Leaked desktop roadmap suggests Haswell refresh in 2014, no Broadwell in sight

I hope you like Haswell, because it looks like the chip might be around for a while. An official-looking roadmap published by VR-Zone suggests that Broadwell, Intel’s next-next-gen processor, won’t arrive on the desktop until 2015 at the earliest. Broadwell is expected to be fabricated using 14-nm process technology, and it could come to notebooks before migrating to the desktop. The leaked slide only addresses desktop processors, which will apparently sport refreshed Haswell silicon about a year from now. The incremental upgrade may not deliver much more than a clock speed bump, though.

According to the slide, the Haswell refresh will arrive with a new 9-series platform that adds SATA Express support. The updated Serial ATA interface is based on PCI Express, and it’s expected to offer 1GB/s of bandwidth. This is the second time we’ve seen SATA Express mentioned in conjunction with a 9-series chipset. It seems the feature will be available on all 9-series platforms rather than just the high-end Z97 variant.

Speaking of the high end, the leaked roadmap points to a late-2013 release for Ivy Bridge-E, which will purportedly slot into Intel’s existing X79 platform. That platform’s dual 6Gbps SATA ports and lack of native USB 3.0 connectivity look especially weak next to current 8-series platforms—and most 7-series ones, too. However, we haven’t heard anything an upgraded X89 chipset. An X99 platform is apparently due out with Haswell-E at the end of 2014, and we may not get an LGA2011 platform refresh until then.

At least as far as Haswell is concerned, the roadmap posted by VR-Zone matches what we’ve heard from one of our trusted sources in the motherboard industry. I hope it’s wrong about the X79 persisting until next year, though. Given Haswell’s modest performance gains and somewhat limited overclocking headroom, Ivy Bridge-E could be a good option for PC enthusiasts seeking a more potent processor. It just needs an up-to-date platform hub to match.

Comments closed
    • moose17145
    • 7 years ago

    So… I won’t have to worry about upgrading from my old i7-920 (currently OCed to 3.2GHz) for a while yet then by the sounds of it…

    • Abdulahad
    • 7 years ago

    This is the just the start of the “Falling of the Empire”.

    • Klimax
    • 7 years ago

    Apparently some people still missed some things:

    We are nearing limits. Significant increases of computation power are now mostly tied to significant increase in complexity of chips. See GPUs and how well that works for them…
    No amount of competition will fix this… only significant research into improving algorithms for use in processors.

    And AMD will hit the same limits. And same will go for ARM.

      • tipoo
      • 7 years ago

      People don’t seem to like what you said, but I think it makes sense. Gone are the days of a new architecture posting huge single threaded gains over the last. Performance gains will now be tied to shrinks allowing more transistors to be thrown at the problem. Perhaps on 14nm 6 or 8 cores will be pushed.

        • Klimax
        • 7 years ago

        I expected that, especially from AMD and ARM fans. (That bit about wall…)

        As for cores, not likely, because there is no software to use it. (At least base libraries need to be greatly upgraded and available)

          • NeelyCam
          • 7 years ago

          I downthumbed you because of the arrogance present in the first sentence

    • RdVi
    • 7 years ago

    Haswell’s supposed tock was more like a tick, now we’re set for another two ticks? Maybe a boom will follow – and no, not the sales kind.

      • tipoo
      • 7 years ago

      I think the days of new architectures posting huge gains are simply over. Gains may be had by way of better multicore use and more cores at smaller process nodes, but I don’t think we’ll be seeing huge single threaded non-extension IPC gains again. Someone else got downvoted for saying the same thing, but it makes sense to me, there has to be an architectural law of diminishing returns where things are already optimized to the point where only small gains can be had. Even with a massive L4 cache for instance the CPU was barely sped up in most tasks.

    • yammerpickle2
    • 7 years ago

    Haswell and the follow up Broadwell based chips do not look to be a big enough improvement for me to upgrade my desktop system. The power requirements are nice for mobile, but I’ll probably have to wait for Skylake-E for their to be enough of a performance increase to make it worthwhile for a new desktop. As a desktop PC person this breaks my heart, but these tiny increases in CPU performance just are not worth building a new rig around. Maybe AMD can surprise me with some compelling reason to build a new desktop performance system.

    • moog
    • 7 years ago

    Back in 2005 Moore claimed we would hit the fundamental limit in 10 – 20 yrs. On schedule, that time is upon us.

    [url<]http://news.techworld.com/operating-systems/3477/moores-law-is-dead-says-gordon-moore/[/url<]

      • Thrashdog
      • 7 years ago

      That’s my read on this as well. Intel, on the bleeding edge of fab tech, is slowing down their node transitions (and getting less benefit out of them), and the other big players like TSMC and GloFo are running up against walls as well. I’d expect process tech to stagnate for a while, until somebody gets a successor technology to silicon photolithography out of the labs and into production.

        • Diplomacy42
        • 7 years ago

        well, the wall is coming. it will hit around 2017-2020 for intel. they are trying to stretch for 2020.

    • DavidC1
    • 7 years ago

    Tick Tock is not dead. But the Desktop might be. Sure, they will continue to exist, but sales willl move away from them.

    Good for them. We had boring beige box PCs, and oversized so-called “Notebooks” that didn’t deserve to be called one for 10 years now.

    PCs, benefitted by Moore’s Law, got more portable, and smaller every generation, until few years ago, when Intel decided to go squarely against AMD. Then it stayed there for YEARS. We reached “good enough” computing for vast majority of the masses back in 2006, with Core 2! Yet people kept saying more! more! And Laptops kept getting bigger with 19 and 21 inch screens. When it goes AGAINST what used to be for 3 decades, its was inevitable that somebody would come and challenge the industry.

    Then Apple and Google came with Tablets and Smartphones. The true epitome of what Moore’s Law should have brought.

    The hell with what AMD is doing, Intel should continue to focus on going increasingly mobile. They shouldn’t kill traditional PC market entirely, but maybe its time that its no longer the sole driver.

    • riviera74
    • 7 years ago

    WOW. Nobody gets it.

    Haswell is nice for the desktop. It is exactly what the doctor ordered for mobile computing! If/when you buy your next notebook within the next month or three (especially ultrabooks), you will be ecstatic about Haswell in your mobile PC.

    Since software is not pushing the performance envelope all that hard anymore, most of the energy is in smartphones, tablets and especially laptops/ultrabooks. Haswell is what Intel needs for today’s marketplace. Sure, Intel could have had an ARM killer five years ago, but Intel is now dedicated to mobile even more so than ever. A Haswell-based notebook with much better battery life than IB or SB is something to be truly impressed with.

      • DeadOfKnight
      • 7 years ago

      Nobody is arguing that Haswell isn’t good for mobile.

      • JrdBeau
      • 7 years ago

      I “get” Haswell and completely agree with you that it’s a huge deal in the mobile space.

      What I DON’T get is why a nearly 5 year old i7-920 is still not worth upgrading.

    • JrdBeau
    • 7 years ago

    This seems very un-Intel-like by leaving the door open for AMD to inch its way back into the processor game. Intel’s had their boot on AMD’s throat for quite a while now and now they seem to be letting up a little. Come on AMD, kick the door wide open and show us what you can do!

    • Bensam123
    • 7 years ago

    Wow, it’s like Intel is on vacation. The Haswell tick was meh and now we get more then two years of it? I hope AMD has good things in store with Steamroller.

    I still think Intel should take a good hard look at splitting it’s processors into a desktop and mobile segment again and have design teams for each, instead of trying to maximize one chip for one specific area. Currently this was at the expense of desktop performance, as there really hasn’t been any notable gains… for two generations.

      • 0g1
      • 7 years ago

      They do split their processors into a desktop and mobile segment. The mobile segment is phones and uses the Atom processor.

    • DavidC1
    • 7 years ago

    Roadmap is only for the Desktop. It doesn’t mean they abandoned 14nm Broadwell.

    Desktop: Haswell Refresh(Same core, same 22nm)
    Notebook/Ultrabook: Broadwell(14nm)

    • Tristan
    • 7 years ago

    Haswell refresh = Broadwell. Next gen is Skylake, not Broadwell

      • NeelyCam
      • 7 years ago

      This.

    • kamikaziechameleon
    • 7 years ago

    Ivy Bridge – E, haswell-E what is that? new terminology to me.

      • OneArmedScissor
      • 7 years ago

      Rebadged Xeons for socket 2011.

        • kamikaziechameleon
        • 7 years ago

        ok, will haswell – E have any benefits to it for that consumer base?

        Why does it seem that the Xeon consumer products like the E processors are 6-24 months behind the low end processors. I mean they are 2 gens behind now right?

        How is that supposed to demand the premium price they do?

          • WillBach
          • 7 years ago

          They generally have more cores and an extra memory channel to feed said cores.

          • OneArmedScissor
          • 7 years ago

          [quote<]Why does it seem that the Xeon consumer products like the E processors are 6-24 months behind the low end processors.[/quote<] Server parts are always behind because they can't launch with bugs and fix it later. The chips are also several times the size of the common mainstream parts and can't be produced in volume on a brand new process. [quote<]I mean they are 2 gens behind now right?[/quote<] Ivy Bridge E is coming soon. It was pushed back by the Sandy Bridge Xeons, which were delayed due to PCIe 3.0 issues. Haswell is also 22nm, so the launches have to be staggered, and there are still a few other variations of Haswell to go. [quote<]ok, will haswell - E have any benefits to it for that consumer base?[/quote<] It will use DDR4, but we'll have to wait and see which Xeons they rebadge. The largest Haswell Xeon chip has 14 cores and 8 memory channels, but that will cost thousands of dollars. They'll probably sell an 8 or 10 core version at the $600 and $999 tiers for desktops, and an additional chip which is cheaper to manufacture. [quote<]How is that supposed to demand the premium price they do?[/quote<] Their Xeon equivalent sets the price. Intel can't sell the $2,500 chip for $999 just because it says i7 instead of Xeon. The reason plenty of people pay it for workstations, where time is money.

          • Diplomacy42
          • 7 years ago

          LGA2011 is imo worth the upgrade to the high end processor all by itself. it is a joy to own, really.

          that said, I’ll put my hex core w/HT @ 5.1ghz up against any(not E) Ivy/Haswell/Broadwell chip (when its released) that you get. “old technology” or not.

    • HisDivineOrder
    • 7 years ago

    Isn’t it amusing that in the face of everyone saying, “People aren’t upgrading their systems because they have nothing to gain!” Intel creates a scenario where they’ve essentially given this opportunity now to their most hardcore users, too?

    Whereas before, it was just the casual users who weren’t upgrading, now they’ve expanded this effect to reach even the enthusiasts. There’s very little reason to go from a desktop SB/IVB to a desktop Haswell. I hope the mobile/tablet market blows up for Intel because they’re pulling a MS and sacrificing their current, given market to try and win over a fickle and cheap (yet theoretically larger) group of more casual users.

    This gambit did not work well for MS. I don’t think it’s destined to work well for Intel, either. The problem they have is that they think that these users they’re lusting after will spend the same as the hardcore users they’re sacrificing, but those “more casual users” are notoriously cheap and both Intel and MS refuse to lower their prices on their main lines to try and win them.

    They’re like children who think growing up is all SLI’s and Eyefinity setups instead of bills and a game every now and again. They look at what Apple and Android and ARM are doing and they think they want a piece of that, but they don’t understand the bill that comes with that.

      • oldog
      • 7 years ago

      This article might interest you.

      [url<]http://gerardmclean.com/the-last-buggy-whip-maker.html[/url<]

      • Bensam123
      • 7 years ago

      Interesting…

      I only imagine the current lack of a performance increase is caused by AMD though. Hopefully Steamroller will get wipe the floor with Intel (at least performance wise) and it’ll make Intel scramble to get better performance out of their chips.

        • Klimax
        • 7 years ago

        No, we are nearing limits. Significant increases of computation power are now mostly tied to significant increase in complexity of chips. See GPUs and how well that works for them…
        No amount of competition will fix this… only significant research into improving algorithms for use in processors.

        And AMD will hit the same limits. And same will go for ARM.

    • smilingcrow
    • 7 years ago

    My fairly recent system upgrade/downgrade will seem perverse to some but it shows why for some the chase for ever faster desktop CPUs is waning in the mainstream sector.

    I had an i5- 2500K/H67 setup for over 2 years and now have an i3-3225/Z77 combo.
    Strange right!
    I just don’t do enough to warrant MOAR cores and sold the i5 for over £50 more than I paid for the i3; I get the better iGPU as well.
    The upgraded mobo was for the extra fan headers, fan controller (Asus), native/more USB 3.0 etc So I have gained in all the areas that are important to me and it was a cash flow positive upgrade; well okay, not if I include the SSD upgrade.
    The only downside is that my penis has shrunk and I’ve started wearing skirts so am now spending more on razors. I quite fancy those Quattro blade razors and apparently if you buy the right batch number they can be unlocked to 5 blades. But then you need to use higher grade shaving paste which has to be applied in a certain way for best results.
    One step forward, one step backwards; more hesitantly now as I’m not used to high heels yet.

      • RDFSteve
      • 7 years ago

      Typical Intel fanboi?

        • smilingcrow
        • 7 years ago

        Darling, you say the sweetest things.

          • dpaus
          • 7 years ago

          Is that you, SSK??

          EDIT: I did laugh out loud when I read it, though – in the lunchroom, too. There’s going to be ‘geeks in heels’ jokes around here for at least a week.

            • smilingcrow
            • 7 years ago

            A week! It takes a month just to break them in.

        • smilingcrow
        • 7 years ago

        The irony is that you appear to be the fangirlygirl.
        It didn’t matter to me if I had gone:

        Intel to Intel
        AMD to AMD
        Intel to AMD
        AMD to Intel

        I just buy what suits me and that varies depending on what’s on the market.
        It was AMD exclusively for years and it has been Intel only for recent years. I focus a lot on power efficiency as silent computing is my bag although more recently Gucci has been more my bag. And to think that some people complain about Intel’s pricing! Darling, all my Gucci’s are Extreme Editions.

          • RDFSteve
          • 7 years ago

          I’m an Intel fanboi myself; I’m just worried that you’re setting impossibly high standards for us. Gucci Extreme Editions???

          Still, good to know that you, um, you know, go both ways. That’s hotter than a 5 GHz FX.

            • MustangSally
            • 7 years ago

            I can’t believe you left me out of this discussion. I’m so hurt.

            • smilingcrow
            • 7 years ago

            Both ways! Gucci and Ferragamo is just the start, I prefer 4 way action if possible but Xeons of course. 🙂

      • Mr. Eco
      • 7 years ago

      Great post, man :))

    • OneArmedScissor
    • 7 years ago

    I hope this is because Broadwell is an actual SoC.

    If the Haswell refresh is a dedicated desktop part, that would actually be a good thing.

    Manufacturing processes are becoming more specialized, just as CPUs did.

    • south side sammy
    • 7 years ago

    why would they have a new release before DDR4 or DDR5 ( if they decide to bypass 4 ) comes out? As soon as the new ram comes out it will make everything before it obsolete. I thought the release of that was supposed to be 2014? Is this an Intel “last hoorah” ($$) on “old technology”.?

      • OneArmedScissor
      • 7 years ago

      The 2014 platforms which support DDR4 are for servers, which is really the only place it’s relevant.

      The Haswell U and Y “system on a package” support LPDDR, which already has DDR4-like voltage levels. LPDDR4 will go mainstream before DDR4.

    • ltcommander.data
    • 7 years ago

    So without 14nm and Broadwell, does that mean giving up hope for an embedded DRAM IGP in ultrabook ~15W TDP processors any time soon? A high-performance IGP with embedded DRAM is the perfect fit for ultrabooks and that Intel can’t or won’t release them is very disappointing.

      • OneArmedScissor
      • 7 years ago

      This is about the desktop roadmap. Broadwell is still coming in 2014, but it’s meant specifically for mobile parts.

      Both Intel and AMD tend to split one CPU design across two markets. It is reaching a point where the mainstream chips have to be for laptops and tablets, instead of laptops and desktops.

      Who knows, maybe they will even put Broadwell in a “phablet” sized phone?

        • mganai
        • 7 years ago

        Yeah. IIRC Broadwell will come out in AIOs at the highest level. Kind of reminds me how Westmere at the consumer level wasn’t meant for quad cores. (Lynnfield was still used for subsequent quads prior to Sandy Bridge.)

    • cheerful hamster
    • 7 years ago

    For some specific applications, namely audio and video production, some DAW builders are showing the quad-core 4770K OC @4.7Ghz beating a six-core 3930K OC @4.5Ghz. Admittedly we’re a small market and our app has always parallelized nicely with more cores and threads, but for us Haswell is looking good.

      • chuckula
      • 7 years ago

      While the Linux benchmarks for Haswell are very raw, for workloads that actually take advantage of the new instructions there are some worthwhile performance gains. Once again, it requires a specific workload and properly compiled code though.

      • OneArmedScissor
      • 7 years ago

      The L1 cache bandwidth doubled for Haswell. That helps audio processing at low latency, which is very memory bandwidth constrained.

      However, the latency did not improve, which is typically what benefits games. The L3 cache is actually a bit slower.

      What I’m very curious about is the impact of the parts with eDRAM on audio. Though it increases total memory latency, the bandwidth is potentially even greater than SB-E’s. And yet, the “L4” latency is actually half of AMD’s L3! 😮

    • Star Brood
    • 7 years ago

    Would be nice if AMD will catch up to/surpass Intel on the desktop side. Given Intel’s blind side towards desktop computing, they may not even care.

      • ronch
      • 7 years ago

      No way they wouldn’t care. Intel still gets much of their profit from big cores, and hell would freeze over before they wave the white flag and deliberately let AMD come in and steal their thunder(bolt?).

    • jjj
    • 7 years ago

    Wish AMD would go 16 cores on 20nm early next year to give us something worth buying, but that doesn’t look likely. If they only go 28nm next year ,it’s not much.

      • smilingcrow
      • 7 years ago

      Give me 20 cores at 16nm with a side salad to go and make it snappy my good man as I have to save all the internetz single handedly.

        • RDFSteve
        • 7 years ago

        Isn’t that the chip slated for the 2021 ‘budget’ Nokia phone for the 3rd world?

        • derFunkenstein
        • 7 years ago

        Give me 64 cores with 2GB of L3 cache at 10nm and a Radeon HD 9970 (ZOMG OVER 9000!!!) with a chicken salad sandwich and a bag of Sunchips. And make it snappy!

          • chuckula
          • 7 years ago

          [quote<]chicken salad sandwich[/quote<] TUNA OR GTFO!

            • MustangSally
            • 7 years ago

            Remember good taste, Charlie?

            • smilingcrow
            • 7 years ago

            I remember Good Time Charlie and he was firing on all cores; amongst other things.

      • flip-mode
      • 7 years ago

      If that’s what happens I’ll be a happy Haswell owner.

    • DeadOfKnight
    • 7 years ago

    And so the era of “good enough computing” truly begins when even enthusiasts are not excited about what’s to come. Even Intel themselves no longer want to further invest in future DIY products such as manufacturing of motherboards.

      • mganai
      • 7 years ago

      People are disinterested because Intel’s slowed down the progression of core count/speed to a near halt. It’s been 5 years since quad cores entered the mainstream market. Hex cores? Still nowhere in sight.

        • DeadOfKnight
        • 7 years ago

        You don’t even need more than 2-3, but 4 is nice for multitasking.

        • Diplomacy42
        • 7 years ago

        i’m running an intel hexcore right now…

        what really fouls my bones is when they disable HT on supposedly HIGH END parts like the i5 and i7. how long ago did HT enter mainstream? Pentium 4?

      • Krogoth
      • 7 years ago

      The truth is we already are in “good enough computing” era and it started back with Athlon 64 X2 and Conroe generation, since then there hasn’t been a mainstream application that has pushed mainstream chips to their proverbial knees. It has been evolutionary steps (throw more cache and core logic onto the CPU) and reductions in loaded power consumption.

      The reason is that it is difficult for software developers to effectively harness more then two or three threads with their program without spending a ton on time and debugging. The costs outweigh the potential benefits. Developers cannot justify this for their mainstream projects and audiences. That’s why applications where more threads = better are limited to professional suites. As a result, Intel has shifted their performance CPU platforms to this audience. Non-Xeon LGA1366 and LGA2011 chips are really workstation/server chips guise as an ultra-high enthusiast platform. The early adopters of these chips are unknowingly, beta-testers for Intel. 😉

        • DeadOfKnight
        • 7 years ago

        Even Krogoth has been unimpressed for the past several years!

        • Diplomacy42
        • 7 years ago

        knowingly beta testers for intel, thank you very much

      • Wirko
      • 7 years ago

      There are as many upgrade paths as ever for those who want more. If you’re cheap and havent bought an SSD yet, you buy one. At the other end, if you have the desire and funds, you can build a twelve core dual Xeon system with 2 or 4 GPUs. Or a cluster of those. It’s just that you can’t have more than so much performance in a single chip.

        • DeadOfKnight
        • 7 years ago

        Well sure there is still life in the GPU market and SSDs are a relatively new one, but other than that there’s really nothing barring $10,000 setups. This isn’t 1999 anymore where the real enthusiasts had to give up the car for a bit of processing power on the high end. A 4770k has more than enough power for consumer needs. So what if you need to wait an extra 5 minutes on top of your 1 hour of video encoding? Good enough is good enough.

          • Wirko
          • 7 years ago

          Even a 4770 non-K is good enough for consumer needs as overclocking on cheap won’t get you far. I guess there will be fewer K parts sold than in the Sandy Bridge era just because of this.

      • bcronce
      • 7 years ago

      Customers: We want faster
      Intel: We’re up against a ghz brick wall that is hard to move
      Customers: We want more cores then!
      Intel: Why isn’t anyone buying the more core CPUs?
      Customers: Because your dual cores are fast enough
      Intel: Well here’s some really low power, just as fast dual and quad cores
      Customers: Intel is trying to slow down progress by not releasing more cores!
      Intel: WTF?

      To really increase core counts, we need to break x86 by getting rid of cache coherency or a hybrid approach. But before we worry about many core CPUs, get programmers to make decent multi-threaded apps, otherwise it’s just a waste of die space.

        • NeelyCam
        • 7 years ago

        +1. It’s funny because it’s true.

    • Xantia27
    • 7 years ago

    It’s a little sad to hear this.

    I’m in the process of building a new rig, and I have the feeling it will be my last big PC.

    This comes from a hardware enthusiast, with also operates a PC Boutique.

    If you start to analize the global picture, the money for R&D is going to mobile devices, (celular, latptops, tablets, etc) the desktop i’m afraid will no longer get the constant updates of the previous years. In my opinion, a great shift is taking place in the industry. The best years of the PC are behind.
    Maybe some years ahead on the future, it will flourish again, but now it’s almost at standstill, al least compared to the recent past.

      • bfar
      • 7 years ago

      Good post 🙂

      If a gap opens up in the market for enthusiast hardware, someone will fill it eventually. Arguably the best days of enthusiast hardware were pre 2004, when it was far less mainstream than it is today. The form factors will definitely have to change. I still have hope.

    • Wildchild
    • 7 years ago

    The upside to this is that I don’t feel like my 2500k is even remotely outdated.

      • jossie
      • 7 years ago

      I’ve only just been convinced to upgrade from an e8400. You’re good for a long while.

      • ronch
      • 7 years ago

      So Rory Read was right: we have enough computing power.

      He would’ve sounded more credible if he worked for Intel though…

    • Chrispy_
    • 7 years ago

    Okay, AMD – Intel have officially declared lazy fatcat status again.

    They have a good performance lead but now is your chance to prove to everyone that competition is good.

      • indeego
      • 7 years ago

      Everyone? Nobody but enthusiast/workstation markets really cares.

    • ronch
    • 7 years ago

    Either Intel is stumbling on its self-imposed Tick-Tock cadence or they’re deliberately letting AMD catch up. I’m fancying the second scenario.

    Steamroller’s gonna be awesome, judging by the [url=http://tinypic.com/r/34pmdmh/5<]die shot[/url<] alone.

      • Diplomacy42
      • 7 years ago

      its true,

      intel always has amazing die shots, but when amd can make theirs look good, you know its special

    • phez
    • 7 years ago

    what ever happened to gpu-less chips?

      • ronch
      • 7 years ago

      Have you ever heard about the AMD FX series?

        • smilingcrow
        • 7 years ago

        Those are CPU-less chips. JOKE. Relax.

          • ronch
          • 7 years ago

          No, I won’t relax. Grrrrr…!!!

            • smilingcrow
            • 7 years ago

            Someone down voted you for answering a question accurately! People are Strange and I should know. 🙂

          • derFunkenstein
          • 7 years ago

          I laughed.

        • Anonymous Coward
        • 7 years ago

        The thing I dislike most about those is the graphics included in the chipset.

          • ronch
          • 7 years ago

          900-series AMD chipsets don’t come with graphics, do they?

            • Anonymous Coward
            • 7 years ago

            I could be an idiot, but I thought those motherboards always had graphics.

            • derFunkenstein
            • 7 years ago

            Not always, but the 880G and 980G chipsets have onboard graphics still. It’s old, old graphics – Radeon 4250 hasn’t changed much, but yet they still sell them. Well, they still offer them, I doubt anybody actually buys them.

    • jdaven
    • 7 years ago

    To all you madmanoriginals, neelycams and chuckulas, out there, as I have been saying all along, tick tock is dead. There will be no intel 14 nm chips next year at all. Intel may not be the first with 14 nm features as well.

    Silicon running out, lack of competition, shrinking pc demand, etc is contributing to this. So you get your wish. No AMD, no competition, slower chip releases, higher prices, failing pc companies that tried to get apple like margins. Thank you for your undying, unquestionable intel loyalty and apple hatred. That sentiment has killed the industry.

      • chuckula
      • 7 years ago

      If you can prove that Intel has stopped construction of all those 14nm fabs then maybe… otherwise it would be stupid of them to drop the money on fabs that are just going to sit idle producing nothing. It’s one thing to say that Desktop is getting less attention… duh… but those slides that you are relying on say nothing about a complete lack of 14nm mobile products in 2014 either.

      • albundy
      • 7 years ago

      well, now, you’re welcome! i am glad to see someone appreciates me!

      • Sargent Duck
      • 7 years ago

      What does Apple hatred have anything to do with an article about Intel processors?

      I hate Apple and currently have an AMD and Intel chip. So where does that leave me?

        • chuckula
        • 7 years ago

        So I’m supposed to love Intel while hating Apple.. which buys a huge number of chips from Intel every year… I’m very conflicted.

          • peartart
          • 7 years ago

          Intel is the Apple of x86. Better product, all the profits.

          • mganai
          • 7 years ago

          Lol. Doesn’t make sense. Intel’s been providing CPUs for their computers since they bailed on PowerPC.

          Still rocking this Core 2 Duo iMac.

        • smilingcrow
        • 7 years ago

        You sound quakers so it probably leaves you in bed with Harpo Marx slurping on Duck soup.

      • ronch
      • 7 years ago

      I don’t think Intel themselves expect Tick-Tock to replace Moore’s Law. Unless they can throw an infinite amount of resources at chip design and manufacturing there’s no way Tickity-Tock can go on forever. It’s gonna slow down increasingly.

      • danny e.
      • 7 years ago

      [quote<]Silicon running out [/quote<] what?

        • Buzzard44
        • 7 years ago

        He means there’s not enough sand.

        Or that we’re reaching the physical limits of silicon’s ability to scale down to make smaller transistors.

        Either one of those meanings works for me.

      • NeelyCam
      • 7 years ago

      Maybe I should bookmark this so we can get back to it in 2014

      • MarkG509
      • 7 years ago

      The general water-cooler chat consensus among my friends at work a few years ago was that Intel was trying to kill every competitor by 22nm, because the tail of the curve would given anyone still alive a chance to catch up.

      AMD? Pretty much.
      IBM and Oracle? Their high-end and software is sustaining them.

      …then along came ARM on the low-end.

    • dashbarron
    • 7 years ago

    I am pretty disappointed at the meager benefits of Haswell. I thought it might give me enough incentives to replace the ole Q9450. Maybe some of Scott’s new CPU reviews will give me the comparisons.

      • ronch
      • 7 years ago

      I expect Steamroller-based 8-core CPUs this year to handily outperform Haswell in multi-threaded apps while dramatically closing in on per core performance. Even now, Vishera can show signs of competence compared to Haswell, a far wider machine than Bulldozer/Piledriver are. Many expect Steamroller to be 30% to 45% faster than the last iteration, which means it will be pretty close to Haswell per core, and dominating it with twice the number of cores.

        • dpaus
        • 7 years ago

        If Steamroller shows 30% – 45% IPC gains, I’ll buy you [i<]and[/i<] NeelyCam beer-n-wings just for the hell of it! 🙂 Even AMD's overly-optimistic marketing department is only talking about 10% - 15% - 20% gains. If they manage to do that with some power reduction, though, that'll still be plenty competitive performance-wise with Haswell, and will almost certainly beat it on TR's famous 'value' scatter plot (even more than the 8350 is already doing, that is)

          • ronch
          • 7 years ago

          Well, it does look like the integer cores are being starved, so doubling the decode bandwidth as well as many data structures across the board should net some nice perf gains. We shall see.

          • MustangSally
          • 7 years ago

          Go big or go home: offer to sponsor the Annual TR BBQ!

          • Ryhadar
          • 7 years ago

          I’ll pitch in if AMD releases desktop Steamroller on schedule as well.

          • ronch
          • 7 years ago

          Can I swap the beer and wings for a high end AMD CPU instead?

            • chuckula
            • 7 years ago

            Why do you want to take an economic loss on the deal?

            • ronch
            • 7 years ago

            Who said I was going to use it? I could sell it for a nice sum, surely worth more than the beer and wings.

            • RDFSteve
            • 7 years ago

            $230 of beer and wings?!!? :-0

            No wonder NeelyCam backed out of the bet!

            • GeneS
            • 7 years ago

            I’m pretty sure he means the $800 FX chip with the 5 GHz turbo setting.

            • dpaus
            • 7 years ago

            If you were chuckula, I’d just assume you were fishing for downvotes, but now you want chips too??

            So it’s either beer-n-wings for you and Neely, or fish-n-chips for you and chuckula?

          • ronch
          • 7 years ago

          Considering how Neely and I are on opposite sides of the fence (he loves Intel while I root for AMD), if Steamroller ends up really fast Neely will probably end up having all the beer to himself while I feast on the wings.

        • chuckula
        • 7 years ago

        [quote<]I expect Steamroller-based 8-core CPUs [b<][i<]this year[/i<][/b<][/quote<] Really??!?!? Could you tell us where they will be? We'd really like to see them too! -- AMD management ... I wonder if the same people who got excited over the slide showing Kaveri launching this year have a cognitive dissonance since the exact same slide shows that Piledriver is here for at least the rest of 2013, and don't give me some line about how a four "coar" part with an IGP can challenge Haswell at CPU loads ...

          • dpaus
          • 7 years ago

          ^^ See? See?

          Fish-n-chips, right there, that’s what I’m talking about! Did you see him repressing me?

            • HTarlek
            • 7 years ago

            I wanted to give a one-half plus for the Monty Python reference, but I had to round up.

            • ronch
            • 7 years ago

            All this talk about food makes me want to have that imaginary case with the food compartment.

          • ronch
          • 7 years ago

          It was so fast it just zipped by. I didn’t see it either.

        • Vaughn
        • 7 years ago

        ” Vishera can show signs of competence compared to Haswell”

        Have any links to back this up?

          • Anonymous Coward
          • 7 years ago

          The statement leaves a lot of room for interpretation, and there are in fact some benchmarks where Vishera posts competitive scores. I don’t think you have any room for argument.

      • ronch
      • 7 years ago

      Come on, man. Get over it already!

      • Vaughn
      • 7 years ago

      ???

      Haswell offers what 50% more IPC than Core2quad that isn’t enough for you?

        • briskly
        • 7 years ago

        Not only that, but the bump in clock speed, usb3, and general platform upgrades weren’t incentive enough? Some people…

          • dashbarron
          • 7 years ago

          I know right, it’s like the OP is one of those $.1/GB people.

          The USB 3 is the next enticing thing about it (besides pure sp33d), but it sounds like there have been some questionable mumbling about the implementation of it, bugs, or finding a multitude of cases that have actual headers. This is minor stuff, but if I can hold out, I’d rather wait a generation for this stuff to be ironed out.

          Plus PCI-E 3 full integration, SATA 3.2 revision, etc. I don’t know how prevalent these are yet so instead of wasting the energy researching, I’ll wait a tick* to see what pops up.

          • Diplomacy42
          • 7 years ago

          it doesnt matter what improvements are made to the platform, the actual chip never gets the credit. one of the main reasons i adopted lg2011 was for the platform improvements.

        • Star Brood
        • 7 years ago

        Hey it’s all a matter of perspective. Personally, I won’t upgrade unless I see at least 2.5x the performance increase (and at a compelling enough price).

        I think Core2Quad is more than sufficient for most tasks as it is. The only thing I do that is still CPU bottlenecked is play StarCraft 2, but I’m not going to shell out a $500 overhaul just for one little game. I’d rather blame Blizzard for not updating the game engine (especially if they want to charge full retail price for their expansions).

    • PrincipalSkinner
    • 7 years ago

    Fabrication problems with 14nm and/or lack of competition.

      • NeelyCam
      • 7 years ago

      Or a misunderstanding of what “refresh” means (i.e., it may mean Broadwell)

        • Antimatter
        • 7 years ago

        I also think that may be a possibility. But there’s another leaked roadmap on Xbit-labs that indicates that the Haswell refresh is also on 22nm and Skylake will follow in 2015 with 14nm.

          • NeelyCam
          • 7 years ago

          Link?

            • Antimatter
            • 7 years ago

            [url=http://www.xbitlabs.com/news/cpu/display/20130422193800_Intel_Cans_Rockwell_Processors_for_Desktops_Spy_Shot_From_Intel_s_Roadmap.html<]X-bit labs: Intel cans Rockwell processors for desktops[/url<]

            • NeelyCam
            • 7 years ago

            Hmm.. looks like the same exact slide..

        • chuckula
        • 7 years ago

        Neely: It ain’t broadwell on the desktop. That doesn’t mean Broadwell doesn’t exist in 2014, it just means that it is targeted at notebooks and tablets. OK, I’m sure there will be NUC-like systems and all-in-ones that use Broadwell, but that’s just repurposing a mobile part and isn’t the same thing as launching a desktop Broadwell.

        If it makes you feel better, I think that a Haswell refresh that is actually targeting the desktop could be a good thing though since we are obviously just getting overclocked notebook parts on the desktop now. It’s not all bad.

          • NeelyCam
          • 7 years ago

          [quote<]OK, I'm sure there will be NUC-like systems and all-in-ones that use Broadwell, but that's just repurposing a mobile part and isn't the same thing as launching a desktop Broadwell.[/quote<] What's the difference if you put it in a desktop case? Or is your point that there won't be any 70-90W TDP Broadwells, so they don't qualify as desktop processors? Is that even confirmed? I know Charlie thinks there won't be any LGA Broadwells, but is there any reliable source for any of this?

    • chuckula
    • 7 years ago

    If Haswell Refresh == Intel finally getting its act together with a 6-core LGA-1150 part in 2014, then it’s OK.

    A die-shrunk 4-core Broadwell desktop part in 2014 will produce an even bigger Meh than Haswell did since Broadwell is going to be even more mobile-focused than Haswell already is.

    [Edit: Whenever I put forth a suggestion for how Intel might improve and it gets downthumbed, I take it as proof of what the AMD fanboys truly fear. Despite all the rhetoric about Steamroller, even they know that it wouldn’t stand a chance against a 6-core Haswell part.]

      • Kurotetsu
      • 7 years ago

      With the way Intel segments its desktop parts, there’s no way in hell they’ll ever produce a 6-core chip for a mainstream platform like LGA1150. If you have a real, legitimate need for a 6-core chip, then you’re already running on LGA2011 (and paying that nice premium for it). So why would they bother?

        • chuckula
        • 7 years ago

        [quote<]there's no way in hell they'll ever produce a 6-core chip for a mainstream platform like LGA1150[/quote<] Yeah there is: If Steamroller lives up to the hype. If Steamroller doesn't live up to the hype, then I can agree with your point. Intel doesn't need a brand-new architecture to compete with AMD, but adding a couple of cores (and potentially dropping the IGP to keep the die size reasonable) could be a legitimate response to Steamroller.

          • indeego
          • 7 years ago

          People stopped caring about AMD on the low-end of x86 when tablets/smartphones came out.

          • Anonymous Coward
          • 7 years ago

          [quote<]Yeah there is: If Steamroller lives up to the hype.[/quote<] Yay for competition. Anyway other than the 28nm process, I can't see any way for Steamroller to [i<]not[/i<] improve nicely on Piledriver. They just need to keep refining things, and its hard to go wrong with refinements.

            • ermo
            • 7 years ago

            [quote<]"(...) and its hard to go wrong with refinements."[/quote<] True, but you typically also need at least a couple of iterations before you can really feel the difference (re. BD -> PD being a rather, er, [i<]incremental[/i<] improvement)... I'm certainly hoping for the top Steamroller part to beat my current 2600K, which will make it difficult for me to resist upgrading, due to features such as AMD's equivalent of VT-d, ECC RAM support, 6 SATA-600 slots, working USB 3.0, at least 2 full PCIe x16 slots AND the support for CPUs with unlocked multipliers, all for less than what a gimped IB or Haswell setup would cost. Did I mention that I truly [i<]hate[/i<] Intel's product segmentation strategy?

        • MustangSally
        • 7 years ago

        [quote<]If you have a real, legitimate need for a 6-core chip...[/quote<] ...you're probably already running an AMD 8350 🙂

          • mganai
          • 7 years ago

          Their modules = Intel hyperthreaded cores.

      • ronch
      • 7 years ago

      Well, since Ivy Bridge cores are roughly 50% faster than Piledriver cores at common speed bins (say, a 3570K vs. 8350), and since many ‘smart’ little kids are expecting Steamroller to be about 30% to 45% faster than the outgoing generation, then perhaps AMD can catch up to Intel or at least play in the same ballpark. With 8 cores, AMD is gonna kick any 4-core Intel part particularly in highly threaded stuff. Single thread may still be Intel’s domain but at least AMD wouldn’t look so embarrassing.

      The only way Intel’s gonna plop two more cores in there is if AMD starts looking pretty good with their SR cores. Otherwise we can all forget about it.

        • DavidC1
        • 7 years ago

        Or maybe, Intel will go “meh” and continue developing architectures and chips with peak perf/power where AMD can’t compete with brute cores anymore.

      • WillBach
      • 7 years ago

      Without more memory bandwidth, which desktop scenarios benefit from another two Haswell cores? Of those, which are important enough that customers will pay for another 30 mm[super]2[.super] of high margin silicon and the associated tapeout and validation costs? Of those, how many are not also so important that customer wouldn’t also to pay for Haswell-E, Knights/CUDA/FPGA accelerator cards, cloud/LAN server offloading, etc.? The value proposition for a six-core “big core” consumer chip from Intel is very difficult to make.

        • chuckula
        • 7 years ago

        [quote<]Without more memory bandwidth, which desktop scenarios benefit from another two Haswell cores? [/quote<] I'll answer out of order: 1. Which workloads benefit? Exactly the same workloads where Steamroller will have its highest performance, which is to say massively parallel ones. If you don't think those workloads are important that's great, but at the same time you can't say Steamroller is wonderful and a 6 core Haswell would be useless: Either they would both be big improvements or both would be meh. 2. Bandwidth: Haswell has gobs & gobs to spare. Pump up the DDR3 speed to 2133 or higher and bandwidth is not an issue at all.. if anything unless AMD has completely redone the memory and cache architecture in Steamroller, it is AMD who should be worried about keeping more cores fed.

          • WillBach
          • 7 years ago

          Are we still talking about the desktop?
          [quote<]workloads where Steamroller will have its highest performance[/quote<] Steamroller's big billing is improved single-threaded performance. If anything, AMD would rather sell a less expensive expensive APU with fewer integer modules and that is possible with better single-threaded performance. The use cases that matter most on the desktop are single or lightly-threaded on the CPU side, whether or not they utilize the graphics cores.

            • chuckula
            • 7 years ago

            [quote<] Steamroller's big billing is improved single-threaded performance.[/quote<] Yes, but even AMD's optimistic estimates put it well short of Haswell. However, per core improvements x lots of cores yield more interesting improvements in highly threaded workloads. The trick is keeping all those cores fed.. can AMD do it properly? Once again: Steamroller improves single-thread performance but is by no means fully caught up to Haswell. OTOH, a 6 core Haswell negates the areas where Steamroller has its strongest performance.

    • BoilerGamer
    • 7 years ago

    Not sure I would give Haswell two years given how minimal the gain is from Ivy Bridge. Hopefully AMD hit a home run with Steamroller and knock some sense into Intel.

      • ronch
      • 7 years ago

      In case you missed it.

      [url<]http://tinypic.com/r/34pmdmh/5[/url<] All beefed up! Edit - I go through all the trouble to resize the image and upload it and post the link here, and all I get is a... downthumb? Ingrate.

        • flip-mode
        • 7 years ago

        Wow, that die shot tells me all about the performance. Looks to be 70% faster than Haswell.

          • Anonymous Coward
          • 7 years ago

          You think it [i<]looks[/i<] fast, well just wait till you smell one. It [i<]smells really fast[/i<].

          • ronch
          • 7 years ago

          It’s called performance profiling… at a glance.

          You want some benchmarks?

            • RDFSteve
            • 7 years ago

            It’s the engineering version of a drive-by shooting.

          • chuckula
          • 7 years ago

          70% faster? Flip-mode, you truly are an Intel optimist. I’d say 90% on a bad day… look at that chiaroscuro shading!

        • Alexko
        • 7 years ago

        It’s probably Excavator, though.

      • DeadOfKnight
      • 7 years ago

      It could just be overhyped as usual, but it seems to me that AMD is a lot more confident about Steamroller than they were about Piledriver.

      • Geistbar
      • 7 years ago

      Well the refresh might provide a bit more of a boost than we got with Haswell to start with, and if Intel sticks with their standard practice (not guaranteed) it would also be die-shrunk as well, likely allowing — at the very least — a reduction in power consumption.

        • willmore
        • 7 years ago

        Broadwell was supposed to be the die shrink of Haswell.

          • Geistbar
          • 7 years ago

          Was it? I guess I over-interpreted “refresh” then. If it’s just an especially good new stepping then yeah, I suppose we shouldn’t expect too much new in the meantime. My bad.

Pin It on Pinterest

Share This