Hacker News Reader

Comment by cwyers | Hacker News Reader

Perhaps relevant. Linus Torvalds' opinion on GeekBench:

http://www.realworldtech.com/forum/?threadid=136526&curposti...

> So basically a quarter to a third of the "integer" workloads are just utter BS. They are not comparable across architectures due to the crypto units, and even within one architecture the numbers just don't mean much of anything.

>And quite frankly, it's not even just the crypto ones. Looking at the other GB3 "benchmarks", they are mainly small kernels: not really much different from dhrystone. I suspect most of them have a code footprint that basically fits in a L1I cache.

It's a few years out of date, and maybe Geekbench has improved since then, but I wouldn't put a lot of stock into these kinds of numbers until I knew that those kinds of issues had been addressed.

View on HN

If these phone processors are that amazing why don't we see them being used outside phones and tablets? I mean take the A10 Fusion give it four cores instead of two and put it on a server mainboard with all the usual things like PCIe, Gigabit Ethernet and SATA. Basically do what AMD did with their A1100 except with a good processors and actually sell more than a dev kit. The lack of total raw processing power per server could be easily offset by the density and lower price of the ARM server. Why isn't this happening? I don't think lack of software is an explanation for that. Things like AWS Lambda could probably be immediately ported to run on ARM servers.

After Apple left the server space when they shuttered the Xserve, I don't think they are really interested into getting involved into that space. I think they view themselves as a consumer goods company that makes good products and uses technological innovation to serve those goals, rather then create that innovation and figure out how to productise it later.

They don't need to sell them to enterprise. They can be their own customer. They have some of the bigger datacenters after all.

ARM for data storage pools seems like a good fit.

How can we know that they're not already doing that?

(Asking in a purely epistemological sense. I'm not familiar with Apple's datacenters at all.)

Is because the cost of an CPU, within the server, And the cost of Server within the whole DC, and then the cost of Running the DC. In the grand scheme of things CPU cost is pretty small. Not to mention we are getting another round of CPU upgrade with Intel Broadwell-E. And then some price competition from AMD's Zen.

And Intel Xeon isn't just about the CPU, it is the enterprise grade QA and testing, Network Controller, SATA controller etc. And the Xeon-D is a very Decent offering.

Then there is scale, even though Apple's Cloud is huge, it is still no where near Amazon AWS scale, not Google and Microsoft Azure either. And even if you add all these three Cloud Giant Combined, I will bet their annual buying CPU core doesn't even come close to Apple's annual iPhone sales.

So to finalize, there just isn't any financial incentive of switching.

I half-way expect the next line of Apple laptops to be powered by such a chip architecture. Imagine a MBP consisting of 4 A12 chips with bucketloads of RAM, etc. Not only could it run macOS but you could run iOS apps as well. Not that I think that's a good thing (looking at you Microsoft Metro) but it should be possible.

Updated thoughts regarding Geekbench 4, the newer version. http://www.realworldtech.com/forum/?threadid=159853&curposti...

TLDR: Much better.

GB3 -> GB4 removed some sha1(?) tests for which Snapdragon introduced fixed-function hardware.

Reply on news.ycombinator.com