[0] https://www.anandtech.com/show/14892/the-apple-iphone-11-pro...

[1] https://www.anandtech.com/show/15708/amds-mobile-revival-red...

Depends on what you mean by performance. The same speed with double or triple battery life would be worth it.

For native apps, yes. But these will need to run older Intel software as well. It needs to be able to run the x86 version of Microsoft Office or Photoshop at reasonable speeds. The only way that is going to happen is with a blazing fast CPU with a good emulation layer. That is more-or-less how they went from PowerPC to Intel, it's likely how they will get from Intel to ARM as well.

Reply on news.ycombinator.com