D
Deleted member 13524
Guest
If they're getting rid of big.LITTLE, I wonder if Qualcomm will go back to quadcores for their flagship SoCs.
Qualcomm SoC & ARMv8 custom core discussions
- Thread starter Nebuchadnezzar
- Start date
loekf
Regular
"will begin sampling "on a leading edge FinFET process" in the second half of 2015. It should be in consumer devices by this time next year."
Why so vague ? Could be either 14 nm Samsung/GF or 16 nm TSMC. Actually learned today that 14nm Samsung/GF FinFET is FinFET transistors with a 20 nm backend (interconnect). Fast and efficient interconnect appearently requires new materials, which Samsung (and TSMC ??? ) didn't have the time for yet. So, 14nm @ Samsung is just a marketing name, likely it gives you the benefits of a half-node.
TSMC is in the same situation, and their 16nm process also relies on a 20nm BEOL.
That's not a PowerVR product, it's Kryo, not Kyro
Same here. We were given so few details, very frustrating...
That's not a PowerVR product, it's Kryo, not Kyro
Who cares? It's both greek to you literally
Kryo stands for cold in greek by the way Nebuchadnezzar
Legend
200-300mV reduction is hardly a half node. Don't really understand why people are ignoring the power and perf benefits so much just because the die shrink is only 15%.
I think not everyone is quite believing this is really all due to the new 14nm node, as opposed to samsung being lazy before and not properly optimizing it... That said if just half of it is really due to the new 14nm node this would already be quite a big deal...
Lazy ? As in game developer lazy ?
The reduction in power is due to the FinFET transistors used. You can lower supply voltage without suffering massive leakage. The low density is probably due to the interconnect lagging behind.
Cheers
loekf
Regular
Dutch website:
http://tweakers.net/nieuws/101871/htc-one-m9-heeft-last-van-oververhitting-bij-zware-belasting.html
Sounds like that the HTC One M9 with its Qualcomm SD 810 gets rather hot under load. Even up to 55 celsius. They measured this with an IR camera:
IMHO.. > 50 celsius is too hot too handle, unless you need a heater or a way to slowly cook an egg..
http://tweakers.net/nieuws/101871/htc-one-m9-heeft-last-van-oververhitting-bij-zware-belasting.html
Sounds like that the HTC One M9 with its Qualcomm SD 810 gets rather hot under load. Even up to 55 celsius. They measured this with an IR camera:
IMHO.. > 50 celsius is too hot too handle, unless you need a heater or a way to slowly cook an egg..
Nebuchadnezzar
Legend
Virtue of the Snapdragon 810. Keeps you warm in winter.
D
Deleted member 13524
Guest
It could just be a malfunctioning unit or SoC. This is just the first batch of the S810, after all.
I expect the M9 to be cooler than its competitors using the same S810, due to the aluminum unibody being used as a heatspreader.
I expect the M9 to be cooler than its competitors using the same S810, due to the aluminum unibody being used as a heatspreader.
Nebuchadnezzar
Legend
I wouldn't be so sure about that.The device goes on sale in a few days.
D
Deleted member 13524
Guest
So.. that means it can't be a malfunctioning unit?
Nebuchadnezzar
Legend
Unless HTC likes sending malfunctioning units for review testing...
Bad news is that all the other tested S810s I know of also have massive overheating/throttling/power consumption problems.
D
Deleted member 13524
Guest
It wouldn't be the first or the last time that a reviewer gets a faulty unit, or an early review sample with buggy software.
But you probably know better than I do about this M9 issue..
Nebuchadnezzar
Legend
There was nothing faulty about the Honor 6 in the end - the antenna just was horrible. I got only slightly better reception on the Mate 7 and other users of the Chinese version of the H6 had the same issues.
We'll see what HTC does, they're claiming it's a software issue. Which plainly means they're not throttling the SoC hard enough.
Meanwhile, the throttling is really bad on other devices. Not just because the peak CPU performance decreases to like half of what it could get within maybe a minute, but also because the single threaded performance goes down every bit as much as the multithreaded performance. Talk about poorly managing the power budget.
Do you know what the emergency temperature limits tend to be set to for devices like this? Where it'll just shut down? It could be getting pretty close...
Similar threads
