Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
A9 Chip Manufacturing Split 60/40 Between TSMC and Samsung, Not Segmented by Device Size
- Thread starter MacRumors
- Start date
- Sort by reaction score
TSMC here on iPhone 6s. Battery life is pretty bad first day into phone, hoping it'll be better as I go along. Also had a weird freeze up but moved to iOS beta, no trouble since. Well, after me getting pissed at Samsung over no removable battery or SD card, I'll take the lower battery life just to be fickle.
Edit: I just got another freeze up. =(
Edit: I just got another freeze up. =(
Last edited:
i don't know about that..
if you take the lowest geekbench score posted by a tsmc and compare it to the highest mark posted by a samsung a9 (and definitely ignore the majority sampling in the middle.. also ignore that you have no idea the conditions present during those two tests)
.. you'll find a 1% difference.
that's like driving 49.5mph instead of 50mph (on a closed circut/non-real-world/very specific test)
..the average drivers could certainly tell the difference between the two speeds.. /s
You can't sell 2 different chips and call them both the same name, because that's fraud. If you got the slowe,r hotter running TSMC "A9" then return it until you get what you paid for.
When Intel does this its called a "Tick" and even they name the chip something different like changing the name from Haswell to Broadwell.
When Intel does this its called a "Tick" and even they name the chip something different like changing the name from Haswell to Broadwell.
I would guess marginally better battery life. Not a huge die size difference. I have the Samsung chip and speeds are in line with tsmc chip. No faster, no slower. I would imagine battery life would be minutes longer, not hours.
Well it's going to be about 10% more efficient so over an hour of extra battery life for the Samsung chip. Pretty significant if you ask me.
If they're doing that, do you think that could be a cause for the freezing? I'm not one to nitpick at my devices but I've never had an iPhone freeze mid swipe in between home screens before... Shoot. Just.. Ugh.
you may just have a bad phone - i would take that back to apple and show them. they'll swap it out right away and send it for failure analysis.
no, not really. they have a target performance they want - the chip should run at 1.8GHz. if you sign off at 2GHz, then you've given yourself some voltage margin (and temperature margin, as things will get slower at high temperature). the chip still runs at 1.8Ghz but it can do so at a slightly lower voltage, and therefore at a lower power. but it came at the cost of closing timing at a higher frequency, which may be difficult depending on what your critical paths look like.
by the way, when you produce a chip like this, there is a LOT of timing characterization going on. the vendor will give you timing numbers for fast silicon, slow silicon and typical silicon at high and low temperatures and high and low voltages. you then pick the worst case corners (slow silicon, high temp, low voltage), (fast silicon, low temp, high voltage) and make sure you can still meet timing. if you can't you add pipeline stages, or resize buffers inside the chip and simulate again. signoff and tapeout will not occur until you've closed timing at these corners.
then after the silicon is made, the vendor will give you samples of these corner lots so you can test them in-system. believe me when i say apple has tested the living crap out of these devices at voltages and temperatures far outside normal. this is how you make sure you can make millions of devices and they all work properly.
I would say no chance the freezing is due to tsmc versus Samsung. TSMC has been manufacturing hugely complex processors for nvidia for many years. I would do a clean load without restoring a backup and see if you still have a problem. If you do I would take it back and get another. Probably just a defective phone.
I've never loaded a backup, only ever clean installs. I'm totally not looking forward to trying to exchange it... I hate doing that. I'm gonna give it a few days and see if it worsens or improves. Who knows, it may be settling in. It's a long shot but I've got 14 days...
I believe the Samsung chip is the slower hotter one. Also, they don't have the same name. They have different designators printed on them internally.
If you're going to install this app, list which device you have Decode the serial number of the device so we can triangulate which factory has which parts.
As for Benchmarks: iPhone 6S 128GB Silver
For completeness, Geekbench in airplane mode:
As for Benchmarks: iPhone 6S 128GB Silver
- Factory: DNP - China, Chengdu (Foxconn)
- Year: 2015
- Week: 37 (14.09 - 20.09)
For completeness, Geekbench in airplane mode:
Last edited:
If people honestly believe they can correlate errors to what chip they have... #losinghopeinhumanity
Nooooooo. Holy jeez. No. Just. No. The manufacturer of your chip will not determine if your phone freezes. My lord.
Last edited by a moderator:
No but the manufacturing processes will especially if the chip is being throttled to match the thermal envelope of the more efficient Samsung counterpart.
The TSMC chip is bigger. The obvious is there. However, the not so obvious is WHY is it bigger? The technology to make the smaller chip to match Samsung is there, so why the discrepancy? You can have an Intel and AMD both running the same GHz but do you think all processors are created equal?
I said something similar in the other thread on here about this. Watch someone over the next handful of months figure out how to unlock the voltage on the TSMC chip to outperform the Samsung chip by a small margin. People will be clamoring for the TSMC chip.
uh, no. you have to re-read what i wrote. this stuff is not magic, it's engineering. you pick a point in the design space you want to hit, then work all the parameters to yield that solution. if there were no samsung A9 to compare to, the TSMC A9 would be what it was and we would not be talking about all this stuff. both SoCs do what they were designed to do.
by the way, did you know that when the 40nm geometry first came out, it had much higher power consumption than 65nm due to leakage current? they eventually solved that problem but for a while 40nm parts were faster and hotter than 65nm, not faster and cooler.
it's bigger simply because the transistors in a 16nm process are physically larger than the transistors in a 14nm process. since the two designs have a comparable number of transistors (they are the same design after all) then the 16nm chip has to be a little bigger.
14nm/16nm is a length - it refers to the effective length of the channel in a MOSFET transistor in nanometers - which is what all these things are made out of.
Last edited by a moderator:
Yes. You are probably correct. It is probably way faster, but also way hotter and needs some serious juice. Apple is well known for under clocking chips to meet there stringent thermal guidelines.I said something similar in the other thread on here about this. Watch someone over the next handful of months figure out how to unlock the voltage on the TSMC chip to outperform the Samsung chip by a small margin. People will be clamoring for the TSMC chip.