Back to Article

  • watersb - Tuesday, July 23, 2013 - link

    Nice catch. Stunning that the CCI shipped. Reply
  • Sivar - Tuesday, July 23, 2013 - link

    Not at all.

    Remember Samsung Galaxy S? It shipped with an obviously broken GPS. No replacement offered.

    Remember the Galaxy X III and its rash of mainboard failures?

    Remember the class action lawsuit over Samsung LCD TV failures? Their refrigerators?

    This is par for Samsung.
  • steven75 - Tuesday, July 23, 2013 - link

    Part of the problem with Samsung's "throw everything at the wall and see what sticks" approach is that some of it *won't* stick. Reply
  • chizow - Thursday, July 25, 2013 - link

    Yes, Samsung's approach is certainly "if it's broke, we'll fix it in the next iteration", but that has led them to continuously innovate, like clockwork. You can expect an update from Samsung in a year, sometimes even less. But ultimately, if a feature is that important to you, make sure it is working when you buy the product from Samsung, as their support and backward compatibility track record is pretty poor. Reply
  • theduckofdeath - Monday, September 02, 2013 - link

    Is it better to offer a free bumper to the first few hundred thousand victims for a design failure, and after that just insist you're holding it wrong?

    It is really rare that companies calls back all devices to fix issues that can be worked around. IT would take a successful class action lawsuit, which does happen in the industry from time to time.

    I really don't see what your point is trying to single Samsung out with a behaviour that all companies have?
  • jameskatt - Friday, August 16, 2013 - link

    It is stunning that Samsung would break quality and ship an obviously buggy product without offering a replacement. Reply
  • jjj - Tuesday, July 23, 2013 - link

    With the 5410 at 122mm2, I wonder how big this one is. Reply
  • twotwotwo - Tuesday, July 23, 2013 - link

    Huh. There were bench #s posted before where the Exynos wasn't much faster than Snapdragon; wonder if that changes when/if the coherence bug is fixed. Also, wonder what the target device is. A Note III could have a bigger battery than most phones, so maybe clocking a little higher than usual might work there. The amount of GPU would seem to suggest full-size tablet, though, since the T604 was enough for the very-high-res Nexus 10. Color me confused (and interested to see what's next), I guess. Reply
  • sherlockwing - Tuesday, July 23, 2013 - link

    Note 3 seem like the likely target if this chip will launch this year(Nexus 10 refresh also a likely target), however there have been rumors about Note 3 using Snapdragon 800 as well. Which side is more powerful & which side end up with the design win will be interesting to watch. Reply
  • jjj - Tuesday, July 23, 2013 - link

    Samsung is using Snapdragon in LTE markets not on perf, it's just more practical.
    Mass production is in august and others don't like to buy from Samsung so it's safe to assume that this will go into non-LTE Note 3 -if they don't use it there is no point making it - and maybe some updated version of S4 , if there is one. A tablet, maybe a Chromebook could also show up with it, we'll see.
  • thexile - Tuesday, July 23, 2013 - link

    Galaxy Note 2 LTE is using Exynos 4412. Reply
  • MrCommunistGen - Tuesday, July 23, 2013 - link

    The Note 2 was surprising (at least to me) in that the SoC didn't vary by market as is the case with a number of Samsung devices. The US variant of the original Note used a Snapdragon... MSM8660 I think. Most international variants used Exynos 4210. The gist is that its still a crapshoot what Samsung decides to deploy in the Note 3 and if it will vary based on the market.

    Being a Note 2 owner myself and having used the original US LTE Note for more than a month, I'd be happy to see either Snapdragon 800 or this new Exynos 5420 in the next Note.
  • speculatrix - Thursday, July 25, 2013 - link

    my Note2, N7105 (European/ETSI bands for GSM, 3G/UMTS-2100 and LTE) is an exynos 4412. Reply
  • aicom - Tuesday, July 23, 2013 - link

    The main memory flush penalty is only paid when switching CPU islands, so it should happen very infrequently during the benchmarks. It probably wouldn't make much of a difference unless the benchmarks were pathologically switching CPU complexes. Reply
  • tuxRoller - Wednesday, July 24, 2013 - link

    I doubt the benchmarking would change much since the CCI bug should only rear its head whem moving between core types and benchmarks should be keeping the cores pegged mostly. Reply
  • coolcfan - Tuesday, August 20, 2013 - link

    An article I read recently said the switching does happen during Benchmarking.

    Benchmark --> high load on A15 --> a lot of heat --> lower frequency --> lower frequency, still a lot of heat --> switch to A7 --> switch back to A15.......
  • jokeyrhyme - Tuesday, July 23, 2013 - link

    This still seems like really ingenuous marketing: calling it "Octa" but really only ever having 4 cores active at a time. As the author suggests, wake me when big.LITTLE MP is available. :) Reply
  • steven75 - Tuesday, July 23, 2013 - link

    I think "misleading" is a better descriptor than "ingenious." Reply
  • FowLang - Tuesday, July 23, 2013 - link

    even still, its only misleading if you dont read about it right ... everyone that complains about it knows whats really going on. so it isnt really ... its a 4+4 which is 8 cores ... they make no claim about how the 5410 has all 8 cores active. Reply
  • phillyry - Thursday, July 25, 2013 - link

    But misleading, by definition, means leading people in the wrong direction. In this case, that would be accurate. Samsung's branding leads people to initially believe that the Octa parts are eight core, simply by its name. So, the point stands. Reply
  • Krysto - Tuesday, July 23, 2013 - link

    "Samsung's new Exynos 5 Octa (product code : Exynos 5420), based on ARM Mali™-T628 MP6 cores, boosts 3D graphic processing capabilities that are over two times greater than the Exynos 5 Octa predecessor"

    The GPU seems to be 2x faster than the PowerVR one in Octa. Also could it be that the CCI bug came from the fact that Mali is more optimized to work with that than the old PowerVR chips are? So we might not see the same bug again.
  • alex3run - Tuesday, July 23, 2013 - link

    The new Mali T628 has twice the count of ALUs inside and the clock speed is supposed to be a bit higher. Don't forget that SGX 544 is an old GPU, which is not blazing fast. Reply
  • darkich - Tuesday, July 23, 2013 - link

    Can't wait for this to show up in Note 3, with the performance and battery life to scream in Anand's face.

    How he fails to see the amazing big.LITTLE potential is beyond me.
    He's not even interested in this top of the shelf ARM SoC!

    While Bay Trail with its FAR inferior(likely) GPU and tripple the price gets him all excited.
    Get a grip, mister.
  • flyingpants1 - Tuesday, July 23, 2013 - link

    Anand likes Intel a lot. Reply
  • ancientarcher - Tuesday, July 23, 2013 - link

    I wonder why...
    Anand, do you have a commercial relationship with Intel?
  • krumme - Tuesday, July 23, 2013 - link

    Obviously the prior A15 version was not working. Performance a few percent over s600, and bad battery life as shown by gsmarena benchmarks. I doubt the memory bug is the only reason for it? The process needs maturement or perhaps 20nm, and the software side needs to mature too?

    But besides that, this concept have great potential in reducing fab cost and helping battery life, moving complexity from production to design. Never in history have that been more important. And in perspective ARM and Samsung is working to fix the bug, while Intel is in the process of positioning Bay Trail against A12 while fixing benchmarks the usual Intel way; Look we came here first, and look we are way faster. Its a total disaster and and its a pain to watch. My eyes bleed watching this old style spin machine. Deliver the results or shut up.

    This time, Intel is not the gorilla, but a tiny unimportant player, still working to get its old monopoly, with a proven record of failures on the mobile side. Intel better used their ressources on building the business side, and on the cloud and server side instead of this old battling AMD habbits.
  • Ortanon - Tuesday, July 23, 2013 - link

    Let's not get crazy; Exynos 5 Octa has a higher IPC than Snapdragon 600, but the Snapdragon has higher battery life. Since Snapdragon isn't weak in CPU, having better battery life makes it a better phone part. After throwing so many CPU cores at the Battery Problem and still coming up short, it really isn't all that difficult to imagine someone being disappointed in the 5 Octa. Reply
  • lmcd - Wednesday, July 31, 2013 - link

    The new Octa should fix that. That's the point. It'll be able to switch CPU clusters quickly and aggressively and save more power. Reply
  • Krysto - Tuesday, July 23, 2013 - link

    I see people making this mistake a lot, but - Bay Trail is NOT A SMARTPHONE CHIP. It's a tablet chip. Merrifield is the one that will be in smartphones, and it won't be until Q2 2014 most likely.

    And I agree the GPU will be much weaker than ARM competition, too, even though it will use SNB GPU architecture (but not as many cores, or as high clock speed).
  • Krysto - Tuesday, July 23, 2013 - link

    Mali cores usually came in 2x numbers compared to PowerVR anyway. The cores are no equivalent, so that's not where the "2x" performance comes from. It most likely comes from the increase in Mali T628's architecture performance, and also because of the higher clock speed.

    Oh, and I think the GPU was at least as fast as the one in iPhone 5 and as Adreno 320, so I disagree that it wasn't fast.
  • lmcd - Tuesday, July 23, 2013 - link

    A bit weaker than Adreno 320, and iPhone 5 does well against Adreno 320 in most tests I believe. Reply
  • Łukasz Markiewicz - Tuesday, July 23, 2013 - link

    Does anyone have a link to a write up on the cache coherency issue? Can't find anything on Anand under the "exynos 5 octa" tag, but I'm sure this has been mentioned before. I vaguely remember hearing about it on one of the podcasts. Maybe around the time of the S4 review, though Anandtech's reviews was of the S600 version. Hmm... Reply
  • mrtanner70 - Tuesday, July 23, 2013 - link

    I think the bug might actually have been more of a deliberate cripple to match the s600. It would have been a marketing disaster in LTE markets for Samsung to have the international version of the s4 show any serious advantage. They had the CCI working fine for demos months earlier and it's a big issue to get through if not by design....if it was deliberate they would certainly not talk about it. Reply
  • mrtanner70 - Tuesday, July 23, 2013 - link

    And for Anand, 8 cores may well be visible.
  • rd_nest - Tuesday, July 23, 2013 - link

    Some info here:

    Andrei always has good knowledge on Exynos SOCs. He even mentioned the 5420 info before it was officially announced.
  • yowanvista - Tuesday, July 23, 2013 - link

    It appears that the 5420 can effectively use any combination of the eight CPU cores at a time if Global Task Scheduling is used instead of In-Kernel Switcher. Of course that's up to Samsung.
  • rd_nest - Tuesday, July 23, 2013 - link

    AndreiLux says GTS is working.
  • FowLang - Tuesday, July 23, 2013 - link

    The HMP (Linaro) would be dependent on which kernel they would be running if im not mistaken. I think that it was implemented in 3.9, so until Android (the assumption here is that these SoC will be used in the Note and an updated S4) upgrades from 3.4 or whatever it is at now, this wont be possible, unless Samsung can port/patch the implementation in their build of the OS. Hopefully the CCI-400 issue is resolved. Reply
  • Krysto - Tuesday, July 23, 2013 - link

    Btw, now I think I know why Samsung went with PowerVR. I don't think they had a Mali alternative for the late spring 2013 launch. There was Mali T604 in fall 2012, and Mali T628 in fall 2013, and I don't think Mali T624 was ready either.

    So that's why they probably went with PowerVR "all of the sudden" (it surprised everyone), and probably why the cache coherency didn't work too well, either, because I think ARM mainly optimized it for its Mali chips, and Imagination, if anything, will optimize for that in PowerVR Series6, and didn't revist the old PowerVR 5 to make it work well with ARM's cache interconnect.
  • aegisofrime - Tuesday, July 23, 2013 - link

    I'm a prospective Note 3 buyer... So this concerns me personally I guess.

    So, the problem with big.LITTLE from an Android perspective is not just that the Linux scheduler can't take full advantage of it, but also there's a hardware bug causing performance and power efficiency to suffer?
  • vcarvega - Thursday, July 25, 2013 - link

    I'm REALLY hoping for the Snapdragon 800 in the Nexus 10 refresh. But I guess we have just a few more months to find out either way. Reply
  • spinportal - Monday, July 29, 2013 - link

    When can we see pound for pound CPU benchmarks of Nvidia Tegra 3 vs Tegra 4 vs Intel Atom Z2580 (Clovertrail+) vs. Atom Z2760 (Clovertrail) vs i5 4200U vs i5 4650 vs Qualcom APQ8974 (Krait 800) vs APQ8064T (Krait 600) vs APQ8064 (S4 Pro) vs AMD Jaguar Kibini & Temash A6-1450 (silicon 5/23) vs... Reply
  • 0xF2 - Friday, October 11, 2013 - link

    from userland, how can one tell which of the two sets of cores is currently in use? And, is there a way to trigger the migration from the userland, other than manipulating load (perhaps CPUfreq, or something of the kind)?

    Best -OxF2
  • Wolfpup - Monday, July 20, 2015 - link

    I care very little for Android at all, but really can't care about tablets that won't be safe to use online after what, like 6 months? The state of Android is a disgrace. Hell, it took six months for my Nexus 7 to get updated, and it's a freaking Nexus! Reply

Log in

Don't have an account? Sign up now