Comments Locked

23 Comments

Back to Article

  • hirschma - Wednesday, November 30, 2011 - link

    From the article:

    "Today, in Korea, they announced their first Cortex-A15 SoC, the Exynos 5250. Clocked at 2 GHz and produced with a 32 nm process, the chip should be competitive with the Qualcomm and NVIDIA designs that will be rolling out soon, though quad-core parts will be the first products they'll be shipping."

    Besides being a horrendous run-on sentence, it is unclear who "they" is, as in "they'll be shipping". So, is Samsung going to be shipping quad core (despite the article title)? Is Qualcomm? I already know about Nvidia.

    Can someone rewrite this for clarity?
  • Glibous - Wednesday, November 30, 2011 - link

    I think they'll is Nvidia as their Tegra 3 is a quad-core part. But when comparing the architecture between the Nvidia, Qualcomm, and Samsung Exynos. Their competitive. By this i'm assuming that Nvidia's processor will be superior due to the number of cores but the Exynos will have a higher clock so light threaded applications will see a benifit from the Samsung processor where heavily threaded tasks will see a benifit from Nvidia's processor.
  • dagamer34 - Wednesday, November 30, 2011 - link

    An increased number of cores does not indicate noticeable superiority, especially when taking about a CPU at architecture advance. Just take aloof at the latest Sandy Bridge E review to see that even with 6 cores, there are only a few tasks when the performance difference is noticeable, and even fewer that justify the cost difference. In a mobile environment where there's usually only one main task running thats not coded to be multi-threaded because it doesn't really need to be, I think a better dual core system will outgun a quad core one.

    Let's face it, quad core is nice, but outside of games, what's is really going to take advantage of it? I certainly hope you won't be doing video rendering on your device as your battery life will be horrible.
  • Glibous - Wednesday, November 30, 2011 - link

    Did you read my entire post?

    "Exynos will have a higher clock so light threaded applications will see a benifit from the Samsung processor where heavily threaded tasks will see a benifit from Nvidia's processor."

    I'm aware there's no advantage with a quad-core running an application that can't run on all cores. Assuming both Nvidia's and Samsung's processors are similar in performance clock for clock then Samsung's higher clock should have a general advantage.
  • Exodite - Wednesday, November 30, 2011 - link

    They won't though, at least as long as you're talking about the Tegra 3.

    Cortex A15-based designs will have significantly higher performance per clock compared to Cortex A9-based designs like the Tegra 3.

    As noted by Samsung themselves in the press release, they expect a 2GHz A15 design to be /twice/ as fast as a 1.5GHz A9 design.

    This implies a 50% performance advantage, clock per clock, over Cortex A9.

    No doubt this chip will be a screamer.
  • Glibous - Wednesday, November 30, 2011 - link

    If that's the case then Samsung's processor will be very impressive. I don't know too much about the performance differences between A9 and A15. I was just going by the article stating that Samsung's processor should be competitive with Qualcomm and Nvidia's processor. but if thats the case then yea definately changes things. I still don't know how heavily threaded applications would fare against Samsung's or Nvidia's processors (Architeture advantage vs. # of cores, In-order vs. Out-of-order execution). I'm going to do a bit of research on this for myself.
  • JasonInofuentes - Wednesday, November 30, 2011 - link

    So, to ensure that everyone is on the same page let's talk about the products we're looking at in a little more detail. Two manufacturers have committed to releasing A15 based SoCs in 2012, TI and Samsung with OMAP 5 and Exynos 5xxx respectively. Qualcomm will be releasing their latest iteration of a bespoke ARM core in the form of Krait (S4). NVIDIA will be trying to compete the same way they did with Tegra 2, by being first. All four will have quad core parts shipping at some point. Tegra 3 will ship as a quad-core first, and soon. Krait will ship in dual and quad core configurations, and it's unclear which will ship first, but that won't be for a few months. TI will probably ship dual-core first, and that won't be till later in 2012. Samsung is today announcing that their FIRST shipping Cortex A15 based product will be a dual-core Exynos.

    I'm sorry for any confusion my earlier phrasing caused, it's often harder than it looks to paint a clear picture of such muddled and overlapping road maps. I will try harder to ensure further clarity in the future. The ultimate point remains, this will be a potent chip, and we want to find out more about it, as soon as we can.
  • joebrooks - Wednesday, November 30, 2011 - link

    How is that a run-on sentence?
  • JarredWalton - Wednesday, November 30, 2011 - link

    When in doubt, throw out terms that suggest poor grammar without knowing what they mean! Though I do think the last part is a bit confusing and thus I'll edit it (e.g. "though quad-core parts" to "quad-core parts").
  • JasonInofuentes - Wednesday, November 30, 2011 - link

    Thanks Jarred. Apologies for any syntax or grammar confusion. Hope everything is clearer now.
  • biassj - Wednesday, November 30, 2011 - link

    So Galaxy S III =1080 screen and plays Crysis?
  • SimpleLance - Wednesday, November 30, 2011 - link

    A15: 3-issue, 15-stage pipeline
    A9: 2-issue, 8-stage pipeline

    ARM estimates that the A15 is 40% faster than the A9 for the same clock speed.
  • tyger11 - Wednesday, November 30, 2011 - link

    I'd love to see what kind of power draw they could get down to by making this at 22nm. Sure, Ivy Bridge will be great and all, but the chips that really need to be built on such a small process are mobile chips.
  • JasonInofuentes - Wednesday, November 30, 2011 - link

    That's an interesting thought, but imagine what kind of a power hungry, hot mess Sandy Bridge was if it was twice as big as it is now. The cutting edge will for some time be PC CPUs, because their higher IPC is a result of higher transistor counts. Plus, PCs remain where most computing happens. At some point, as more and more computing is done in the mobile space, then node parity between ARM and x86 might happen. If it does, though, Intel and AMD had better have designs that transcend their current performance envelopes, because their performance per watt is still nowhere near ARM designs.
  • etamin - Wednesday, November 30, 2011 - link

    Does anyone have a guess as to how powerful these SoCs are vs older desktop chips like Pentium III or Pentium 4 maybe?
  • SimpleLance - Wednesday, November 30, 2011 - link

    When I tested the same single-threaded chess code on an HTC HD2 (1Ghz A8 ARM) and on a 2.3Ghz, Core2-duo, the C2D was 3 times faster.

    This means that the newer A15 will be faster than a Core2, for the same clock speed.
  • Exodite - Wednesday, November 30, 2011 - link

    Kinda depends a lot on what instructions you use though, doesn't it?

    The x86 instruction space, all extensions included, is vast and while that's often a disadvantage I'd imagine it would widen the gap considerably if taken advantage of properly.

    Personally I find it difficult to believe that even the Cortex A15 could hope to approach the same IPC as C2D.
  • tipoo - Thursday, December 1, 2011 - link

    I was just about to say that, I'll bet the code didn't use the x86 extensions available in the Core 2 Duo otherwise it would be much faster.
  • psychobriggsy - Wednesday, November 30, 2011 - link

    With over 1000 balls on that BGA, we're talking about a significantly higher-end chip than your average mobile processor! I guess that it has off-chip DRAM (via SO-DIMMs) rather than stacked on die, and a lot of those balls are memory interface connections.

    The graphics sound like they'll be twice as powerful as the ones on the A5, certainly more than adequate for 2012's tablets and ARM netbooks / ultrabooks. And possibly Android-based media hubs that can also play games (because Apple TV 3 will surely do that soon).

    Wonder what the power consumption is...
  • rd_nest - Thursday, December 1, 2011 - link

    How many SoC would Sammy make?

    I understand that 4412 would be Quad Core. AFAIK, all of them use 32nm fab. But what's the difference (benchmark/practical usage) between all these SoC?

    Off-topic: Do Samsung allow it's employees to take vacation or week-offs?
  • Hunt3rj2 - Thursday, December 1, 2011 - link

    Qualcomm is in trouble if their Snapdragon S4 can't stay competitive with this.
  • ven - Thursday, December 1, 2011 - link

    Though this not the place to ask I can't able to stop it, so please bear it

    ARM provides architectural design details and fab company's like samsung or TSMC will implement that in the silicon for their customers like (Apple),then what role does apple do in their A5 or A4 chip?. . Just printing their Apple White Logo on top of the chip package?..

    i ask this because

    samsung= they make their chips.

    qualcomm= they borrow only the ARM instruction but they design the core themselves

    Nvidia= obviously they have their graphics engine

    Apple=?

    As you can see except Apple all other SOC manufactures do something on their part. so, how does apple claim itself as one of the SOC manufacture without doing anything?..

    Someone clarify my doubt.
  • juicytuna - Thursday, December 1, 2011 - link

    Apple bought Intrinsity, a company who design SOCs using off the shelf IP (PowerVR GPU, ARM CPU), just like Samsung or TI. They lack the manufacturing capability of those two however, and rely on third-party fabs to make their SOCs.

Log in

Don't have an account? Sign up now