QQ Snapdragon 810 overheating issues: delay?

[witeken]

Problems in Qualcomm Snapdragon Set Alarm Bells Ringing for Samsung, LG

ARM's A15 successor seems to have power consumption problems, 'despite' the 20nm process.

Local smartphone makers are nervous at the prospect of a delay in the launch of new models next year, including the Galaxy S6 and the G4. It is unclear whether or not the supply of the Snapdragon 810 will exist in the first half of next year due to technical problems such as overheating and a decline in speed.

Samsung is likely to solve the problem by featuring its own Exynos chips in the Galaxy S6, but LG seems to be in trouble. Even though the company has its first AP, NUCLUN, it is not better than entry-level APs. If Qualcomm cannot supply the Snapdragon 810, it won't be easy for LG to find an alternative for the G4.

“Qualcomm is faced with hard-to-solve problems. The Snapdragon 810 overheats when it reaches a specific voltage. It also slows down owing to problems with the RAM controller connected to the AP. In addition, there is an error in the driver of the Adreno 430 GPU,” said an industry source on Dec. 2.

This would be the ideal moment for Samsung, LG, etc. to switch to Moorefield, since there won't be any better alternative.

 

[rushmore]

If this is true, they would be set back at least three months and that assumes they know what to fix now.

 

[Arachnotronic]

it won't be easy for LG to find an alternative for the G4.

Balderdash! That 1.5 GHz big.LITTLE Cortex A15 NUCLUN they've got will do the trick.

Doesn't LG know that anyone can design a state-of-the-art mobile apps processor?

 

[geoxile]

Maybe they should stop putting 8-core processors in 7mm thin phones and pretending like the processor is the problem.

 

[Arachnotronic]

This would be the ideal moment for Samsung, LG, etc. to switch to Moorefield, since there won't be any better alternative.

Moorefield's ISP only supports a 13MP rear camera. Literally could not power a next gen flagship Galaxy S phone even if Samsung wanted to use it.

 

[ShintaiDK]

Maybe they should stop putting 8-core processors in 7mm thin phones and pretending like the processor is the problem.

Its purely PR, mixed with the artificial limitation of OS upgrades.

Its hard to sell a new phone just because its....same?

The "moar cores" however is so desperate. Apple is still laughing.

 

[Nothingness]

This would be the ideal moment for Samsung, LG, etc. to switch to Moorefield, since there won't be any better alternative.

Yeah it's not like a Mediatek chip in a phone already beats Moorefiled in a tablet on Geekbench. Oh and Exynos does the same to Moorefield. You perhaps meant something else than Moorefield?

Now the companies designing CPU and SoC are just waiting for Intel to take over the world :biggrin:

 

[Arachnotronic]

Yeah it's not like a Mediatek chip in a phone already beats Moorefiled in a tablet on Geekbench. Oh and Exynos does the same to Moorefield. You perhaps meant something else than Moorefield?

Now the companies designing CPU and SoC are just waiting for Intel to take over the world :biggrin:

Moorefield smokes even the A57 in the Note 4 in 3DMark Physics, for what that's worth. That said, I do agree that Silvermont is getting old...

I wish the folks at AnandTech would run SPEC2k6 on these ARM cores as well as Silvermont. Again, not a perfect benchmark, but would give us another good low-level benchmark datapoint.

Heck, if a bunch of folks on this forum want to chip in for a copy of Spec2k6, we could potentially get the ball rolling on benchmarking these mobile processors...

 

[Sweepr]

Problems in Qualcomm Snapdragon Set Alarm Bells Ringing for Samsung, LG

ARM's A15 successor seems to have power consumption problems, 'despite' the 20nm process.


This would be the ideal moment for Samsung, LG, etc. to switch to Moorefield, since there won't be any better alternative.

I think that's related to Qualcomm's implementation, not the ARM cores. My Exynos 7 Octa (20nm A57/A53) device delivers 5-6 hours on-screen time battery life with mixed use including 3G/4G web browsing, gaming, video playback, etc. That's on par if not better than most flagships. The device barely gets warm, definitely cooler than my old S801-based LG G3 and according to some users the A57/A53 variant is also cooler than the Snapdragon 805 units. I wonder if Samsung 20nm process is better than TSCM's 20nm process.

 

[liahos1]

Yeah it's not like a Mediatek chip in a phone already beats Moorefiled in a tablet on Geekbench. Oh and Exynos does the same to Moorefield. You perhaps meant something else than Moorefield?

Now the companies designing CPU and SoC are just waiting for Intel to take over the world :biggrin:

yes moorefield is a little long in tooth. its a shame it seems like the merchant soc smartphone roadmap has gotten pushed back for next year.

 

[Abwx]

Maybe they should stop putting 8-core processors in 7mm thin phones and pretending like the processor is the problem.

Make me think that a local north african brand is producing a 1.7 8C phone that is 5.5mm thin...

Who said that iphones are the thinest.??.

I might pick one for 280$ when i return there..

http://www.condor.dz/fr/component/condor/?view=produit&p=292&template=condorinformatique

 

[krumme]

Problems in Qualcomm Snapdragon Set Alarm Bells Ringing for Samsung, LG

ARM's A15 successor seems to have power consumption problems, 'despite' the 20nm process.


This would be the ideal moment for Samsung, LG, etc. to switch to Moorefield, since there won't be any better alternative.

Its about as possible as Intel using ss soc for their next mobile give product for free and loads of cash at the same time attack.

Translated; Even if Intel product was compettitive - witch it isnt for a long stretch - it wouldnt stand a chance.

Look at contrarevenue for non samsung and apple oem. 4B a year. Its way off. How many B to get to apple or ss. 20B a year? Naa probably way more. Ss and apple want 100% control of their extremely profitable market. They dont earn anything worth compared to mobile phones market. And they dont want to share a dime with anyone.

 

[liahos1]

Its about as possible as Intel using ss soc for their next mobile give product for free and loads of cash at the same time attack.

Translated; Even if Intel product was compettitive - witch it isnt for a long stretch - it wouldnt stand a chance.

Look at contrarevenue for non samsung and apple oem. 4B a year. Its way off. How many B to get to apple or ss. 20B a year? Naa probably way more. Ss and apple want 100% control of their extremely profitable market. They dont earn anything worth compared to mobile phones market. And they dont want to share a dime with anyone.

i dont think contrarevenue applied to moorefield/merryfield did it?

 

[Fjodor2001]

Its purely PR, mixed with the artificial limitation of OS upgrades.

Its hard to sell a new phone just because its....same?

The "moar cores" however is so desperate. Apple is still laughing.

What makes you say that?

ARM, Qualcomm and Samsung are all going in the direction of big.LITTLE and 4+4/4+2/4+1 CPU core configurations. They are the biggest players in the mobile phone industry. Obviously they consider this the best option given the alternatives.

Apple may be laughing all they want, but they are loosing share in the smart phone market. They are down to ~10% now.

 

[liahos1]

What makes you say that?

ARM, Qualcomm and Samsung are all going in the direction of big.LITTLE and 4+4/4+2/4+1 CPU core configurations. They are the biggest players in the mobile phone industry. Obviously they consider this the best option given the alternatives.

Apple may be laughing all they want, but they are loosing share in the smart phone market. They are down to ~10% now.

i think qcom is going back to a non big little approach once they get a custom v8 implementation out.

 

[Thala]

Moorefield smokes even the A57 in the Note 4 in 3DMark Physics, for what that's worth. That said, I do agree that Silvermont is getting old...

I have yet to see Moorefield reach 19k+ in 3d mark physics in a phone configuration. I have yet to see Moorefield reaching 4300+ MC score in Geekbench. Moorefield is not even in the same league as A57.
It puzzles me, what expectations people have from the old Silvermont architecture.

 

[Arachnotronic]

I have yet to see Moorefield reach 19k+ in 3d mark physics in a phone configuration.Moorefield is not even in the same league as A57.
It puzzles me, what expectations people have from the old silvermont architecture.

I was merely making an observation. I have seen benchmarks run on Intel's Moorefield FFRD, and it scores ~21-22k in 3DMark Physics.

Whether this is representative of real-world performance, or whether it's applicable across a broad selection of workloads, I'm not making a claim either way based on that one datapoint. That said, I suspect that you are correct that A57 scales to higher performance levels than Silvermont, though, across a broad set of workloads.

 

[Thala]

I have seen benchmarks run on Intel's Moorefield FFRD

Ok, but please compare likes with likes i.e. production devices and if possible the same form factor. Otherwise i agree, a few more low level benchmarks would be desirable.
Looking forward to the first batch of 64bit Benchmarks of A57 - in the end the architecture is optimized for 64 bit execution.

 

[geoxile]

i think qcom is going back to a non big little approach once they get a custom v8 implementation out.

I don't see anything wrong with big.LITTLE. But what is the point of have 4+4?

 

[oobydoobydoo]

Balderdash! That 1.5 GHz big.LITTLE Cortex A15 NUCLUN they've got will do the trick.

Doesn't LG know that anyone can design a state-of-the-art mobile apps processor?

Can't LG just pay Samsung or TSMC to produce a stock ARM Cortex/Mali SoC? I would think that Qualcomm and LG both would have secondary plans to fall back on. Obviously it wouldn't be ideal and probably wouldn't compete with Samsung Exynos but it would allow them to deliver a product on time. This article here describes the snapdragon roadmap:

http://www.dailytech.com/Leaked+Qua...core+Smartphone+SoCs+Cometh/article36417.html

Snapdragon 810 is a big step! Going from quad core 32 bit to octacore big.LITTLE may have been too big a step to do in one revision.

 

[Arachnotronic]

Can't LG just pay Samsung or TSMC to produce a stock ARM Cortex/Mali SoC? I would think that Qualcomm and LG both would have secondary plans to fall back on. Obviously it wouldn't be ideal and probably wouldn't compete with Samsung Exynos but it would allow them to deliver a product on time. This article here describes the snapdragon roadmap:

http://www.dailytech.com/Leaked+Qua...core+Smartphone+SoCs+Cometh/article36417.html

Snapdragon 810 is a big step! Going from quad core 32 bit to octacore big.LITTLE may have been too big a step to do in one revision.

oobydoobydoo

It's not that easy. Even for a "stock" A15/A57, the quality of the physical implementation of the core will vary depending on the design team. A company like Qualcomm or Samsung will be able to implement the "stock" core in a much better way (higher frequencies, lower power) than a company with minimal chip design experience like LG.

That's why the LG NUCLUN clocks at just 1.5GHz for the A15; other companies have done better.

And who is to say that the problem with S810 is the CPU? That's a ~2B+ transistor SoC we're talking about and the interactions among the IP blocks within such a chip, and the fact that you have a bunch of different IPs all running at different frequencies, adds a LOT of complexity.

The CPU core is just a small part of a mobile SoC, which is why I find it perplexing that everyone cares so much about whether QCOM or whomever does a custom core or uses an ARM core...there's so much more than that to an SoC...

 

[Exophase]

Heck, if a bunch of folks on this forum want to chip in for a copy of Spec2k6, we could potentially get the ball rolling on benchmarking these mobile processors...

I'll chip in if you agree to do the actual work configuring the hardware, setting up the benchmark and writing the article

 

[Idontcare]

oobydoobydoo

It's not that easy. Even for a "stock" A15/A57, the quality of the physical implementation of the core will vary depending on the design team. A company like Qualcomm or Samsung will be able to implement the "stock" core in a much better way (higher frequencies, lower power) than a company with minimal chip design experience like LG.

That's why the LG NUCLUN clocks at just 1.5GHz for the A15; other companies have done better.

And who is to say that the problem with S810 is the CPU? That's a ~2B+ transistor SoC we're talking about and the interactions among the IP blocks within such a chip, and the fact that you have a bunch of different IPs all running at different frequencies, adds a LOT of complexity.

The CPU core is just a small part of a mobile SoC, which is why I find it perplexing that everyone cares so much about whether QCOM or whomever does a custom core or uses an ARM core...there's so much more than that to an SoC...

That is a very astute point which I hope others appreciate.

With so much integrated IP on a single SoC, the likelihood of this having anything to do with the tiny portion of die area that is ARM related (be it big.LITTLE or something within any given LITTLE or big core) is really quite small.

If the problem is power, or clockspeed, then the finger is definitely going to point to (1) process node or (2) layout (including critical path and/or borked power saving functionality).

And since this is made on the same 20nm as Apple's chip, which doesn't have any reportable yield or power issues, the finger would tend to lay pointing at Qualcomm's layout team as having muxed something up with a critical path or two, or a key power-saving feature has been disabled pending a mask respin.

 

[CHADBOGA]

I'll bet Intel are kicking themselves that they don't have a suitable product that would allow them to swoop in here and take advantage.

 

[Abwx]

oobydoobydoo
That's why the LG NUCLUN clocks at just 1.5GHz for the A15; other companies have done better.

It could be a deliberate choice to improve battery life, ARM supply the core layout and physical implementation in a given process, LG surely did nothing else than connecting the relevant bus to the cores.