LightBlog

mercredi 6 mai 2015

On Qualcomm’s Damage Control: Marketing and Rumors?

qualcomm1

The Heated Snapdragon chronicles seemingly see no end, and 5 months into 2015 we are still discussing the Snapdragon 810. This is unfortunate, because it is rather clear that the market is moving away from it, and it’d be better to simply forget about this black stain once and for all. However, many websites can’t let it go – and in some ways, we are one of them – but only in response to claims that we think affect consumers or users.

 

XDA is a community of power users. As such, it follows that we want to empower consumers to give them more control over their purchases. A large part of this is the software our thousands of developers work on, but there’s also a more empowering aspect: knowledge. The XDA Mantra has always been about this – XDA is about learning as much as it is about flashing fancytweaks. Because of this, whenever we see false or misleading claims, we are quick to put them into proper context so that our users have greater control through informed decisions – that is the real virtue of a power user.

So under this light, I want to bring up a recent article where we discussed the shortcomings of the Snapdragon 810 that other publications chose to sweep under the rug or claim as non-issues. It began like this: the Snapdragon 810 saw early rumors that said the chipset suffered from overheating. The rumors escalated and evolved into gossip regarding Samsung not using a Qualcomm chip in their Galaxy S6 for this very reason. The Snapdragon 810 in the G Flex 2 did not heat up dramatically, but it certainly wouldn’t – it throttled before it could. The same was true for the M9, and early reports on unfinished software builds showed overheating that was then fixed by installing a severe performance cap in the CPU’s frequency. To top it all off, the S6 did, indeed, not feature a Qualcomm chipset.

Now that we have those pieces laid out, we want to bring up this interview that Forbes had with Qualcomm’s VP of marketing Tim McDonough. In it, we see the Snapdragon 810 overheating brought up yet again, with more official statements denying such a thing, and more. Let’s take a look:

“The rumours are rubbish, there was not an overheating problem with the Snapdragon 810 in commercial devices”

There were no overheating problems on commercial devices, that is true. The rumors, however, were about the processor (as a singular entity) seeing these problems. When you put throttles in place to prevent the design flaw from making the device overheat, it does not eliminate the design flaw. With the M9’s frequency capped to 1.6GHz for regular use, the device doesn’t overheat. It is also worth noting that performance does take a hit from this. Moreover, throttling on the G Flex 2 is awful in comparison to every competitor, meaning performance is gimped almost immediately.

“If that’s true, which we’re saying it is, why was there so much rumour? Why was someone spreading false information about the 810? Our point of view is that those rumours happened with the LG G Flex 2 and Qualcomm 810 being first to market with the premium-tier application processor. Then somebody decided to put out some false rumours about that, which is unfortunate but sometimes that’s how business is done. That has forced us to spend a lot of time addressing the false rumours.”

It is not rare for companies to start rumors about other companies to hurt them, and it could very well be the case that the rumors were put out specifically to damage Qualcomm. That being said, the Snapdragon 810 as seen in the G Flex 2 was not quite the “premium-tier” application processor that it is made out to be in this statement. During its release, performance complaints were addressed by claiming that the initial software was unfinished, and stutters, re-draws, force-closes and generally bad performance were common in reviews.

“Everything you’re saying is fair. But we all build pre-released products to find bugs and do performance optimisation. So when pre-released hardware doesn’t act like commercial hardware, it’s just part of the development process.”

It is a normal part of the development process… but when you have OEMs deliver “unfinished software” in demo, review and/or consumer, and then you see the consequences of that, you can not blame the development process nor those who disclose the findings. The development process of a new release should end before release, and pre-release hardware or software shouldn’t be given out to reviewers. If you do give it out, there can be no excuses. If you must push out an OTA with several performance fixes (for UI fluidity, camera, or whatever) early on, there can be no excuses. If you have to delay a release to address these issues after they surfaced in review or demo units, there can be no excuses.

This is a matter of transparency: if reviewers or testers crucify a device for having sub-par performance on unfinished software, they are not to blame. They can state and should state that it is unfinished software, that it might get fixed, but it is ultimately the OEM’s responsibility and it shouldn’t be shrugged off nor sugar coated. If, for example, reviewers or testers hold back possible faults using the “unfinished software card”, the objectivity of a review is lost. Many faults can be fixed down the way – something we’ve seen many times in Nexus devices, for example – and we know that. Realizing that some shortcomings come from present software is a must, but using it as an excuse “because future software might (or might not) fix it” should never be a practice.

“I think someone very artfully took that and used it to fuel the rumours and took something that’s completely normal and sent it to some less sophisticated news outlets to give them a story.”

Ignoring the fact that practically every major Android website reported these rumors at some point (and all jumped on the overheating tests, with all sorts of title puns), this is irrelevant when you consider that review units  saw these problems too. The very writer of the Forbes interview noted that review units had to be re-called and a software update “fixed” many performance issues. The throttling was put in place to prevent overheating, but this is something controlled by software switches, and doesn’t change the physical design flaws that the Snapdragon 810 might have. When it comes to said switches, HTC also enabled a special “high performance mode” to cheat on benchmarks by using the maximum frequency despite it not being attainable nor sustainable in real-world use, which is deplorable as you cannot lift the cap on regular use-cases. This leads to misleading specification sheets where the listed CPU frequency is not what consumers end up getting.

“Who stands to gain from the rumour?”

Obviously companies in the semi-conductor game, like Samsung. McDonough denied to specify if he believes Samsung started the rumors, but even then, the rumors were not entirely false. Variants of the M9 that didn’t have the software updates pushed to reviewers and installed in commercial devices did see overheating, and it seemingly got fixed by throttling the SoC to prevent it from running too hot at the maximum frequency.

When asked about the Snapdragon 808 in the LG G4:

“These decisions get made 18 months before a phone shows its face. When we were working with LG on the G flex 2 and G4, Qualcomm produced the 810 and 808 around the same time. The simple reason is the design goals of the G4 2k experience (…)

(…) The difference between the chips is that the 810 is scaled for end-to-end 4K and it’s our flagship. The 808 scales for end-to-end 2k and it still captures 4k video but it’s not built to display 4k natively on a screen (…)

(…) If you were to look at these two side-by-side, they have the exactly the same modem, they both use an Adreno GPU, low-power audio and they both compresses 4k video so it doesn’t kill your phone. They’re almost exactly the same. (…)”

This is, perhaps, the biggest offender here. First of all, the Snapdragon 808 and 810 are not “almost exactly the same”. They have the exact same modem, low-power audio and both compress 4K video. What about the CPU and the GPU though? Different – very different. In fact, I do not know why he said that they both use an “Adreno GPU” and used it as a point of comparison. The 805, 800, 801, 615, 610, 600, 410, 400, 210, 200, S4/Plus, S3, S2 and S1 all use Adreno GPUs. Did I just list all Qualcomm chipsets? I might have. Saying that they both have an Adreno GPU means nothing. In fact, the Adreno 430 in the Snapdragon 810 is around 50% faster than the one found in the Snapdragon 808. The CPU in the Snapdragon 810 is different in core count, listed frequency, and more. The Snapdragon 810 also allows for DDR4 RAM, while the 808 does not.

Bringing up the resolutions is also largely irrelevant. The difference between the chips is that the 810 can display 4K natively, while the 808 can’t. Neither the 810 phones nor the G4 have 4K displays, and in fact, the M9 has a 1080p display. The G4 has a 1440p screen which means the G4 should be allocated more GPU power than the M9, not less. As it stands, the G4’s GPU is 20% faster than the Adreno 330 in the Snapdragon 801 used in the G3, which had serious difficulties pushing the pixels. The Adreno 420 in the 805 was enough to mitigate the pixel gain, but the GPU in the 808 is still worse than the one in the 805 which was 40% faster than the G3’s 801. To say that processors are exactly the same when they have vastly different CPUs and GPUs is just misleading marketing, as is listing CPU frequencies that can’t be obtained in real-world use.

 

That is all we have to say. As you can see, Qualcomm’s marketing is doing very well, because a key aspect of marketing is the omission of details. The Snapdragon 810 is not a bad chipset, it can be scaled down to perform very well. We analyzed this not too long ago, and saw that it actually outperforms the Exynos 7420 in various cases. Regarding the Snapdragon 808 in the G4, we also arrived to the conclusion that performance would suffer in some areas, and the recent benchmarks and reviews show it did. However, real -world/UI performance is still reported to be solid due to optimizations that Qualcomm and LG worked on. A chipset alone can not condemn a phone, and software can mitigate many shortcomings. But performance downgrades in such a heated business are, in fact, issues – especially when marketing teams and websites go out of their way to mitigate that they are. I hope we at XDA can always go against the narrative to inform our readers, and in turn, empower our users.

The post On Qualcomm’s Damage Control: Marketing and Rumors? appeared first on xda-developers.



from xda-developers http://ift.tt/1EhLkgr
via IFTTT

from XDA http://ift.tt/1IQeREm
via IFTTT

Aucun commentaire:

Enregistrer un commentaire