While Qualcomm already announced the Snapdragon 821, with the announcement details were rather sparse. Fortunately, today Qualcomm followed up with more details. Those that followed the announcement might recall that the only information disclosed at the time was that the CPU big cluster was now at 2.4 GHz. Today, we also get a GPU clock disclosure, with full details as seen below.

  Snapdragon 820 Snapdragon 821
CPU Perf Cluster 2x Kryo 2.15 GHz 2x Kryo 2.34 GHz
CPU Power Cluster 2x Kryo 1.59 GHz 2x Kryo 2.19 GHz
GPU Adreno 530 624 MHz Adreno 530 653 MHz

Interestingly enough, Qualcomm is also claiming a 5% bump in power efficiency, which sounds like it’s actually referring to platform power but could just be overall SoC efficiency. Other marketing bullet points include support for Snapdragon VR SDK which allows for Daydream support as well as dual phase detection. I’m not sure what this is unless this is referring to support for two separate phase detect auto focus systems similar to the Sony Alpha SLT-A99, but Qualcomm is claiming that this will improve autofocus speed compared to a traditional PDAF solution. The ISP also now supports extended ranges for laser AF, so systems like those seen in the LG G5 and HTC 10 will be able to better guide contrast AF for devices where PDAF isn’t available or can’t be used.

Qualcomm is also citing some interesting statistics for user experience with the Snapdragon 821, such as 10% faster boot speed, 10% faster app loads, and some BSP changes combined with faster processing to enable smoother scrolling and improved web browsing performance. The Snapdragon 821 SoC is already shipping in devices like the ASUS ZenFone 3, so we shouldn’t be far off from seeing major launches using this SoC. It's interesting to note here that last year we got details of Snapdragon 820 by September but we have yet to see what Qualcomm plans to launch for next year's flagships. It'll be interesting to see whether they stay with a custom CPU core or elect to go with an ARM Cortex big.LITTLE configuration similar to the Kirin 950.

Comments Locked

32 Comments

View All Comments

  • MrSpadge - Wednesday, August 31, 2016 - link

    Valid points!
  • ImSpartacus - Wednesday, August 31, 2016 - link

    Isn't power increase from clock increase linear? I thought it was voltage increases that were non-linear. Obviously, you often need higher voltages to get to those higher clocks, but it's sometimes possible to just increase clocks.
  • MrSpadge - Wednesday, August 31, 2016 - link

    I think he was implying a voltage increase. Apart from that there's a minor non-linearity from the higher temperature which the higher power draw at higher clock speeds causes, but this can easily be neglected when talking about a few percent clock speed differences.
  • Slap_and_Tickle - Wednesday, August 31, 2016 - link

    The power increase with clocks is almost never linear at the top speeds because a voltage increase is required.
  • TylerGrunter - Wednesday, August 31, 2016 - link

    Is the link supposed to be of the new Kirin? Because the results are the same as the old Kirin 950. So any hope for 2x their current GPU performance is lost.
    https://gfxbench.com/device.jsp?benchmark=gfx40&am...
  • jjj - Wednesday, August 31, 2016 - link

    Car chase 11.7 vs 6.3 FPS
    Manhattan 3.1 20.1 vs 10.4
  • jjj - Wednesday, August 31, 2016 - link

    We'll see if it's MP8 at moderate clocks or MP6. MP8 at moderate clocks would be great and hopefully avoid heavy throttling.
    It can't be on 10nm due to timing so MP8 would be just right and A73 at 2.6GHz on 16ff would fit what we know about A73.
  • name99 - Wednesday, August 31, 2016 - link

    Intel seems to have squeezed around 12% performance improvement from their 14nm optimization at more or less equal power. One suspects the same holds for TSMC's improvement from 16nmFF to 16nmFF+. Obviously none of us know the details, but the Intel number likely results from not just frequency improvements but also some compensating power management improvements. QC has done much less to tweak their SoC apart from just "re-compiling" it for the new process, so we'd expect less benefit this round.

    More interesting is what Apple will manage to do with the same process improvement. The same order of frequency boosting would get us to something like A10@2.0GHz and A10X@2.5GHz. If they throw in 15% or so micro-architecture improvements, then we get an overall boost of around 25%, which is very much analogous to the A7->A8 transition. And presumably Apple have enough manpower and time that they will also boost their power management somewhat, giving us those improvements with no real drop in efficiency and, I would assume, like the A9 the ability to maintain that performance (mostly) even under extreme long-duration conditions. (At some point I expect Apple will run out of other knobs to tweak, and will move to the Intel "ramp up performance into fireball mode for a few seconds to maintain short bursts of interactivity" but they seem unwilling to do that yet.)
  • name99 - Wednesday, August 31, 2016 - link

    semiaccurate suggests
    http://semiaccurate.com/2016/08/31/qualcomm-outs-b...
    what I assumed as well --- the process improvements give BOTH 9% speed boost AND 5% reduced power, not either/or.
  • lilmoe - Wednesday, August 31, 2016 - link

    Looks like the little cores are getting the larger bump in clocks. Most modern OEM optimizations for Android revolve around routing as much UI and scrolling workloads to the little cores to save power. Kind of explains their claims of better UI and smoother scrolling.

    Big cores are already fast enough, what we now need is even better performance and efficiency in the little cores. I'm looking at you Sammy; either ramp up the small cores to full speed when my finger touches the screen, or schedule the entire foreground application to a big core or two at lower clocks (keeping the little cores for background processes, don't care if those slow down).

Log in

Don't have an account? Sign up now