HTC One M9 overheating issue FIXED via software update

[dann127]

They committed to their own processor BECAUSE the 810 was overheating. How hard is this to understand?

News broke that the 810 was overheating when? The beginning of December? Let's give Samsung the benefit of the doubt and assume they knew about this issue early. Say mid October. At that point they decide to do a full 180 and use only their own chip for the phone. In 5.5ish months they would need to do a major board redesign while updating the firmware to be compatible with the new chip and buttoning up the software so it ran correctly on that new firmware. All of this taking place in probably half the time of their normal development process (assuming ~1 year) plus placing orders and receiving finished products before MWC. This would take an absurd amount of engineering effort. These parts aren't just swappable. Their who knows if they even have the same number of outputs.

You really think Samsung would waste their money focusing on HTC when HTC's marketshare was already plummeting all around the world?

Yes. Assume HTC only sells 1 million M9s. Maybe all of those buyers would have bought an S6 or S6 edge instead. Maybe only half or a quarter. It's still a significant number. If you can get Android people doubting another Android product you're likely removing them from the competition. Sounds like a pretty fantastic plan if you're trying to keep up with Apple's total number of devices sold. Samsung isn't most likely to steal Apple customers. They're most likely to steal customers from HTC, LG, and Sony. People already in the Android ecosystem.

 

[John Kar]

News broke that the 810 was overheating when? The beginning of December? Let's give Samsung the benefit of the doubt and assume they knew about this issue early. Say mid October. At that point they decide to do a full 180 and use only their own chip for the phone. In 5.5ish months they would need to do a major board redesign while updating the firmware to be compatible with the new chip and buttoning up the software so it ran correctly on that new firmware. All of this taking place in probably half the time of their normal development process (assuming ~1 year) plus placing orders and receiving finished products before MWC. This would take an absurd amount of engineering effort. These parts aren't just swappable. Their who knows if they even have the same number of outputs.

Are you conveniently forgetting that Samsung always releases 2 variants of the Galaxy S phones? One with SD and the other with Exynos?

Did you also forget what I said concerning Samsung and their side-projects as backups?

It doesn't matter what Qualcomm released as the SD 810. It could be the best chip in the entire world with 500 petaflops performance with no throttling whatsoever, but Samsung would have still made the Exynos 7420 even if they were never going to use it.

You are operating under the assumption that I think that the Exynos 7420 started development in October when everything I'm saying contradicts my ever believing that.
I believe the 7420 was under development since the S5's launch, since Samsung's original plan was using the Exynos in non-US markets like they always do. They only decided to expand it to the US market after finding out the 810 overheated.

It wasn't "hard" like you're claiming. By that time when Samsung was testing the 810, the Exynos 7420 was already finished. Giving the green-light to manufacture more 7420s could be done with a simple e-mail.

Yes. Assume HTC only sells 1 million M9s. Maybe all of those buyers would have bought an S6 or S6 edge instead. Maybe only half or a quarter. It's still a significant number. If you can get Android people doubting another Android product you're likely removing them from the competition. Sounds like a pretty fantastic plan if you're trying to keep up with Apple's total number of devices sold. Samsung isn't most likely to steal Apple customers. They're most likely to steal customers from HTC, LG, and Sony. People already in the Android ecosystem.

Apple's customers are more loyal than other customers of other companies, but we're already seeing lots of iPhone users take note of the S6, especially the Edge version. Even Apple is a little worried about the S6/S6 Edge, which is why they set the iWatch order date to the same date as the S6 release date.

 

[dann127]

I apologize, I did forget. Although not conveniently like you assumed. Thank you for the condescension though. I would still contend that it is not as easy as an e-mail. Giving the green light is simple. Having the capacity to manufacture enough chips for the US-market is not. I'm not saying that they don't, but making a few million more chips isn't that simple. Plus eating the loss of possibly millions of dollars for R&D on the 810 variant isn't something to sneeze at.

Apple customers are more "loyal" because if they switch they lose any and all app purchases that they've made in iOS and the App Store. I don't believe you can claim with any sort of certainty that Apple is worried. They may be but releasing a new device the same day as a non-competing product isn't necessarily worry.

Please don't misconstrue any of my post as making a definite claim. Everything being said by anyone on these forums concerning devices which have not been released is speculation and should be taken as such.

 

[TimmE]

Can you name the highly regarded tech websites that proved the 810 overheats after the article debunking the rumor?

Posted via the Android Central App

 

[John Kar]

I apologize, I did forget. Although not conveniently like you assumed. Thank you for the condescension though. I would still contend that it is not as easy as an e-mail. Giving the green light is simple. Having the capacity to manufacture enough chips for the US-market is not. I'm not saying that they don't, but making a few million more chips isn't that simple. Plus eating the loss of possibly millions of dollars for R&D on the 810 variant isn't something to sneeze at.

Apple customers are more "loyal" because if they switch they lose any and all app purchases that they've made in iOS and the App Store. I don't believe you can claim with any sort of certainty that Apple is worried. They may be but releasing a new device the same day as a non-competing product isn't necessarily worry.

Please don't misconstrue any of my post as making a definite claim. Everything being said by anyone on these forums concerning devices which have not been released is speculation and should be taken as such.

It is easy for Samsung, since no one else is using their 14nm finfet fabs yet.
The iPhone 6s/7 or w/e they're going to call it probably isn't finished yet, so I doubt they're mass-producing the chips right now. Qualcomm isn't going to use Samsung's fabs until next year, so Samsung has their entire capacity all to themselves.

 

[dann127]

It is easy for Samsung, since no one else is using their 14nm finfet fabs yet.
The iPhone 6s/7 or w/e they're going to call it probably isn't finished yet, so I doubt they're mass-producing the chips right now. Qualcomm isn't going to use Samsung's fabs until next year, so Samsung has their entire capacity all to themselves.

Assuming all of their equipment is compatible with the 14nm process. Again too many assumptions to say anything definitive.

Posted via Android Central App

 

[John Kar]

So I just checked the M9's new benchmark score and it seems there is quite a bit of throttling.



m9 - Geekbench Search - Geekbench Browser

I've checked all the benchmarks of the M9 in geekbench and the single-core performance dropped from 1200-1300 before March 18th to 700-900 after March 18th.
Multicore performance also dropped from 3800-4000 to 2900-3500.

 

[Habiib]

So I just checked the M9's new benchmark score and it seems there is quite a bit of throttling.

View attachment 167613

m9 - Geekbench Search - Geekbench Browser

I've checked all the benchmarks of the M9 in geekbench and the single-core performance dropped from 1200-1300 before March 18th to 700-900 after March 18th.
Multicore performance also dropped from 3800-4000 to 2900-3500.

Looking at the data, I grabbed the minimum and maximum single/multi-core benchmark numbers:

From 2/16 - 3/18
[table="width: 500"]
[tr]
[td]Minimum Single-Core[/td]
[td]Maximum Single-Core[/td]
[/tr]
[tr]
[td]698[/td]
[td]1335[/td]
[/tr]
[/table]

[table="width: 500"]
[tr]
[td]Minimum Multi-Core[/td]
[td]Maximum Multi-Core[/td]
[/tr]
[tr]
[td]2270[/td]
[td]4288[/td]
[/tr]
[/table]


From 3/19 - Present
[table="width: 500"]
[tr]
[td]Minimum Single-Core[/td]
[td]Maximum Single-Core[/td]
[/tr]
[tr]
[td]715[/td]
[td]1335[/td]
[/tr]
[/table]

[table="width: 500"]
[tr]
[td]Minimum Multi-Core[/td]
[td]Maximum Multi-Core[/td]
[/tr]
[tr]
[td]2219[/td]
[td]4131[/td]
[/tr]
[/table]

It appears that the highest single-core score dropped 13.25% and the highest multi-core score dropped 3.6% after 3/18. I'm not sure how conclusive this all is being that we are only looking at ~1 1/2 days worth of data for the post 3/18 submissions. What's interesting to me are the memory performance and floating point scores during those time frames.

 

[John Kar]

Looking at the data, I grabbed the minimum and maximum single/multi-core benchmark numbers:

From 2/16 - 3/18
[table="width: 500"]
[tr]
[td]Minimum Single-Core[/td]
[td]Maximum Single-Core[/td]
[/tr]
[tr]
[td]698[/td]
[td]1335[/td]
[/tr]
[/table]

[table="width: 500"]
[tr]
[td]Minimum Multi-Core[/td]
[td]Maximum Multi-Core[/td]
[/tr]
[tr]
[td]2270[/td]
[td]4288[/td]
[/tr]
[/table]


From 3/19 - Present
[table="width: 500"]
[tr]
[td]Minimum Single-Core[/td]
[td]Maximum Single-Core[/td]
[/tr]
[tr]
[td]715[/td]
[td]1335[/td]
[/tr]
[/table]

[table="width: 500"]
[tr]
[td]Minimum Multi-Core[/td]
[td]Maximum Multi-Core[/td]
[/tr]
[tr]
[td]2219[/td]
[td]4131[/td]
[/tr]
[/table]

It appears that the highest single-core score dropped 13.25% and the highest multi-core score dropped 3.6% after 3/18. I'm not sure how conclusive this all is being that we are only looking at ~1 1/2 days worth of data for the post 3/18 submissions. What's interesting to me are the memory performance and floating point scores during those time frames.

I don't think looking at the range is the best idea.
You should look at the median/average.

 

[neo905]

The best idea is waiting for the M9 to be released and see if it suits your personal needs. These numbers don't really mean squat to me or 99% of people heading into their carriers to buy any phone.

Posted via the Android Central App

 

[Alex_NJ]

As soon as the embargo is lifted all will be well

 

[Adam_ation]

So I just checked the M9's new benchmark score and it seems there is quite a bit of throttling.

I had wondered but its nice to confirm that you are the guy known as "Applesexual", XDA's resident troll.

You hide your troll bait behind a facade of concern for the HTC M9.

I'm so glad the exact same thing you posted on XDA got thread closed by a mod straight away, when will you learn? The sooner your banned the better - Adam

 

[John Kar]

The best idea is waiting for the M9 to be released and see if it suits your personal needs. These numbers don't really mean squat to me or 99% of people heading into their carriers to buy any phone.

Posted via the Android Central App

Seeing people say that the S6 is the fastest phone they've ever used could be down to their storage speed and high single-core performance.

 

[dann127]

I don't think looking at the range is the best idea.
You should look at the median/average.

Any because it doesn't conform to the point you want to make? Having a sample size of 1.5 days is not nearly enough to get a reliable average as one value could skew the data significantly. Min max isn't the best way to represent the data either but if you present all of it together maybe you can start to draw conclusions after you have an adequate sample size.

In the end benchmarks are relatively meaningless and don't determine how a phone will perform for everyday users.

Posted via Android Central App

 

[TimmE]

Seeing people say that the S6 is the fastest phone they've ever used could be down to their storage speed and high single-core performance.

Seeing people say that about the S6 leads me to believe they are intoxicated because nothing running touch wiz is fast.

Posted via the Android Central App

 

[John Kar]

Any because it doesn't conform to the point you want to make? Having a sample size of 1.5 days is not nearly enough to get a reliable average as one value could skew the data significantly. Min max isn't the best way to represent the data either but if you present all of it together maybe you can start to draw conclusions after you have an adequate sample size.

In the end benchmarks are relatively meaningless and don't determine how a phone will perform for everyday users.

Posted via Android Central App

Assuming that benchmark scores of a phone are normally distributed, we have more than enough data points already to make a solid guess that there is throttling going around.

Using the highest and minimum score to determine performance hits is a terrible way to assess throttling. as there will always be outliers in the sample size.

A good way to think about it is wealth.
In all countries, there are poor people and rich people.
Mexico's richest man, Carlos Slim, is worth $73 billion. In the US, the richest man, Bill Gates, is worth $79.3.

In Mexico, the poorest man is worth $0. In the US, the poorest man is worth $0.

Would you say the typical Mexican man is as wealthy as the typical American man because there's not much difference between the poorest citizens and between the wealthiest citizens?

 

[John Kar]

Seeing people say that about the S6 leads me to believe they are intoxicated because nothing running touch wiz is fast.

Posted via the Android Central App

Really? Because I just tried the S6 yesterday and it is by far the smoothest phone I've EVER used. It even beats the iPhone in terms of smoothness.

There are people talking about it on reddit: Bestbuy with Samsung experience getting Galaxy S6 3/28 for display : Android

 

[TimmE]

Got any video of that?

Posted via the Android Central App

 

[John Kar]

Got any video of that?

Posted via the Android Central App

Unfortunately, no, as I'm didn't want to look like a dork in front of everyone and record my using a display device.

 

[Habiib]

Assuming that benchmark scores of a phone are normally distributed, we have more than enough data points already to make a solid guess that there is throttling going around.

Using the highest and minimum score to determine performance hits is a terrible way to assess throttling. as there will always be outliers in the sample size.

A good way to think about it is wealth.
In all countries, there are poor people and rich people.
Mexico's richest man, Carlos Slim, is worth $73 billion. In the US, the richest man, Bill Gates, is worth $79.3.

In Mexico, the poorest man is worth $0. In the US, the poorest man is worth $0.

Would you say the typical Mexican man is as wealthy as the typical American man because there's not much difference between the poorest citizens and between the wealthiest citizens?

I drew no conclusion at all, I only gave the measured performance differences. There's a difference between giving calculations and coming up with an explanation based on some analysis.

To compare < 2 days worth of data against ~20 days worth of data and draw a conclusion doesn't sound realistic. The disparity in the sample size is pretty significant, so it doesn't matter what approach is taken to compile the data IF the goal is to reach a conclusion. Also there are so many factors that can impact the results of a synthetic benchmark test, that using these numbers alone for making such a determination is kind of odd.

Posted via the Android Central App