Thanks to Samsung's new models, heat-pipes might become popular in smartphones

Samsung apparently wasn't convinced that the Snapdragon 820 chipset wouldn't overheat like its predecessor, the Snapdragon 810 chipset. After all, it was Samsung that practically led manufacturers to avoid the Snapdragon 810 SoC whenever possible last year. For Samsung, this led the company to employ their own Exynos 7420 chipset on all of its high end models, even in regions like the U.S. where traditionally a Snapdragon solution was used...
This is a discussion for a news article. To read the whole news, click here

106 Comments

75. TyrionLannister unregistered

LOL, ever heard about something called IPC? By your logic, Snapdragon 805 should obliterate single core benchmarks, and AMD's 100$ CPUs should destroy intel 5960x.

85. S.R.K.

Posts: 678; Member since: Feb 11, 2016

A53 to A57 is a big difference in power.

43. vincelongman

Posts: 5606; Member since: Feb 10, 2013

From the initial teardown, it seems like it doesn't have the magnesium mid frame like older Samsung devices (probably because it has a stronger frame, so the metal mid frame isn't necessary) So a glass back and no metal mid frame like older models, so its not too surprising there's a heat pipe I really hope not There seems to be mixed reports so far about the G5 and S7 AnandTech's hands on with the Mi 5 had some early power consumption numbers 820 (2x 2.15GHz Kryo HP + 2x 1.59GHz Kryo LP) 1 core | 2 cores | 3 cores | 4 cores 2055 | 3330 | 4147 | 4735 7420 (4x 2.1 Ghz A57 + 4x 1.5 Ghz A53) 1 core | 2 cores | 3 cores | 4 cores 1619 | 2969 | 4186 | 5486

45. vincelongman

Posts: 5606; Member since: Feb 10, 2013

I looked again, it looks like the metal mid frame is still there (also sorry mid frame might not be a good way to describe it) Hard to tell without full teardown

46. TyrionLannister unregistered

You see, it's higher power cores consumes way more power than even the 7420. And unlike 7420, it doesn't have A53s to use in normal scenarios. I see you pulled these from Anantech mi5 article. There was Huawei mate 8 too whose power consumption destroyed the 820. Considering 820 is on a newer architecture and process, this much power consumption is scary. And having worse perf/watt compared to A72 even after process advantage of 14 nm LPP is downright shameful. Also in the same article: "Again it's to be noted that the device can get hot and preliminary power measurements show that it's very unlikely that the Mi 5 will be able to sustain these scores for extended periods of time." 820 seems to be a heater. They simply tried too hard to match Apple with an inferior 3-way superscalar design, and the results show.

51. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Those were the first A72 cores on any Android device.

58. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Are you seriously unaware ? SD820 has 2+2 big little core, but still consumes more power than 4+4 7420. 7420 came out in April while 820 will be manufactured later this year, yet 7420 outperforms 820. Samsung would release M processor that is claimed to score geekbench over 7,000, and SD820 would be ditched by Samsung again. Snapdragon has never performed better than Exynos overall.

59. TyrionLannister unregistered

did you even read what I was replying to? We are talking about power consumption here. And no, 820 consumes more power per core, not overall. Please stop replying to me without even understanding what I'm talking about. I beg you.

61. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Since it's a 2+2 core (2 'High' speed Kryo + 2 'Medium' speed Kryo) no wonder it couldn't beat the 4+4 Big.Little Exynos. But in all probability on a per core basis Kryo doesn't seem as ahead of the pack like Krait (which was faster/more efficient than A9 and equal/much more efficient than A1) , but more like their Scorpion core which though faster/better than the A8, couldn't keep up with the later A9. Their approach seems to be a lot like Apple which have just 2 but 'very powerful' cores in theie phones. Qualcomm seems to be doing the same but better with 2 high-speed 'very powerful' Kryo's (faster than 2 high-speed A72's) and 2 medium-speed but still 'powerful' Kryo's (equal to 2 high-speed A72's). Ofcourse this is assuming that the clock speed of 2+2 Kryo is 2.2 & 1.7 GHz respectively

62. S.R.K.

Posts: 678; Member since: Feb 11, 2016

At 10twp. Old Exynos A15s were at 12twp, and..... Why does it bother you if it's higher? I understand it drains battery. But wth wants efficiency crap. We are talking about pure brutal power? Why don't you use crappy 6 series Snapdragon. Let the extreme power users enjoy hight torque and higher clock speed.

64. TyrionLannister unregistered

Who wants efficiency crap? Are you kidding me! Phones are made to last long. I want the best perf/watt. 6xx snapdragon series is slow as all hell. I want something that's fast and efficient. And most users want that. What good is power if your phone can't last a day. Extreme users! LMAO. I have a 5820k at home, OCed to 4.5 GHz. But power is fine there as it runs on wall power. Not here.

67. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Carry a powerbank aswell if you can to recover what and why it's clocked at 12twp.

68. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Phones are made to last long then ask oems to use 4000 or 5000 Mah battery, you have 2 option either the weight goes up to 230g or make the width at 11mm. What would you like? Also ask oems to use 9nm, then the smaller the Cpu die the rubbish it is. I'd take 22nm big die Cpu over crappy rubbish small 14nm Bs. Look at Apple, they went for higher 22nm.

63. vincelongman

Posts: 5606; Member since: Feb 10, 2013

The 950 is 16FF+ Should be very close to 14LPP Maybe even better considering 16FF outperformed 14LPE for the A9 IMO his comparison with the 950 isn't really fair since he's counting all 4 cores for the 820 but only 4/8 cores for the 950 (although that's a bit unfair on the 950) 820 (2+2) = 4735 mW 950 (4x A72) = 3734 mW 950 (4x A53) = ~800 mW (my rough estimation, he didn't measure it) 950 (4+4) = ~ 4500 mW 7420 (4x A72) = 5486 mW 7420 (4x A53) = 1026 mW 7420 (4+4) = 6512 mW BTW those figures are max values. Qualcomm can still use the Kryo LPs at say 1GHz, for probably similar perf/watt to the A53s in the 7420 The perf/watt curves weren't given But I agree, the perf/watt of the Kyros does seem to be worse than the A72s in this case Very disappointing considering we'd expect the Kyros to be improved A72s Hopefully they get improved as 14LPP matures and there's a 821/825 in 6 months

65. TyrionLannister unregistered

Yeah, they do look pretty bad. Since 4+4 configs usually only use 2 big cores in 99% of the tasks(from Anandtech).

69. S.R.K.

Posts: 678; Member since: Feb 11, 2016

And my Z5 rooted with custom kernel uses all 4 Cpu at power mode by SetCpu from Android store. 2 cores are crap and slow.

105. pmsap

Posts: 105; Member since: May 26, 2015

What I don't get is that apple's SoC's are out-there to be tested, analyzed, "copied". I don't get why can't samsung improve their power output while maintaining heat low as apple does. It's not like they can«t test the technology. It's there.

109. alex3run

Posts: 715; Member since: May 18, 2014

4,7W TDP isn't scary. It is pretty acceptable though quad stock A72s are faster and eat less at the same time what I didn't expect. And E7420 eats more tnan SD820, check Anandtech. As for core architecture: while we know nothing how to build a CPU core if I were you I wouldn't think QC engineers are stupid.

60. S.R.K.

Posts: 678; Member since: Feb 11, 2016

By the way Kyro is even slower than a A57 at the same clock speed. Kyro: only a Krait with some improvements and 64bit extensions

42. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Why don't you use a cover and stop complaining? Why is the heat bothering you? It's not that hot that's it's unbearable to hold? What you can do is start buying midrange Android phones. You won't achieve the power but have a standard spec phone.

47. TyrionLannister unregistered

You have no idea how bad is it in tropical countries with 45 degree celcius temperatures. A heater phone gets really uncomfortable. I had an S4 which heated like the 810. It just feels bad when your phone even warms up. Now on the S6 and it's awesome, never gets even warm to touch.

50. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Why don't you put on a cover. My Z5 is extremely overclocked to 2.7ghz,. 45degree is bearable, my Z5 reaches 80 degrees. I ain't nagging, but for sure it's 50% faster then it was as standard. All these people complained about lags on Touchwiz, I can multitasking extremely on my Note5 with not a single Bs lag. I understand the battery in getting rinsed, but I carry a powerbank 3000Mah. The entire phone does not spread heat, only the top left section heats up. I mean if it was the entire phone spreading heat, that's a different issue.

56. TyrionLannister unregistered

Cover is a bad idea. And overclocking to 2.7 GHz is terrible idea. The SoC will wear down pretty fast and the performance increase in real life would be marginal as most of the time the bottleneck is storage. Cover just keeps the heat inside phone, damaging the internals. Not to mention the design gets compromised and your phone basically becomes a thick cheap plastic phone.

71. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Not marginal, I've even took the voltages to the highest extreme. The difference is huge vs the standard.

76. TyrionLannister unregistered

The issue is, you're not going to be benefit even if your CPU is faster. It's called Ahmdahl's law.

83. adrianzka

Posts: 64; Member since: Apr 25, 2013

Does your S6 heat up when fast charging if I may ask? My note5 gets scorching hot after about 15 minutes around the metal sides. It cools down later on though.

52. adrianzka

Posts: 64; Member since: Apr 25, 2013

Actually, I'm pretty sure throttling would happen about the same time as with no heat sink. Because i think the measured heat would be that of the SoC, and not the smartphone itself. When the SoC went to a certain temperature threshold, it would throttle down performance. It's just that a heatsink would spread out the heat generated slightly better than usual, as well as be easier during repairs in comparison to heat films (i dont know what those thin copper films that are used normally is called). So while the heatpipes aren't absolutely necessary, they are quite welcome.

54. S.R.K.

Posts: 678; Member since: Feb 11, 2016

Even better are the liquid cool variants.

55. TyrionLannister unregistered

If the heat-pipes spread heat, the SoC temperature will drop down and it won't throttle.

82. adrianzka

Posts: 64; Member since: Apr 25, 2013

It still will, albeit would take a little longer compared to without one. During the time it doesn't throttle, more heat will build up as expected. However, the more even distribution should still be more than able to compensate for the added heat as more heat exhange occur with the surroundings. The good thing about heat pipes are when the SoC throttles, the temperature of the SoC will definitely be higher than the heatpipes, so the SoC will cooldown faster after throttling, as compared to relying on heat convection without one, since the higher internal temperature for the phone could lead to potential problems towards other components, most notably batteries. Sorry for the long essay :^
This copy is for your personal, non-commercial use only. You can order presentation-ready copies for distribution to your colleagues, clients or customers at https://www.parsintl.com/phonearena or use the Reprints & Permissions tool that appears at the bottom of each web page. Visit https://www.parsintl.com/ for samples and additional information.