Qualcomm Snapdragon 8150, Apple A12X Bionic benchmarks leaked

1. Now, all these are awesome processors , but what creates confusion is the attempt of those benchmarks like geekbench and antutu to measure “realistic” performance.
   Realistic performance is not that difficult to assess , even if , no “real” testing is done. What we don’t know , or at least directly being informed is theoretical performance, which is fishy to say the least.Not that we can’t search, but it is not the same….

   I may be interested to know if my phone can spike in a demanding load or throttle in lighter loads. But , what is more interesting ,computationally speaking ,is sustained performance under a long heavy load, that gives no space to excessively throttle or spike , which is what makes the real performance in my opinion.

   So when being informed about performance, we do need to know, operational frequency, instructions per cycle(dp and sp), memory bandwidth, branch prediction and so on. A computer’s performance can’t be bigger than the predicited theoretical maximum.

   From the little info , “mobile” companies let out in the light of the day (and now unfortunately Intel and AMD start following bad influences), it seems that theoretical maxima are far smaller for mobile than desktop and laptop processors and if this is so , all this circus about phone processors being equal to desktop is fake. But they do need to come out clean or I will never trust anything they say. Because it seems “mobile” and “desktop” processors are tested in loads that suit best “mobile” processors and leave demanding tasks untouched.

   Additionally indirectly one can see that what these people claim about their bionics and snapdragons can’t be true, otherwise top supercomputers would be built from them as building blocks. Saving cost and energy is a big deal when you have to run millions of cores together and spend MWs of power. But I haven’t actually seen anyone rushing for these so efficient and so fast processors. Ergo… they are not so efficient and so fast after all! 🙂

2. “Porting macOS to run on ARM without breaking compatibility with legacy apps is also likely to take some time, so it could be a few years before that happens.”

   Or they could have done it years ago.

3. Why is there the shitty SD 845 and Kirin 980 and not the beastly Exynos 9810 ???

   1. The QSD 845 has obvious latency issues, and despite that, its actually faster than the Exy 9810.
      The Exy 9810 uses more power and throttles immediately, its first/high benchmark scores are not representative of the actual chipset. Run the benchmark again (actually 10 times is best) and you see massive heat, crawling performance, and even random shutdowns.

      Over at anadtech and xda, people have started tweaking the drivers, clocks, and schedulers to “tame the beast” and it seems to produce great results. Much better actual performance, but the theoretical benchmark scores take a nose-dive (but still higher than non-tampered throttled scores).

      Overall, I’d still put it lower than the QSD 845. The predecessor might fix some of the low-hanging issues and see it live to a fuller potential, smashing the Exy 9810, QSD 845, and the Kirin 980 out of the water. But then there’s Qualcomm’s next Snapdragon which may win again…. or it might lose by a huge margin. Time will tell (Initial Mar ’19, Detailed June ’19).

4. Make a Macbook Pocket with 7 inch screen once this chip or its successors start being used in Macs. Slap some gaming controls on, too, to make it a gaming UMPC and then I’ll finally voluntarily buy an Apple product.

5. I would prefer to see Google Docs benchmarks rather than Geekbench. I don’t care what potentially my tablet can do, but what it actually does. Real word benchmarks on Apps that run both on iOS and Android. Webkit benchmarks are good too, as they run on both OSs.

   1. I believe that this year’s ipad pro is the first ipad to include (in the box) a wall adapter that can do 9V charging. Also, a usb c type cable too. Not even the latest iphones have included 9V wall adapters.
      Well, I like how Korean phones (Samsung, LG) have always had 9V wall adapters included with the phone. This has nothing to do with application processor performance, just the user experience of the base unit.

      1. Who cares about the voltage?

         It’s the wattage that matters.

   2. So you use Google Docs as a benchmark and I’ll use Apple Pages and benchmarks will cease to be benchmarks

6. Does this take into account the forced throttling of iPhones?

7. Bring in actual sustained performance with something like 3DMark Physics or TabletMark and the results will be significantly different. These Geekbench results are a result of the aggressive boost clocks and associated aggressive peak power usage of the Apple A series.

   Interestingly, unlike many other tablet and smartphone processors, the A series’ aggressive peak power demands is precisely why Apple iPhones with aged batteries have abrupt power shutdown events. In fact, these lithium batteries, which are some of the best in the industry and would otherwise supply more than sufficient power for practically any other mobile processor including even the 15W Intel Core line at default power curves, are still not powerful enough for these processors’ peak power.

   1. Most power draw measurement tools cannot see the simultaneous power draw on the A series with fine enough resolution to show what I am talking about. What I have seen is very sharp, nigh-exponential ramping far beyond the sub-10W average TDP ratings people often associate with the A series.

      1. Data or it didn’t happen… Transients are hard to measure, and the technique to use the tools (A little more expensive than an iPhone X) comes dearer than the tools.

   2. Now, to be fair to Apple, they have at present the fastest reacting dynamic clock frequency generation in the industry. Nothing comes close, even Intel with the significant improvements they made starting with Broadwell Core M. This allows them to quickly ramp up and down to complete a task and expend a lot less unnecessary energy during subsequent idle clock cycles. That is why the total energy (this is power multiplied over time) spent is less, whereas the power transients can get a bit crazy in order to pull off what they are doing.

Comments are closed.