Rumor: Intel puts the pedal to the metal on its HEDT refresh

The notorious rumor mongers at Digitimes report that Intel is planning to move several upcoming releases foward. The first rumor concerns the release date of the rumored Basin Falls high-end desktop (HEDT) platform, the accompanying Skylake-X and Kaby Lake-X CPUs, the X299 chipset, and the rumored LGA 2066 socket. According to "Taiwan-based PC vendors,"  Basin Falls' premiere is moving from the end of July to a release date during the Computex trade show at the end of May. This news comes just days after rumors that the same launch was being moved from late July to late June.

For those who don't follow the minutiae of Intel's ever-shifting codename structure, Skylake-X is expected to be the Skylake incarnation of Intel's long-running HEDT series of chips, which have traditionally born the '-E' suffix. Skylake-X CPUs are rumored to pack six, eight, or ten CPU cores and a quad-channel memory controller inside packages with at least 2000 pads. These chips are the natural successors to Broadwell-E models like the $1650, 10-core Core i7-6950X.

The nature of Kaby Lake-X is not quite as well understood, but rumors suggest these chips will be quad core processors with dual-channel memory controllers that fit into the same motherboards as the Skylake-X chips. The advantage of these processors over standard-issue desktop Kaby Lake chips like the Core i7-7700K may be limited to higher clock speeds allowed by more aggressive TDPs, as well as any feature advantages inherent to the X299 motherboard chipset.

In news that could affect a larger part of our audience, the rumor mill is also churning out speculation that Intel's Coffee Lake mainstream chips' release is being moved forward from January 2018 to August 2017. Coffee Lake is widely expected to be another subtle refinement of Intel's 14-nm architecture, similar to the relatively small changes observed in the Skylake to Kaby Lake transition. Some speculate that top-of-the-line Coffee Lake processors may sport six CPU cores. This would represent the first time that Intel has increased the core count of its mainstream processors since the 2008 release of the first Core 2 Quad processors.

The rumor mongers speculate that Intel is quickening its cadence as a response to AMD's renewed competitiveness in the realm of high-performance desktop CPUs with its existing Ryzen R5 and R7 lineups, the coming Ryzen R3 chips, and the rumored 16-core models and accompanying X399 motherboard chipset. The move could also be somehow related to Intel's cancellation of the Intel Developer Forum conference.

Comments closed
    • Scoop
    • 3 years ago

    Tesla buys AMD …
    <News at 11>

    … Musk takes on 4th CEO title!

    … AMD abandons PC market!.

    … electric cars are the future man! (seriously Al Gore demands it!)

    /bangs head on desk! (over and over again!)

      • K-L-Waster
      • 3 years ago

      /Tries to figure out what in blue blazes an article about Intel’s HEDT lineup has to do with Tesla, AMD, and Al Gore…

    • elitech
    • 3 years ago

    [quote<]This would represent the first time that Intel has increased the core count of its mainstream processors since the 2008 release of the first [url=https://techreport.com/review/14573/intel-core-2-quad-q9300-processor<]Core 2 Quad processors[/url<].[/quote<] You are off by [url=https://techreport.com/news/11565/intel-pulls-curtain-back-on-the-core-2-quad-q6600<]a year[/url<].

    • Kretschmer
    • 3 years ago

    I just bought Kaby Lake; please delay Coffee Lake. πŸ˜€

    • blahsaysblah
    • 3 years ago

    Great timing yet again. PCI-E 4.0 will be out soon. Same as that almost immediately worthless X99 PCH and its DMI 2.0/PCI-E 2.0 lanes.

      • chuckula
      • 3 years ago

      Indeed, and considering that AMD has unequivocally stated that AM4 will not be upgraded for at least 4 years, it’s pretty obvious that not one AMD platform will be supporting PCIe 4.0 this decade.

      Truly a sad state of affairs, and it’s patently obvious that AMD is anti-consumer by forcing us to buy PCIe 3.0 RyZen products instead of waiting until next year to launch RyZen with PCIe 4.0.. How dare they.

        • albundy
        • 3 years ago

        because…pcie 3.0 is fully saturated? guess it makes sense for those running 4 way sli.

        • Krogoth
        • 3 years ago

        This is a complete non-issue for the customer-tier market. PCIe 4.0 make sense for high-end enterprise-tier NVMe devices, high-end enterprise-tier GPGPUs and 100Gbps NICs.

        Intel platforms have similar plans where PCIe 4.0 is going to be Xeon-tier only for a while and you will not see any mainstream platforms with PCIe 4.0 until the next decade.

    • TwoEars
    • 3 years ago

    Speaking of rumors have you heard the one about nvidia accelerating their product cycle and aiming for a Volta Q3 release? Only rumors of course but not good news for AMD if it’s true.

    • Laykun
    • 3 years ago

    The real question is, will I be able to slap a coffee lake CPU in my Z170 motherboard? It’s socket 1151 but I just don’t trust Intel to not force me to upgrade my motherboard.

    • Klimax
    • 3 years ago

    All these rumors have one interesting thing in common. No rumor has anything about target frequencies. There might be fun thing in hiding.

    • herbcanter
    • 3 years ago

    Very tough for Intel to compete if AMD offers similar performance for nearly half the price . Even if Intel cuts prices by hundreds of dollars may not be enough to stop the Ryzen train.

    Intel may charge more but nowhere even remotely close to how overpriced their stuff is now.

      • Klimax
      • 3 years ago

      “Similar”…

        • tipoo
        • 3 years ago

        “for”

      • tacitust
      • 3 years ago

      Something tells me that Intel will do just fine.

      • Mikael33
      • 3 years ago

      You srs bro? Remember when AMD was kicking Intel’s ass back in the K7 days and continued to offer better performance(in most cases) until Intel abandoned Netburst? Intel was able to survive even when they had a sizeable performance deficit to AMD, with Ryzen merely being competitive with them I think they’ll do just fine πŸ˜‰

    • Flying Fox
    • 3 years ago

    DAAMIT, they still leave ECC for Xeons. =(

      • Captain Ned
      • 3 years ago

      We ain’t giving ECC away for free, suckers. And we can get away with it.

      [/Intel]

    • gamoniac
    • 3 years ago

    Nothing like a little competition.

    • Welch
    • 3 years ago

    I’d say if true that it is early, and the consumer parts coming have 6c/12t and possibly clocked a smidgen higher than their current 4 core parts… That this may be an actual answer to Ryzen. I would have thought it would take Intel over a year to get a working consumer 6 core and maybe 8 core part into circulation.

    So either Intel already had their response to Ryzen in the works way back or Intel isn’t the slow moving massive behemoth some people think it is.

    Interested to see how this all pans out.

      • Redocbew
      • 3 years ago

      I would bet they already had it in the works. They might be able to shave a few months off the launch if needed, but going from nothing to finished project in this amount of time seems unlikely.

        • Gadoran
        • 3 years ago

        Bet Intel knows well what is Zen since two years ago?? It really exist the “secret” in a cpu project ? No. Engineers change their cpu team very fast in these days, going from a company to another in a short time.
        Only skilled execution and great manufacturing count today.

        • DancinJack
        • 3 years ago

        Not unlikely, it’s impossible. It just isn’t going to happen. They’ve had this stuff in the works for years, and have been generous with the amount of time they’ve allotted to make it happen. Which is why they can shave off a bit.

        • Welch
        • 3 years ago

        Oh yeah, I doubt they can go from nothing to something. I could see them modifying an existing 6 core to package as a consumer part to compete with AMD though.

        I agree though, I think they had something in the works. Just curious if they knew because what AMD had to offer or their newest chip coming out was their speculation, that is what I was getting at.

        Time will tell.

      • blastdoor
      • 3 years ago

      And/or — Intel has been intentionally dragging its feet in bringing out more powerful CPUs at lower price points because it has had no competition.

      How that there’s a little competition, no more foot dragging.

      • ChicagoDave
      • 3 years ago

      The rumor that Coffee Lake will be 6-cores for the i7 have been around for a long time – at least a few months. Nothing official has been announced so it’s still pure speculation at this point, but assuming it’s true, Intel made that decision well before Ryzen was released.

      One thing I’m curious about is if Coffee Lake is indeed 6 cores, is that overall architecture like one and a half of the standard LGA-11xx or a cut down LGA20xx ? AMD obviously went with two CCX complexes (4 core’s each) so they went with the “take a quad core and slap a second one on there” approach rather than building a much bigger chip with a half dozen to several dozen cores. I’m guessing it’ll be 2x4core parts with two cores disabled/defective, but who knows.

      I’m also very interested to see the pricing for the next round of Skylake/Kabylake -E and -X processors; Intel cannot pretend to exist in a vacuum while Ryzen offers very similar performance for much cheaper. The real differentiating factor is the number of PCIe lanes straight off the CPU, but if that’s not what’s important then Ryzen offers an excellent deal and Intel will have to respond. I also really hope AMD releases a high PCIe lane processor soon, preferably with several NVMe M.2 slots πŸ™‚

        • the
        • 3 years ago

        The internal architecture of all of Intel’s chips has been the same since Nehalem: an internal ring bus. This has been shared in consumer and server Xeons. Intel has ran into a problem on their large core count Xeons with over 12 core plus IO (memory controllers, QPI links, PCIe controllers all on die) so they’ve split up the ring network into two domains but it remains fundamentally the same with the additional bridges. The consumer parts have their on-die GPUs sit on the ring bus as a stop as well. So basically increasing the number of cores on the Intel side is a relatively straight forward process vs. how AMD scales the number of CCX.

        As noted above, Intel’s topology isn’t scaling as well with high core counts. It has been a personal prediction that Sky Lake-EP will bring a new on-die topology forward. Intel’s Xeon Phi chip with its 72 cores is already using a 2D torus design which I would expect to be implemented with Sky Lake-EP. This also would lead to Intel going forward to produce big-little designs with Sky Lake-EP cores and Xeon Phi cores.

    • Gadoran
    • 3 years ago

    I think few realize what is Coffee Lake. Consumer Core arc. is modular and Intel can put 4 6 8 etc cores in line with small effort. The L3 banks are under ring bus so they are free to add more cores without a single issue in a very very short time and without much debug.
    So this “news” is “not” a news because consumer Core is done for this since Intel adopted the full modular design.
    In August we will have a Skylake with 6 cores in 14nm++ at a cheap price and crazy clocks for entusiasts. Thanks AMD……but seriously put out an APU, a server cpu is not done (not even thinked) for games !!!! and in fact it does’t run correctly for the pile of silicon spent at GloFo.
    Intel actually does better with only 50/60 mm2 of a crap i5 or i7.

    • Captain Ned
    • 3 years ago

    So, Intel WAS just idling along and now feels like once again squishing the bug that is AMD (in Intel’s POV). Why innovate if no one else in your space is?

      • Ninjitsu
      • 3 years ago

      well, i mean they’re still not pushing single thread much higher – just adding cores and maybe dropping prices.

      Which was always the case – they were idling along on core counts because there wasn’t much point on the consumer side and AMD hadn’t forced their hand – which it now has. Single thread was limited by physics, and that appears to be.

      • Klimax
      • 3 years ago

      There was always competition: From previous chips. If no innovation then nobody buys new chips and this no money for Intel. It’s just that IPC at target power is mostly mined.

      Note: Ryzen isn’t close in IPC, so there has AMD still some room to improve. But that might kill target clocks…

        • Antimatter
        • 3 years ago

        When there’s no competition Intel gets to decide on the pace of innovation. If AMD competes aggressively Intel is forced to innovate. It’s likely we would have seen mainstream 6 core chips from Intel sooner if AMD had been more competitive.

          • Ninjitsu
          • 3 years ago

          Arguably it’s more like “Intel has to release things faster and maybe cheaper” as opposed to “innovate” – I’m pretty sure they’re just following their roadmap in terms of content.

            • BurntMyBacon
            • 3 years ago

            [quote=”Antimatter”<]When there's no competition Intel gets to decide on the pace of innovation. If AMD competes aggressively Intel is forced to innovate.[/quote<] [quote="Ninjitsu"<]Arguably it's more like "Intel has to release things faster and maybe cheaper" as opposed to "innovate" [/quote<] I'm confused. Or perhaps you're confused. Isn't releasing "things" faster raising the pace of innovation. I'm pretty sure Antimatter isn't disputing the fact that Intel is innovating. They wouldn't be able to sell new chips if they didn't at least add something over their previous models that consumers consider of value. The issue in question is the pace at which these innovations are released.

            • Antimatter
            • 3 years ago

            Yep. Increasing clock speeds, core count, IPC or adding new instructions or features all count as innovation.

            • the
            • 3 years ago

            The 12 core chip is part of their roadmap. Intel has been using the low core count Xeon chip for the highend desktop market since they split the Xeon line up into three tiers back in 2011. The catch is that Intel likely had no plans for a 12 core desktop Core i7 but rather was going to use a 12 die with 2 disabled cores to make 10 total.

            The midrange Xeon die is likely to top out at 20 cores and the high core count die goes all the way up to 32. The LGA 3647 packaging is also quiet massive so they have room to put several CPU dies there if they wanted to go in that direction too. They’re already using that extra packaging space to place some accelerators there (FPGA, Omnipath, Nervana tech etc.).

        • w76
        • 3 years ago

        “There was always competition: From previous chips.”

        Hooray! You solved many woes in economics circles. We can now safely toss out anti-trust/monopoly laws and embrace a re-built AT&T, Comcast, Google, Microsoft and Amazon-dominated future, because hey! They gotta compete with THEMSELVES! We can even toss SpaceX’s rockets in to the sea, because Boeing/Lockheed has to compete with THEMSELVES!

        Edit: And have you ever seen the cars Detroit produced after they consolidated the US market but before the Japanese invasion really took root? Look at a 1969 Mustang and a 1987 Mustang and try saying with a straight face competing with yesteryears products does anyone any good. Detroit quality took a nosedise, Bob Lutz admitted it openly (and famously), and it didn’t stop until they realized everyone was driving a Camry.

          • chuckula
          • 3 years ago

          You’re making one massive error* in that “argument” and that is Intel needs to convince people to upgrade from perfectly operational systems to newer systems in order to make money.

          And Intel’s parts last a *long* time all things considered. Like, I’m running a 9 year old Core 2 Quad chip in a server right now that’s not even close to being retired and replaced.

          You can’t say that about smartphones, which are basically designed to self-destruct, or ongoing services like Cable/wireless, etc. where you can’t just pay for one month’s service and then buy more at some indeterminate time in the future when you feel like an upgrade.

          So yes, the argument that Intel needs to compete with itself is 100% valid as far as Intel’s situation is concerned, and if you think I’m wrong, the just write me a check for $1 multiplied by the number of posts that have been made on TR about how some random guy isn’t upgrading from his 2500K because a new Intel chip isn’t impressive enough. I’ll retire then.

          The fact that his argument applied to Intel may or may not be valid when applied to some other randomly selected company (and he never claimed that it was) isn’t the issue.

          * Actually several, but this one is easy.

            • w76
            • 3 years ago

            Edit: Had a longer post, but not worth it. I had to go back to the 1970s to even find papers debating a notion that’d support the ridiculous concept that Intel is driven forward by more than a bare-minimum (determined by a forecasting model, Intel has very sharp economists on staff) to drive a profit-maximizing upgrade cycle.

            So instead, I’ll just link you here, there’s some decent sources there:

            [url<]https://en.wikipedia.org/wiki/Perfect_competition#References[/url<]

            • Anonymous Coward
            • 3 years ago

            I think that not many companies keep important stuff running on 9-year old hardware these days.

            • K-L-Waster
            • 3 years ago

            I almost spit out my coffee when I read that.

            Do some research on what most banks run their core banking operations on. (Pro-tip: in most cases it’s mainframes from the 60s and 70s….)

            • Anonymous Coward
            • 3 years ago

            I’m no expert on banks, but I call BS on decades old mainframes.

            Also, you won’t catch me running anything important on 9 year old hardware without a solid failover in place. Crusty nasty software and hardware that you’re afraid to touch is a disaster waiting to happen.

            • the
            • 3 years ago

            Legacy is a hard thing to get rid of but the best cases of ancient hardware in use today is simply platforms that no long exist. For example, OpenVMS on Alpha had some popularity in businesses in the 90’s but new hardware hasn’t been shipping for over a decade and the software migrated over to Itanium but it too has effectively been killed. I know of one instance where that is still in use as the migration path appeared for Itanium and then disappeared.

            Mainframes, at least form IBM’s stock, are backwards compatible today with code written and compiled in the 1960’s. Banks still use mainframes but the hardware is far more modern as IBM continues to refresh that line. In fact, main frames hold the highest stock clock speed of any recent hardware at 5.5 Ghz.

        • kuraegomon
        • 3 years ago

        I think AMD will actually be able to improve _both_ their IPC and their clocks in their next silicon revision. Obviously they’ll be working like fiends with GF on process-related optimizations (for headroom and hence clockspeed), and they’ve already discussed identifying low-hanging fruit for architectural optimizations.

        I’m an HEDT consumer, and if the rumored AMD 16-core part with additional PCIe lanes materializes, I’m going to be on that like white on rice unless Intel _really_ lays out with Skylake-X.

          • Anonymous Coward
          • 3 years ago

          How much customization do they do on the fabrication side of things these days anyway? Seems like they would seek to minimize that.

          • the
          • 3 years ago

          Well there is always Skylake-EP and Naples if you want to really go all out. Both of those platforms have something to offer that the high end desktop parts are not getting. The main barrier would be price but this time around you’re actually getting more from these server chips to help justify the added expense.

      • chuckula
      • 3 years ago

      So it really WAS Barzini all along.

        • Captain Ned
        • 3 years ago

        I wll settle ALL family business today.

        Oh, and Tessio was always the smart one.

    • blastdoor
    • 3 years ago

    If true, pulling Coffee Lake forward by 5 months seems like a pretty big deal.

    I would be happy to see a 6 core Coffee Lake in an iMac. A custom Ryzen-based SOC might have been more interesting, but I get the impression that Apple isn’t inclined to try very hard with the Mac. Sticking with Intel is clearly the path of least resistance.

      • tipoo
      • 3 years ago

      I’m with ya. They said there would be a new higher tier of “pro” iMac, 6C Coffee Lake in there would be sweet. Hopefully higher TDP GPUs too.

      [url<]https://techcrunch.com/2017/04/06/transcript-phil-schiller-craig-federighi-and-john-ternus-on-the-state-of-apples-pro-macs/[/url<]

      • Anonymous Coward
      • 3 years ago

      Hmm, I wonder what the economics of a custom chip would be for Apple. Are high-volume desktop machines so much more flexible than consoles these days that the economics differ markedly?

    • chuckula
    • 3 years ago

    If you really want some juicy and massively unconfirmed rumors, just search for how Nate Dogg and Kaby-G are going to regulate.

      • Bumper
      • 3 years ago

      Awww sht… earn ya keep intel!

    • Bauxite
    • 3 years ago

    Kaby-X is going to be a joke, but it was being planned during year 8 or so of no competition so if you’re surprised please get out more.

    The difference between 7700k and 7740k or whatever will be a socket, a pointless TDP number and price. Literally none of the features of the platform get used. They will probably find yet another feature to turn off and make xeon only as well, some bean counter in there found out about X99 vs C612 and will have none of that!

    I wouldn’t be surprised if it is the same old sloppy heatspreader assembly and TIM too. They are both unlocked, the lottery will likely be the only thing that matters for overclocking limits.

    I also expect the pricing in general to still suck, intel is a big behemoth with a lot of inertia and will be slow to react. The broadwell-E launch is a nice unfiltered example of what they think the market should bear.

      • derFunkenstein
      • 3 years ago

      You seem really down on these hypothetical, rumored CPUs.

        • K-L-Waster
        • 3 years ago

        “Whatever Intel is going to do, it’s going to suck and be over-priced.”

        Did I miss anything?

          • Shobai
          • 3 years ago

          <quick re-read>

          Ah, that’d be a ‘no’.

        • Bauxite
        • 3 years ago

        Kaby-X is already in the ovens for the masses and sampling to the OEMs along with the PCH. Plenty of recent platform information is already out there.

      • Krogoth
      • 3 years ago

      Kaby-X isn’t going to be a joke though. It will basically be a Kaby-Lake minus the iGPU but adding in a few more cores in its place (4C/HT, 6C/NO HT, 6C/HT), possibly a few more PCIe lanes and a wider memory bus. The Xeon version will have ECC support but need C6xx chipset boards for it.

      It is meant to go up against Ryzen R7 family. Kaby-X will essentially be replacing the i7-3820K-5820K line-up.

      Skylake-X will focus more on core count.

        • Kougar
        • 3 years ago

        That’s what I don’t understand. They’d be forced to add more PCIe lanes if they want to support that platform, otherwise the M.2 slots and most of the PCIe slots aren’t going to work. The info I heard was that Kaby-X will stay quadcore, but if they’re going custom with the PCIe lanes and nixing the IGP instead of disabling it then they might as well add more.

        Whether one liked it or not the consumer / HEDT platform separation at least made a sort of sense. Kaby-X is going to throw that out the window and not really gain anything by it except sucker people to pay even more for a consumer chip on a more expensive platform. Best case half of that platform the chip won’t even be able to use.

          • Krogoth
          • 3 years ago

          That sounds exactly what i7-3820K-5820K were.

          • Bauxite
          • 3 years ago

          Is no one paying attention? They are not adding anything.

          Kaby-X 7xxx socket 2066 = Quad core, 16 lanes, dual channel.

          Joke.

            • Krogoth
            • 3 years ago

            That doesn’t make any sense since Socket 2066 requires a complete rework of the silicon since none of the pins are going have GPU support. There’s a very good chance there’s some reworking going on with Kaby-Lake X that will end-up being a small Skylake-X chip (4-6 cores) w/ SGX support (which is why it is called Kaby Lake)

        • Bauxite
        • 3 years ago

        [quote<]adding in a few more cores in its place[/quote<] Nope. The gpu is probably staying right there too as a nice chunk of dead silicon. [quote<]possibly a few more PCIe lanes and a wider memory bus[/quote<] The joke is on all of you then, it will have neither of those.

          • Kougar
          • 3 years ago

          If that was true then none of the M.2 ports on a X299 board are going to work, nevermind half the slots. Any such chip would incur heavy returns from consumers when nothing on the board works, and probably board returns as well. Board makers aren’t going to throw in a multiplexer into all x299 boards just to make 16-lane chips feasible.

            • the
            • 3 years ago

            One solution would be to put a PCIe bridge chip in the package to ensure some sort of IO consistency from the socket but that would increase cost.

            This also does nothing to address that half the memory slots wouldn’t work.

            • Anonymous Coward
            • 3 years ago

            Slots that aren’t connected to anything are slightly offensive.

    • chuckula
    • 3 years ago

    [quote<]These chips are the natural successors to Broadwell-E models like the $16500 10-core Core i7-6950X.[/quote<] THIS JUST IN! INTEL HAS CUT THE PRICE OF THE i7-6950X BY A FACTOR OF 10 TO $1650! OBVIOUSLY THEY ARE PANICKED BY RYZEN! GAME OVER MAN! GAME OVER!

      • BurntMyBacon
      • 3 years ago

      Golden

        • sophisticles
        • 3 years ago

        Shower

      • K-L-Waster
      • 3 years ago

      Short Intel now!!1! Tomorrow will be too late! They’ll be out of business before Monday!!!11!!1!

      • Forge
      • 3 years ago

      Noted, and bookmarked in case you ever make a typing error. I know it’ll never happen, but still, just in case.

        • chuckula
        • 3 years ago

        I nevar meke tiping airors!

          • K-L-Waster
          • 3 years ago

          For the record, there are two p’s in tipping…

Pin It on Pinterest

Share This