Comments Locked

55 Comments

Back to Article

  • jaydee - Friday, August 18, 2017 - link

    So essentially, i5-7600 = i3-8100. Real curious to see the actual power draw and pricing.
  • milkywayer - Friday, August 18, 2017 - link

    Suddenly we see Intel become generous. Blessing us with two extra cores at the 4 core processor point after years.

    Amazing what a small kick in the ass from a competitor can achieve. Yeah,
    I'm still salty over the $1700 price on 10core cpu last year.
  • HollyDOL - Friday, August 18, 2017 - link

    There is veru little chance this would be a reaction on Zen. Developing a cpu is not a thing that can be done overnight.
    Imho Ice Lake announcement would be time wise closest to be able to fit in new cpu as a reaction to the competition (not talking about rebranded Xeons).
  • HollyDOL - Friday, August 18, 2017 - link

    very*

    Where's the edit button?
  • Notmyusualid - Saturday, August 19, 2017 - link

    Agreed, people who think Intels' releases are all Ryzan-based reactions are on drugs - hopium.

    If they watched some of the Intel University etc talks on a well-known video site, they too would see CPUs are not 6-month projects.
  • Notmyusualid - Saturday, August 19, 2017 - link

    DAM the edit button, or lack thereof!

    I wanted to add that PRICE adjustments, may well be Ryzen inspired. Not products, just yet.
  • iphadke - Monday, August 21, 2017 - link

    Zen details has been known for years, why wouldn't this be Intel's reaction to AMD? What good reason did they have before May 2017 to give more physical cores for the same price?
  • 0ldman79 - Tuesday, August 22, 2017 - link

    I don't think anyone is claiming that Intel developed a six core CPU in response to AMD.

    It is very likely that they *released* a six core CPU in response to AMD though.

    It would be incredibly short sited for Intel to not have other chips on the back burner with or without AMD as competition.
  • masouth - Tuesday, August 22, 2017 - link

    Not all Intel releases are necessarily Ryzen-based reactions but this one seems pretty obvious that it is.

    Intel has been able to go more cores for years but AMD didn't push them for a reason to so they used the "extra" space trying to improve that integrated graphics sacred cow that they seem to feel is required for all the i series chips and why bother offering an option without it because people don't have another choice.

    The simple fact is that AMD finally provided a marketing reason to start offering more cores when even Ryzen's low end is shipping with 4 or more cores.

    A rising tide lifts all boats....that don't sink first.
  • azazel1024 - Friday, August 18, 2017 - link

    And why wouldn't it be? It isn't like Intel hasn't known AMD has been developing Zen for at least a couple of years now. Even ignoring any actual industrial espionage AMD has been stating their expectations of the new architecture for awhile. So between that, rumors/leaks and possibly some unscrupulous behavior (possible, but unlikely) Intel probably had some good guesses on the shape of Zen.
  • Notmyusualid - Saturday, August 19, 2017 - link

    Have YOU known of Ryzen two years ago?

    I expect not. So I imagine AMD did well to keep their 'cards close to their chest'.
  • baka_toroi - Saturday, August 19, 2017 - link

    Yes, in a way. Everybody knows what Jim Keller joining AMD meant: a guaranteed success. So it's obvious Intel was preparing something better from the get-go.
  • Notmyusualid - Sunday, August 20, 2017 - link

    At the risk of being moderated, I think you are talking shite pal.
  • zodiacfml - Monday, August 21, 2017 - link

    True. CPU fans should know this. AMD and Intel know each others plans more than we can ever know. Intel's current situation is a choice they made.
  • jiffylube1024 - Thursday, August 24, 2017 - link

    You don't think Intel heard through the grapevine about Zen's potential years ago? Such as this public slide presentation from April 2015:
    http://www.guru3d.com/news-story/amd-zen-architect...

    Zen was being talked about back in 2014, and Intel probably heard inklings even back then about its potential. Remember AMD was doubly hamstrung by old/inefficient architecture and TSMC and GloFo being behind in manufacturing nodes compared to Intel. Intel had to be prepared for the eventuality that they could get caught up, especially if Intel started having issues shrinking down to 10nm, etc.
  • Notmyusualid - Saturday, August 19, 2017 - link

    Understood, I'm on X99 too.

    Had any 'fun' with the platform too? Memory issues, other hard-as-hell issues to trace?
  • eSyr - Friday, August 18, 2017 - link

    Not really, i3 traditionally lacks proper AES-NI support, to say the least.
  • extide - Monday, August 21, 2017 - link

    I think as of Haswell and later you get AES-NI on i3 and above, instead of just i5 and above.
  • soliloquist - Friday, August 18, 2017 - link

    So there may finally be a compelling reason to upgrade from Sandybridge.
  • ddriver - Friday, August 18, 2017 - link

    Doubt it will be worth it. You are gonna get like what, 30% improvement?
  • TEAMSWITCHER - Friday, August 18, 2017 - link

    You should upgrade from Sandy Bridge for the new features.
  • A5 - Friday, August 18, 2017 - link

    Depends on what you're doing, but if you're going i5 2500K->i5 8600K I'd expect a significant improvement in single-thread and absolute domination in multi-thread.

    Not to mention you get USB3, M.2 SSDs, DDR4, etc.
  • azazel1024 - Friday, August 18, 2017 - link

    Yeah, as it stands today in most single threaded workloads an i5 7600k is going to give you about a 25-40% increase in single thread performance depending on the workload. Stepping up to i5 8600k is likely to be more than that for single thread and multithreaded with that larger cache plus the extra cores there are likely to be some edge cases were you might see 100% increase in performance. More likely in the 60-75% range in multithread.

    That is a pretty substantial increase, especially once you include all of the other stuff you get. DDR3, USB3.1, etc.

    Intel may have made it harder for me to decide between Zen and Core architecture for my next processor. Going to depending on what the pricing structure is on these new 8th gen processors. If their pricing structure stays the same, jumping from my i5-3570 to an i7-8700k might be the way to go.
  • Notmyusualid - Saturday, August 19, 2017 - link

    Everything you said was true, but if I asked myself, 'would my mother notice / care'?

    The answer would be NOPE, again.

    And I'm betting her usage scenario is typical for the majority of home users.
  • BrokenCrayons - Friday, August 18, 2017 - link

    SB has had a good run and there's no denying the fact that those old processors can still offer up enough CPU performance to handle contemporary demands. Aside from the platform features that others have already pointed out, there's also a point about the age of components like the motherboard, RAM, and processor to consider. I'm not saying a Sandy Bridge system is absolutely going to fail, but it is worth thinking about. (Then again, I'm only just getting around to replacing a Penryn laptop that I use on a daily basis and it's not hardware failure that's driving my upgrade so I don't really have room to talk.)
  • Notmyusualid - Saturday, August 19, 2017 - link

    Fair comment, but I don't worry / expect the failures, as I only fit expensive PSUs.

    I tried a Raidmax 1000AE, and it died, taking out two very expensive GPUs with it, and I only risked it, as it had two of them to hand, from a disused Litecoin miner. It was the only thing that wasn't 'top shelf' in my rig, and it phucked me over. Oh, and you can't say 'it was stressed out due to mining', it was a BRAND NEW replacement for one that had died mining. Enter my AX1500i. So time will tell if I was right.

    But again, for the office PCs I manage, I only fit expensive supplies.
  • Old_Fogie_Late_Bloomer - Friday, August 18, 2017 - link

    Yeah, time to start eating ramen, it's about to be upgrade time for me.
  • CaedenV - Friday, August 18, 2017 - link

    Not too sure about that...
    We no longer live in a world where you ever 'upgrade'. I have sort of come to the conclusion that for my use (light video editing/transcoding, and games) my SB i7 is still 'good enough'. There is just no new load where the chip is the major bottleneck for these use-cases yet. Right now my SSDs and GPU are still the bottleneck most of the time.
    Now, this all changes rather quickly if it is a work scenario. If the speed of my computer was limiting the amount of money I can make and projects that I can do, even then this would not be a good upgrade because an i9 or threadripper would pay for itself. But as my projects don't make me any money, there is no motive to upgrade for the sake of speed.

    But there are other factors other than speed to be considered...
    Thunderbolt, USB-C/3.1 ports, 10gig Ethernet (when it really becomes a thing), m.2 SSDs... heck, I was reminded the other day that my system is still on PCIe v2 instead of v3. These are all nice things. The question in just HOW nice are they? Is it worth spending ~$1000+ to upgrade the core of my system? I am just not sure that it is. Again, these are all nice things to have, but I am not sure that I would see any real-world benefit for the particular things that I do. Stardew Valley, Minecraft, EVE online, and 1080p/4k movies are all going to play just as well on a new system as my old one. Video editing would render a bit faster... but I typically tell it to export and let it chew overnight, or during my work day, so it isn't a big deal. The difference between a 4-6 hour render vs a 2 hour render really makes no difference for how I do things (as long as it is under 10 hours).

    And that is the thing. Someday my computer will die... what do I replace it with? Do I go for another monster i7 with 32GB of RAM and dual SSDs in RAID0? Probably not. My next system will probably be a cute little i3 or i5 with an m.2 SSD, a nice GPU, and maybe a 10gigE card (because 1gbps is the true bottleneck I face all the time!). It won't really be an 'upgrade' it will be a replacement that is smaller, quieter, and more power efficient than my current rig.

    The real issue is that after 6 years of rendering video... I am not sure my SB i7 will ever actually die! The thing just keeps chugging along as fast as it ever has. Every few years I give it some fresh thermal paste and a GPU upgrade, and it is all good.
  • Notmyusualid - Saturday, August 19, 2017 - link

    For most people, just doing office tasks, browsing & paying bills, and Skyping the grandkids - NOPE.

    It is just down to your usage / requirements.

    For mobile, that is a different story.
  • edlee - Friday, August 18, 2017 - link

    Core i3 should have been 3/6 better tha. 4/4
  • Tsu_brO - Friday, August 18, 2017 - link

    yes :/ I wanted i3 = 3/6; i5 = 6/6; i7 = 6/12, and for laptops m3 = 2/4; m5 = 4/4; m7 = 4/8 (extinguishing 'iCore' for laptops and keeping only as 'mCore')
  • jardows2 - Friday, August 18, 2017 - link

    If this all pans out, and the products are released at the same price point as the 7xxx generation, that i3 8100 could be very compelling.

    I've been holding off on a needed office computer upgrade, waiting for Zen APU, but if i3 8100 is a true quad core at the ~$120 price range, that may be my ticket!
  • KaarlisK - Friday, August 18, 2017 - link

    Yup, really hoping price points do not change.
    That would mean my next build being $65 cheaper than expected, which would be the largest price drop on a while.
  • MonkeyPaw - Friday, August 18, 2017 - link

    Might still want to wait since you'll need a new motherboard. The AMD setup may still be cheaper once you get to the platform level. That, and AMD is the only reason this change happened.
  • HomeworldFound - Friday, August 18, 2017 - link

    Then the established platform stability will still be a winner if you can justify the potential extra cost to yourself, and if you actually expect a stable platform as opposed to an enthusiast / overclocking build.
  • MonkeyPaw - Friday, August 18, 2017 - link

    Kinda hard to make a conclusion as to which will be more stable, since we don't have either product in hand. The AMD chips will drop into AM4 boards already in the market, while the new Intel chips will require a new board.
  • IGTrading - Friday, August 18, 2017 - link

    Does AMD have the option to launch CPUs which have a 10% faster frequency ?

    This would be their only answer in such a situation ..
  • A5 - Friday, August 18, 2017 - link

    This is good news for everyone. Glad to have some real competition again.
  • guidryp - Friday, August 18, 2017 - link

    "any 4C/8T component has the potential to surpass a 6C/6T in certain tests"

    That seems dubious to me. The Hyper-threading bonus is maybe 30% gain overall max, often much less. 2 real cores should be + 50% consistently on those same tasks.

    No contest IMO, 2 real cores will always be better on 4 Hyper-threading ones will add.
  • MrSpadge - Friday, August 18, 2017 - link

    > The Hyper-threading bonus is maybe 30% gain overall max

    No, it's 20 - 30% average gain, depending on the test suite. And since we know there are workloads which slightly regress with HT, there have to be tests with far larger gains. 4 HT-cores giving +25% each would already tie 2 physical ones giving +50% each.
  • guidryp - Friday, August 18, 2017 - link

    @MrSpadge
    You are doing the math wrong.

    The best result I have seen from i7 with HT off/on is ~30%.

    That is 30% for 4 HT cores already counted. Typically it is Much less(sometimes even negative). And you don't multiply it by 4, all 4 cores are already counted in that best case 30% boost.

    Adding two real core is a full 50% boost.
  • smilingcrow - Sunday, August 20, 2017 - link

    Well said.
  • ltcommanderdata - Friday, August 18, 2017 - link

    That assumes the 4C/8T and 6C/6T CPUs are clocked the same. It's likely the 4C/8T CPU will be clocked higher and depending on how high, there could be cases where hyper-threading combined with higher frequency wins.
  • guidryp - Saturday, August 19, 2017 - link

    @ltcommanderdata
    That's kind of red herring. 4C/4T can beat 6C/6T if the clock speed difference is high enough.

    If want to talk about HT cores vs Real cores we should leave clock speed out of it.
  • peevee - Friday, August 18, 2017 - link

    On Intel CPUs since Sandy Bridge (if not Pentium 4) HT adds about 20% on average (sometimes even slows things down), because Intel speculatively schedules core's resources very aggressively even in a single thread per core (and the threads are going to fight for this core's cache).

    On AMD Zen it adds 50%. So with HT/SMT enabled, IPC disadvantage of AMD in multi-thread applications disappears as SMT picks up the slack.

    High-performance applications need to care about SMT/HT though, in particular halve their assumptions about caches (if they ignore existence of caches they are incompetently written to begin with).
  • Dug - Friday, August 18, 2017 - link

    Would be nice if people could use same motherboard. If getting a new motherboard is required, any word on possibility of raising pcie lanes (with bandwidth to support them)
  • qlum - Friday, August 18, 2017 - link

    I know this is not final but they did drop the freqency of the i7 part by 400mhz. If this goes without any real improvements in ipc or overclocking headroom we may very well see a performance loss for games and other applications that do not use the extra cores.

    Or in other words the area's where intel is winning against amd right now. If zen 2 manages to increase ipc and clock speeds we may end up in a situation where amd actually wins on all fronts be it to a lesser degree.

    Of course if they do increase the ipc it may just become a very level playing field assuming you don't care about the igp.
  • DanNeely - Friday, August 18, 2017 - link

    I'd expect the lower base clock to be countered by higher turbo levels for 4 and fewer loaded core configurations. Unless TDPs go up by 50% or the cores use 33% less power at load; going from 4-6 cores necessitated a drop in the base frequency to keep power in check.
  • peevee - Friday, August 18, 2017 - link

    What is it with Intel and AMD and disabling HT/SMT? Intel will disable HT for most of their lineup. After all, the hardware is there, Intel has paid for it during development and production, why cripple your own product?

    Marketing people need to be shot.
  • Jhlot - Friday, August 18, 2017 - link

    I predict no turbo on the i3s so their single threaded performance will see no gain or even small step back from prior generation.
  • stephenbrooks - Friday, August 18, 2017 - link

    As far as "fake cores" are concerned, I tried a floating-point-and-RAM-heavy simulation program on an early Intel HT system and an FX-8350 and got:

    4 cores + HT ~= 5 cores (i.e. 25% boost)
    4 Bulldozer modules (8 "cores") ~= 5.5 cores

    So I'd guess 6C/6T will be better than 4C/8T for more-or-less everything.
  • AntonErtl - Saturday, August 19, 2017 - link

    For our LaTeX benchmark (integer, running mostly in caches), I see an SMT speedup by a factor of 1.1 on a Ryzen R5 1600X, a speedup by a factor of about 1.6 from using two cores on the same module on an Athlon X4 845 (Carrizo/Excavator), and 1.14 from HT on a Core i7-6700K over just using one core/one thread.

    (This is a single-threaded application, so I ran an instance of the benchmark on the other thread/core of a core/module while the other thread core is loaded with running the same benchmark, or on one core/thread, and from that computed the speed of running two instances on two threads/cores vs. running them back-to-back on one core).
  • edcoolio - Saturday, August 19, 2017 - link

    My take:

    With no equal performance competition, Intel hiked prices and likely held off on release dates.

    With equal or greater performance competition, Intel equalizes/drops prices and needs to release designs on-time or earlier.

    Intel price drop+new design= WIN for the consumer, thanks to AMD.

    Thank you AMD.

    Signed,

    Capitalism
  • iphadke - Monday, August 21, 2017 - link

    The very first sentence "At a closed-session partner in China" should read as "At a closed partner session in China".
  • TallestJon96 - Wednesday, August 23, 2017 - link

    4c/8t vs 6c/6t at equal clocks will be the most interesting comparison. For productivity I think it'll be a wash, but I think 6c will win in gaming (even if its a small margin). This will probably put the i5s back in the gaming sweet spot again. 6c/6t @ ~4.8ghz with maybe 3200 ddr4 will crush anything you throw at it, maybe even beating the 7700k.

    Very exciting, but my 6700 @4.4ghz will last me a while longer yet. I'll be waiting for cannon lake or Zen 2, or maybe even their HEDT counter parts.

Log in

Don't have an account? Sign up now