While this launch is Korea-only, Samsung recently announced a new version of their Galaxy S5 smartphone, dubbed the Galaxy S5 Broadband LTE-A. Naming aside, this makes this phone the first to launch with APQ8084 and MDM9x35. For those unfamiliar with Snapdragon 805 and MDM9x35, this means that the CPUs are now Krait 450 instead of Krait 400, and the GPU is now Adreno 420 instead of Adreno 330. While the CPU revisions are minor, the GPU is fast enough to have the same level of performance at 1440p as an Adreno 330 at 1080p. The MDM9x35 modem also means that category 6 LTE is supported for speeds of up to 300 Mbps. The MDM9x35 is also the first 20nm SOC part shipping from TSMC, which bodes well for 20nm SoCs in the near future.

Qualcomm also notes that this phone integrates the WTR3925 transceiver, so carrier aggregation is done on a single chip instead of the WTR1625L/WFR1620 dual-chip solution that was previously needed. Samsung also integrated a QHD (2560x1440) OLED display into this model at the same 5.1" display size. The only other difference is that the phone now has 3GB RAM instead of the 2GB present in the international model. Otherwise, the rest of the phone is identical to the international Galaxy S5. It's curious to note that Samsung has chosen to use the 2.5 GHz bin of the APQ8084 line rather than the highest 2.7 GHz bin, although the reasons behind this decision aren't quite clear yet.

Source: Samsung Tomorrow

Comments Locked

40 Comments

View All Comments

  • Gondalf - Friday, June 20, 2014 - link

    But yes, it is always possible to have better computational strongness at a given clock speed (more IPC or throughput), still this have a cost in power consumption because the units (alus, fp blocks etc.) are more active, this draws power even if the units are more advanced (better caches, internal buses, minor bottlenecks). It's a matter of transistor switch, the more are the switching transistors at a given process, higher is the power consumption. Obviously there is the better power management to drop the passive power (leakage) but.....in these days nearly all was done to drop the power in mobile SOCs, we are at the minor incremental enhancement era.
    The real story is that all the dust is hidden by inexistent TDP figures from Qualcomm and by an aggressive throttling after two or three minutes of full load, just in time to make hot the body of your loved high end phone (not much high end after all).
  • UpSpin - Friday, June 20, 2014 - link

    I doubt you know what you're talkig about.
    In benchmarks the units are as active as possible, how on earth can they become more active by a new revision?
    The more the transistors switch, the higher the clock speed. But we're talking about the performance at a fixed identical clock speed. So the 'more switching' is nonsense.
    Power management has little to do with leakage current. Leakage current is a matter of the process. The smaller the process, the higher the leakage current. To handle this issue you can power gate specific parts. Power management is more about how responsive, fast and efficient the GPU can transition between different states, so it can ramp up only when needed and stay there as short as needed.
  • R0H1T - Friday, June 20, 2014 - link

    "It is strange that Anandtech do not stress the top tier phones with a sustained GPU bench to reveal all the weakness of these absurd tablet thinked SOCs" ~ Just take a wild guess which SoC's will throttle the most, hint it starts with an I & ends with L !

    P>S> sorry for yet another rant (-:
  • Homeles - Friday, June 20, 2014 - link

    Do you have any facts to back up your claim? Or are you just being an ignorant shill for AMD, as usual?
  • R0H1T - Friday, June 20, 2014 - link

    Funny you say that when in fact you also come to the defense of Intel seemingly anywhere & everywhere on this planet, guess what that makes you OR is it the usual double standards from your side of the fence ?
  • Homeles - Friday, June 20, 2014 - link

    I actually don't. I'll defend anyone, but your head is too far up your own rear to realize this.

    Since you are far too uneducated about hardware in general, the designs that would be most subject to thermal throttling are the ones that produce the most heat.

    AMD would easily be the worst, seeing as they can't even scale their designs down to smartphone levels, and are barely even tablet worthy. They'd be followed by Nvidia and Apple, in no particular order, that have higher power targets than their competition.
  • R0H1T - Friday, June 20, 2014 - link

    The fact that you brought up AMD in a topic involving Adreno & Intel is no surprise to me but there is no AMD SoC (yet in tablets) & so your extrapolation of "fun facts" also tells me that your arse is not in the right place. FYI before Maxwell AMD had the most efficient GPU in HD 7790, the thing AMD's doing right now is slowly but surely cutting down their APU power consumption with a top down strategy heavily reliant on node shrinks & GPU optimization unlike the other major vendors who've developed SoC's from the bottom up. The new ARM based AMD cores will show where does AMD in the league of established players but hey carry on ranting wherever I mention the "I" word in any of my posts, but since I already know where you & most of the Intel apologists on this site come from I do take pleasure in winding up the so called "neutral" fans !
  • 787b - Friday, June 20, 2014 - link

    I second that. Almost every contemporary android flaghip uses a version of snapdragon 800, but there are packaging differences between different models, so the thermals are bound to be different. Sony, for example, uses a thermal paste between SOC and a metal chassis in Z1 Compact, but I have no ideal if it works in real world.

    I propose a freezer test, similar to the nexus 4 throttling investigation, just with more phones. Also a review of a Sony smartphone would be really nice.
  • Morawka - Saturday, June 21, 2014 - link

    problem is, this soc IS ON a smaller fabrication process (20nm) than all the others you mention. Your argument is entirely invalidated.
  • ruzveh - Friday, June 20, 2014 - link

    It would be interesting to see Full HD 1080p display performance over 20nm SOC. It would either bring more power or drastic reduction in power.. Frankly who needs QHD on phones?

Log in

Don't have an account? Sign up now