If you’ve been paying attention to the right places in the past few months it was probably obvious this was coming, but Qualcomm is announcing a higher tier to their Snapdragon 82x lineup, known as the Snapdragon 821 or MSM8996 Pro. While today’s announcement basically boils down to acknowledging that this SoC exists and that the big CPU cores have a clock speed of 2.4 GHz, it’s likely that in the months since the Snapdragon 820 was released Qualcomm engineering staff have been working on resolving various errata as well as improving their floorplanning and architecture implementation. It’s also likely that we will see a few new or otherwise revised IP blocks.

  Snapdragon 820 Snapdragon 821
CPU Perf Cluster 2x Kryo 2.2 GHz 2x Kryo 2.4 GHz
CPU Power Cluster 2x Kryo 1.6 GHz 2x Kryo >2 GHz
GPU Adreno 530 624 MHz Adreno ??? ~650 MHz

What isn’t in this announcement is that the power cluster will likely be above 2 GHz and GPU clocks look to be around 650 MHz but without knowing whether there are some changes other than clock relative to Adreno 530 we can’t really estimate the performance of this part. However, this information can be subject to change depending upon what happens at Qualcomm. It's important to note here that while these changes may seem to be small that improvements in the implementation of an SoC can have a dramatic effect on performance and power. I’m sure we’ll be learning more about this SoC in the coming months so for now we’ll just have to wait and see what comes next.

POST A COMMENT

78 Comments

View All Comments

  • easp - Monday, July 11, 2016 - link

    What amazing luck! Can we use phones as PCs to get this magical scaling benefit? Reply
  • hapeid - Friday, September 30, 2016 - link

    Enjoy while it last because the rumour said that Kyro would be the last custom core from Qualcomm. If ARM were to release a new arch to replace v8-a or the newer cortex were to much better than Kyro, then we're back to many cores madness.

    http://hapeid.com
    http://iteote.com
    Reply
  • CloudWiz - Monday, July 11, 2016 - link

    So you're telling me that the 950 is faster in single-core and web browsing performance than the A9? I understand that Twister is much larger than A72, but there's no catching it in terms of IPC. When the A10 drops with improved IPC and higher clocks, there's no way A73 should be able to beat it. Reply
  • Geranium - Tuesday, July 12, 2016 - link

    Apple SoC do good in running Apple optimized benchmark. Reply
  • osxandwindows - Monday, July 11, 2016 - link

    Lol, ok.
    How is it that a dual core cpu (apple a9) equals or exceeds 4 and 8 core ARM cpus and almost everything.
    Reply
  • retrospooty - Monday, July 11, 2016 - link

    Surely you get that Apple has their own closed architecture that they totally control. They control everything from the hardware, to the drivers, to the API's and even the benchmarking apps. Open architecture is entirely different especially when benchmarking. Doing it Apple's way is "a way" to go, but it benefits Apple and no-one else. Apples stuff is fast, good CPU's and they "benchmark" extremely fast for sure (in the closed architecture)... But this is a discussion about the tech involved and the advantages of it not an "Apple is best because I say so" argument. Its closed and therefore cant be benchmarked in the same manor. All of todays high end chips, the A9, SD820, Exynos 8890, Kirin 950, even Mediatek x20 are all insanely fast (Girls you are ALL pretty, now stop arguing). Reply
  • easp - Monday, July 11, 2016 - link

    Retrospooty, you are a fool. Apple's approach benefits anyone who buys an iOS device. The excessively multi-core solution benefits almost no one who buys an android phone, except for people who like running multithreaded benchmarks.

    Apple's way of using a few beefy cores isn't better just because they say so, or even because they have tight control over both software and hardware. Some things just don't parallelize well, and many of those that do are better run on the GPU. That's not to say that having lots of cores on a relatively inexpensive SoC isn't useful for some people, it is, though generally not in a phone.

    Most of the android market is SoCs that integrate licensed ARM IP for CPU cores, not original designs. ARMs designs have tended not to push the edge of performance, because doing so would undermine their architecture licensees. The other reason is that one of their architecture licensees, Apple, invested much more aggressively than the others (particularly Qualcomm) and caught the others off guard in-terms of both performance, and the transition to 64-bit.

    Now, though, Qualcomm is back in the game, Samsung has joined it, and it seems that ARM might be a bit more aggressive than they have in the past. Why are more companies developing big cores for ARM? It's not because of doing it Apple's way is "a way" to go, its because it is THE way to go, because, as explained, a lot of things don't parallelize well, and many of those that do are better run on the GPU.
    Reply
  • retrospooty - Tuesday, July 12, 2016 - link

    You are missing what I am saying. The discussion is on CPU's and benchmarking cross platform isnt really accurate. The speed benefits of tight integration are more from the tight integration, not from the CPU itself. Its a good CPU, made to run much better based on the tight integration and control. The end result is very fast, but as I mentioned they are all fast in real world use.There are drawbacks to it as well. The extreme lack of options being the biggest. Reply
  • tuxRoller - Tuesday, July 12, 2016 - link

    In what way are the 2-3 wide cores "big"? Reply
  • Geranium - Thursday, July 14, 2016 - link

    Apple has no nown architechure, they just refine ARM's work. Reply

Log in

Don't have an account? Sign up now