So potentially the Apple customers are participating in a chip lottery? Some will get Samsung chips, some TSMC chips. Both may be clocked at the same frequency, but may e.g throttle differently.
Of course there may be variations between individual chips from the same manufacturer too, but still interesting.
Kind of reminds me of the panel lottery when buying LCD TVs...
Does this include bitcode delivery via Xcode? It's default, but still optional for now - this will be a requirement in the near future. How will you prevent behind the scene optimizations?
Yep. Confirmed.I am still guessing they will bench the same, with throttling not being a significant issue, except under unusual conditions.
The problem is that you won't know if Apple doesn't do specific benchmark optimizations.You don't want to prevent behind the scene optimizations since every iOS developer has access to them.
I wonder the same. But is there any smartphone with a 14nm Intel chip?Now that there are multiple FinFETs in the market and have presumably been investigated, I wonder how they compare.
Is Samsung or TSMC's FinFET more like Intel's 22nm or 14nm in terms of power consumption etc? Please give sources.
The problem is that you won't know if Apple doesn't do specific benchmark optimizations.
I wonder the same. But is there any smartphone with a 14nm Intel chip?
Anyway I'm not aware of any public study about SoC power consumption (as opposed to full device power consumption).
Just rewrite the program and test again.
I wonder the same. But is there any smartphone with a 14nm Intel chip?
Anyway I'm not aware of any public study about SoC power consumption (as opposed to full device power consumption).
Why? He's still incorrect in his conclusions.
Could you provide proof of that. Who were the manufacturers of the different A5 chips ?
How would that help? 'Just rewrite' - Are you aware of the resources needed for contant rewriting on multiple platforms?
you have no longer control over the specifics of your benchmark. There can be a large performance difference between compiler versions
If Apple finds it beneficial to move heavy fp code to GPGPU processing, then what are we actually benchmarking?
Hiraku Wang has created an app that's able to determine whether an iPhone has a TSMC chip or a Samsung chip, and has shared some data on results gathered from users who have installed his app.
According to results from approximately 2,500 iPhones, there are more TSMC chips than Samsung chips. TSMC chips were found to be installed on 58.96 percent of devices, compared to 41.04 percent for Samsung chips.
The iPhone 6s Plus appears to have relatively equal split of Samsung and TSMC chips, with slightly more Samsung chips. Of 1,329 iPhone 6 Plus devices, 56.81 percent have the Samsung chip while 43.19 percent have the TSMC chip.
Looks like IDC was right.
Sorry, but didn't IDC claim that the A9 was 100% TSMC?
So what about creating another thread? You are off-topic here, thanks.I'm not talking about the SoC, I'm asking about the transistor, the fin.
Sorry, but didn't IDC claim that the A9 was 100% TSMC?
At any rate, for those who are curious, at the moment its all TSMC. Samsung was originally apportioned 80%, then 70% due to softening confidence in their yield ramp rate, and that was later on further reduced to 50%...but their yields are so low right now (compared to TSMC's) that practically all the A9's headed to market on the eve of the iphone 6S release were fabbed by TSMC. Should Samsung get their yields up, and same goes for GloFo for that matter, then they'll get to claw back their portion of the volume contracts for the A9 in the coming months and quarters.
The A9x gonna be superinsane, something like twice the performance of xbox 360. Is that correct? Xbox 360 has similiar performance to Kepler soc from nvidia and A9x is more than 2x. I admit I thought samsung would do all a9 and tsmc all a9x seems it was wrong ):