My own TSMC vs Samsung battery run-down test

[RonFromOregon]

I have here two iPhone 6s's, 128GB, Space Gray, one is TSMC and the other Samsung. I've restored them each from the same backup, removed the SIM from each, placed them in Airplane mode, set screen brightness to max, killed all background apps, charged to 100%, and otherwise made them as identical as I know how.

I ran GeekBench 3, and, just as other tests showed, after a time Samsung's battery fell to 48% while the TSMC's battery was at 54% (6% delta).

BUT THEN, I stopped GeekBench 3 and instead ran about 2 hours of slideshow from the camera roll (the two phones stayed exactly synced on the same photos), and then played a movie (again, the moves stayed exactly synced). After both the slideshow and movie run-downs, Samsung was at 21% and the TSMC at 27%. DELTA STILL AT 6%.

I think there's something going on in GeekBench 3 that is not "real world." Perhaps the heavy usage during the GeekBench test causes a heating effect or other phenomena that results in higher power consumption, but this is something that few people would see in everyday use. Based on my testing, I would say the TSMC and Samsung behave near identical in real-world usage. In fact, there may be other qualities about the Samsung that aren't easily put to a benchmark test that may show it to outlast the TSMC in lighter-than-normal use.

I'm going to repeat this testing at least two more times, and I'll post additional info if something comes up. In the meantime, please relax and know that whether you have TSMC or Samsung, you're holding one of the finest pieces of technology in the world. Be happy it's in your hands.

 

[heshirecat]

Any chance you could try looping the GFXBench GL 3.1 battery tests until each phone loses power? Thanks again and keep us posted!

 

[Lobwedgephil]

So the actual real world usage was the same, but the geek bench testing created the difference?

 

[roeiz]

thanks!

but still, a real world usage is at least watching hd videos...
maybe try to put them on YouTube and play some 10 hour HD content or something?

 

[Azzurro]

Thank you for the effort. This is comforting. I got Sammy inside my thingy.

 

[sprucemoose]

The reason you're seeing a smaller delta than other tests is because you put the screen brightness to max.

Think about it: the other things you've done — airplane mode and taking the SIM out — ensure battery is not being used on the radio components. But putting the brightness up does the opposite, it means a greater % of the battery is being used to power the displays, and therefore will minimise differences in the CPUs. It doesn't matter that both devices are at the same brightness, it's that the proportionally higher display drain is masking the differences in the CPUs.

If you look, you'll see that most other people have put the display brightness to MINIMUM when running Geekbench. Try that, and I guarantee you'll see a larger delta.

 

[RonFromOregon]

I removed the SIM because I only have one that's working, and I wanted to keep things even between the phones. I put the screen at full bright because that's what I typically did with my iPhone 6. You're right... by doing so, I moved some of the battery usage away from the CPU and into the screen, but again I think that's more "real world" (for me, at least).

My emphasis here is that I think GeekBench 3 is doing something to either cause an invalid test condition, or it's over-emphasizing the effect of hammering the CPU for hours on end.

I like roeiz's idea of streaming a super-long YouTube video. I'll try that.

 

[roeiz]

I like roeiz's idea of streaming a super-long YouTube video. I'll try that.

this will be great, thanks!

 

[iSunrise]

@RonFromOregon
Every video you watch will be played back by using the fixed-function units inside the A9 (ImgTec PowerVR VXD) and therefore will mostly measure display power consumption. Also, turning the screen brightness all the way up is wrong, you need to dim the display or all you're gonna measure again is also mostly display power consumption.

You should instead try to run something that will put the SoC to work, like Google Maps, a game loop or demo or an app that will need to use your CPU and GPU.

Geekbench is certainly an extreme case, yet, it will show you that even when only using CPU fixed-function units there are quite noticeable differences.

Good examples to benchmark, apart from fixed-function units are:
Google V8, Kraken, 3DMark, GFXBench

 

[mmartins25]

Thank you very much for the info, we're exicted about the new tests.

 

[heshirecat]

@RonFromOregon
Every video you watch will be played back by using the fixed-function units inside the A9 (ImgTec PowerVR VXD) and therefore will mostly measure display power consumption. Also, turning the screen brightness all the way up is wrong, you need to dim the display or all you're gonna measure again is also mostly display power consumption.

You should instead try to run something that will put the SoC to work, like Google Maps, a game loop or demo or an app that will need to use your CPU and GPU.

Geekbench is certainly an extreme case, yet, it will show you that even when only using CPU fixed-function units there are quite noticeable differences.

Good examples to benchmark, apart from fixed-function units are:
Google V8, Kraken, 3DMark, GFXBench

A good post.

 

[RonFromOregon]

I continued my testing by playing a YouTube video (same video playing on each phone). After 2-3 hours, the TMSC was at 58% and Samsung at 56%. Both phones just slightly warm to the touch. 2% Delta.

Then I ran the Kraken 1.1 test in Safari. I did just one iteration. Following that, I downloaded the GFXBench OpenGL test and its data files on each phone. At that point, TMSC at 52%, Samsung at 50%. 2% Delta

After running the GFXBench OpenGL test twice on each phone, TMSC at 31% and Samsung at 27%. 4% Delta. Phones noticeably warm to the touch.

It seems clear that the TMSC is more power efficient during benchmark testing, but in everyday use the difference is almost negligible.

Somebody within Apple is leaning back in a chair right now thinking, "If we had just made TMSC less efficient, this wouldn't be an issue."

 

[j2ee]

I have here two iPhone 6s's, 128GB, Space Gray, one is TSMC and the other Samsung. I've restored them each from the same backup, removed the SIM from each, placed them in Airplane mode, set screen brightness to max, killed all background apps, charged to 100%, and otherwise made them as identical as I know how.

I ran GeekBench 3, and, just as other tests showed, after a time Samsung's battery fell to 48% while the TSMC's battery was at 54% (6% delta).

BUT THEN, I stopped GeekBench 3 and instead ran about 2 hours of slideshow from the camera roll (the two phones stayed exactly synced on the same photos), and then played a movie (again, the moves stayed exactly synced). After both the slideshow and movie run-downs, Samsung was at 21% and the TSMC at 27%. DELTA STILL AT 6%.

I think there's something going on in GeekBench 3 that is not "real world." Perhaps the heavy usage during the GeekBench test causes a heating effect or other phenomena that results in higher power consumption, but this is something that few people would see in everyday use. Based on my testing, I would say the TSMC and Samsung behave near identical in real-world usage. In fact, there may be other qualities about the Samsung that aren't easily put to a benchmark test that may show it to outlast the TSMC in lighter-than-normal use.

I'm going to repeat this testing at least two more times, and I'll post additional info if something comes up. In the meantime, please relax and know that whether you have TSMC or Samsung, you're holding one of the finest pieces of technology in the world. Be happy it's in your hands.

How about you do some "real life" test. Like you just let the two phones play youtube from full battery to no battery. Then you find a script and keep web browsing from full battery to no battery. Then you just play a movie from full battery and no battery. Then we can take the average or look at each timing.

 

[j2ee]

I have here two iPhone 6s's, 128GB, Space Gray, one is TSMC and the other Samsung. I've restored them each from the same backup, removed the SIM from each, placed them in Airplane mode, set screen brightness to max, killed all background apps, charged to 100%, and otherwise made them as identical as I know how.

I ran GeekBench 3, and, just as other tests showed, after a time Samsung's battery fell to 48% while the TSMC's battery was at 54% (6% delta).

BUT THEN, I stopped GeekBench 3 and instead ran about 2 hours of slideshow from the camera roll (the two phones stayed exactly synced on the same photos), and then played a movie (again, the moves stayed exactly synced). After both the slideshow and movie run-downs, Samsung was at 21% and the TSMC at 27%. DELTA STILL AT 6%.

I think there's something going on in GeekBench 3 that is not "real world." Perhaps the heavy usage during the GeekBench test causes a heating effect or other phenomena that results in higher power consumption, but this is something that few people would see in everyday use. Based on my testing, I would say the TSMC and Samsung behave near identical in real-world usage. In fact, there may be other qualities about the Samsung that aren't easily put to a benchmark test that may show it to outlast the TSMC in lighter-than-normal use.

I'm going to repeat this testing at least two more times, and I'll post additional info if something comes up. In the meantime, please relax and know that whether you have TSMC or Samsung, you're holding one of the finest pieces of technology in the world. Be happy it's in your hands.

Can you take a pic and somehow show you have 2 phones with different chips? You are new and no pictures to prove anything at all.

 

[roeiz]

I continued my testing by playing a YouTube video (same video playing on each phone). After 2-3 hours, the TMSC was at 58% and Samsung at 56%. Both phones just slightly warm to the touch. 2% Delta.

Then I ran the Kraken 1.1 test in Safari. I did just one iteration. Following that, I downloaded the GFXBench OpenGL test and its data files on each phone. At that point, TMSC at 52%, Samsung at 50%. 2% Delta

After running the GFXBench OpenGL test twice on each phone, TMSC at 31% and Samsung at 27%. 4% Delta. Phones noticeably warm to the touch.

It seems clear that the TMSC is more power efficient during benchmark testing, but in everyday use the difference is almost negligible.

Somebody within Apple is leaning back in a chair right now thinking, "If we had just made TMSC less efficient, this wouldn't be an issue."

Many thanks.
This was very helpful.

 

[sprucemoose]

I continued my testing by playing a YouTube video (same video playing on each phone). After 2-3 hours, the TMSC was at 58% and Samsung at 56%. Both phones just slightly warm to the touch. 2% Delta.

Then I ran the Kraken 1.1 test in Safari. I did just one iteration. Following that, I downloaded the GFXBench OpenGL test and its data files on each phone. At that point, TMSC at 52%, Samsung at 50%. 2% Delta

After running the GFXBench OpenGL test twice on each phone, TMSC at 31% and Samsung at 27%. 4% Delta. Phones noticeably warm to the touch.

It seems clear that the TMSC is more power efficient during benchmark testing, but in everyday use the difference is almost negligible.

Somebody within Apple is leaning back in a chair right now thinking, "If we had just made TMSC less efficient, this wouldn't be an issue."

Thanks for sharing this testing with us. Much appreciated.

Just another note on the video playback test — the iPhone has dedicated video decoding hardware, so the CPU is doing minimal work while playing video and this is why you don't see any gap increase between the two during that test.

I think your testing highlights that the severity of this issue depends entirely on what you do with your phone. Heavy gamers should be concerned, casual users much less so.

 

[j2ee]

I continued my testing by playing a YouTube video (same video playing on each phone). After 2-3 hours, the TMSC was at 58% and Samsung at 56%. Both phones just slightly warm to the touch. 2% Delta.

Then I ran the Kraken 1.1 test in Safari. I did just one iteration. Following that, I downloaded the GFXBench OpenGL test and its data files on each phone. At that point, TMSC at 52%, Samsung at 50%. 2% Delta

After running the GFXBench OpenGL test twice on each phone, TMSC at 31% and Samsung at 27%. 4% Delta. Phones noticeably warm to the touch.

It seems clear that the TMSC is more power efficient during benchmark testing, but in everyday use the difference is almost negligible.

Somebody within Apple is leaning back in a chair right now thinking, "If we had just made TMSC less efficient, this wouldn't be an issue."

I think all the videos or GFX tests would use GPU part but not CPU part

 

[j2ee]

I continued my testing by playing a YouTube video (same video playing on each phone). After 2-3 hours, the TMSC was at 58% and Samsung at 56%. Both phones just slightly warm to the touch. 2% Delta.

Then I ran the Kraken 1.1 test in Safari. I did just one iteration. Following that, I downloaded the GFXBench OpenGL test and its data files on each phone. At that point, TMSC at 52%, Samsung at 50%. 2% Delta

After running the GFXBench OpenGL test twice on each phone, TMSC at 31% and Samsung at 27%. 4% Delta. Phones noticeably warm to the touch.

It seems clear that the TMSC is more power efficient during benchmark testing, but in everyday use the difference is almost negligible.

Somebody within Apple is leaning back in a chair right now thinking, "If we had just made TMSC less efficient, this wouldn't be an issue."

So base on your tests so far, the different is between 6% to 2%, depend on what the users do.

 

[j2ee]

If only 6% to 2% different is true, it is really not a big deal. I am sure even the battery part is made by different brands, and that would be at least like 1-5% different because of different battery brand with different quality....and there are so many other parts in the phones to affect battery which could be all made by different brands different quality...

 

[roeiz]

Can you take a pic and somehow show you have 2 phones with different chips? You are new and no pictures to prove anything at all.

the man doing tests for you people in this forum,
and you want to see proof??
take what he gives you from his own time..
read the whole thread..
why would he bluff?

 

[roeiz]

If only 6% to 2% different is true, it is really not a big deal. I am sure even the battery part is made by different brands, and that would be at least like 1-5% different because of different battery brand with different quality....and there are so many other parts in the phones to affect battery which could be all made by different brands different quality...

these tests are a good start.
but indeed, maybe the Youtube video run is not the perfect option..
and also, these % can get bigger within more hours of use..
who knows?
take this 4K battery use on a TSMC for instance (quoted from another thread):


I shot stills/video during a 2.5 hour workshop. Total camera time probably ~1 hour.

-9:03 am: ~90% battery

-11:33 am: ~33% battery

Total 4K video capture ~30 minutes plus 60 stills.

*Wifi was on and uploading >5 GB to my iCloud during the shoot.


i bet that if this was done on both chips,
we would definitely see a bif % difference, a lot more than 5%
i might be doing stuff like that with my phone,
and that's a real bummer imo.

 

[j2ee]

the man doing tests for you people in this forum,
and you want to see proof??
take what he gives you from his own time..
read the whole thread..
why would he bluff?

Well, samsung even paid to do marketing stuff in forum, see the news:
http://www.digitaltrends.com/mobile/samsung-fined/
Who knows if Apple does more and better
I have worked for two famous brands in my city, both did a lot of stuff in forums to promote their products, both have serious success in sales. For me, fake comments by companies in forum are normal, it just depends on do you like the product. Of course there are real people share real experience so I want to see if this is fake or real.

 

[roeiz]

Well, samsung even paid to do marketing stuff in forum, see the news:
http://www.digitaltrends.com/mobile/samsung-fined/
Who knows if Apple does more and better

ok.. but it doesn't SEEM like this user is marketing Apple.
he does various test, discusses them, and coming up with ideas to check.
he has various results, and not all of them are so good for Apple...

 

[j2ee]

ok.. but it doesn't SEEM like this user is marketing Apple.
he does various test, discusses them, and coming up with ideas to check.
he has various results, and not all of them are so good for Apple...

I understand, sound like his reviews is real, but not hard to take a photo to prove right? XD

 

[RonFromOregon]

You guys crack me up. No, I'm not paid by anybody to do this. I bought my first 6s, no contract Verizon model from the Apple store. I bought Verizon because I wanted to run it on AT&T and didn't want AT&T to lock it. Right after that, AT&T started allowing WiFi Calling but it failed for me because the Verizon model isn't okayed for WiFi calling yet on AT&T. That day, Apple started selling a "SIM-free" model which a) won't lock to AT&T, b) work with WiFi calling (I haven't tested that yet), and c) give me "Band 30", whatever that is. So I bought a second 6s, this time Sim-free, with plans to return the first one once I got the second one up and going.

That's when I saw this whole "chipgate" thing exploding online. I downloaded BMSSM and saw that I have one of each of the A9 models. So I thought I'd try some real-world testing to see if the controversy is true. Turns out that it's very difficult to do real-world comparisons between two phones, especially when I need to actually use one of them as a phone and I only have one active SIM card.

I'm running GeekBench Battery test now on both phones. I don't want to interrupt the testing, but when it's through I'll run BMSSM on each phone, take a picture, and post it here. I'm not sure how that will help, but I suppose it can't hurt.