Possible Haswell Refresh CPUs listed online

Ah, and just when I was starting to miss e-tail leaks.

As CPU World points out, e-tailer ShopBLT.com has begun listing a small collection of unannounced Intel processors—and those chips could well be part of the rumored Haswell Refresh lineup. Here’s a shot of the listings, which pop up when one searches for the keyword "LGA12C" on ShopBLT.com:

(The search also brings up a handful of existing chips, including the Core i7-3770, i7-3770K, i5-3570K, and i7-4770. I cropped those out of the screenshot above, though.)

The listings suggest that at least some Haswell Refresh CPUs may be clocked 100MHz higher than current models. For example, the Core i5-4670’s peak Turbo speed is 3.8GHz, while the the Core i5-4690 listing says "3.9GHz." CPU World says the 100MHz bumps will apply to base speeds, as well. That would mean the Core i5-4690 could have a base speed of 3.5GHz, up from the i5-4670’s 3.4GHz.

As for prices, those look about in line with what Newegg charges for current members of the Haswell family. It’s not clear how close ShopBLT.com’s listings are to Intel’s volume pricing figures, though. E-tailers often charge small premiums for pre-orders of unreleased parts.

In any case, this all looks a tad underwhelming. Perhaps those 9-series chipsets expected to accompany Haswell Refresh will spice things up a little more. Based on the leaks to date, however, I don’t think I’d bother putting off the purchase of a current-gen Haswell CPU.

Comments closed
    • Welch
    • 6 years ago

    Bah pathetic, god how I wish AMD would get their heads out and start to compete again. DDR4 anyone? Quad channel memory!?

    /sigh, one can dream.

      • Homeles
      • 6 years ago

      Having a competitive AMD wouldn’t bring either of those things.

      • chuckula
      • 6 years ago

      Considering that AMD has retreated to the domain of “BUT WE’RE CHEAPER!” in all areas outside of discrete GPUs, I highly doubt that an undoubtedly expensive platform using even more expensive DDR4 coming out on the bleeding edge would be very helpful.

      They’d be better off finally getting a decent DDR3 controller done to lay the groundwork for moving to DDR4 after Intel has taken the initial hit and brought the prices down.

        • MarkG509
        • 6 years ago

        Yeah, but it’s somewhat “our” fault too, in terms of ROI. The outlook for volumes to these/any things we’ll actually buy are causing them to slow down on the ‘I’ because they’re not getting, or are afraid they (or their stockholders) might not get, the ‘R’.

        Every smart phone or tablet we buy is ~one less PC they’ll sell. Every cloud built is hundreds, perhaps thousands, of sites that don’t each need to buy their own tens to hundreds of chips.

          • Stickmansam
          • 6 years ago

          What if I told you clouds need to use chips (sometimes Intel chips) and that smartphones/tablets can’t replace PC’s for everyone?

            • MarkG509
            • 6 years ago

            Yes, while the clouds are being built, they soak up quite a few chips. Once they’re up, they become a shared resource (presumably efficiently shared), such that the total number of chips is less (significantly so, IMHO). The idea of each/every small/medium business running their own data center is over.

            Most people of “my generation” tend towards PCs, and if they have a smart phone or tablet at all, it tends to be in addition to a PC. But many of the “yutes” (reference to My Cousin Vinney) I know tend to avoid PCs/Macs as much as possible, or get wimpy ones for course-work (writing papers, etc.) in college.

            Edit: See [url<]https://techreport.com/news/26107/tablet-sales-grow-68-in-2013-android-outsells-ios[/url<]

            • Stickmansam
            • 6 years ago

            While clouds are more efficient, they still need to grow though likely not at the same rate as 1 pc = 1 chip. In fact, clouds should drive more sales of chips, esp for consumers. Instead of 1 pc for office work, you got a lower end pc or tablet/hybrid+ cloud which overall can drive more chips, just not all Intel ones (esp in tablet).For business case, there is likely a higher efficiency in moving to cloud in some situations but again you need a client and it can’t be a dumb terminal for sure, not with the latencies involved.

            As someone from the “younger generation”, I actually see PC quite often but they tend to fulfill a more limited role (work, gaming) and mostly reside on their phones. I do agree that the market has reached near saturation rates and any sales are going to be either when the device fails or no longer can perform it’s task.

    • Star Brood
    • 6 years ago

    100MHz! They’re still learning how to overclock, it looks like.

    I agree that the only news is going to be when 9-series chipsets come out.

    Btw who is going to pay the same amount for an i3 what you can get an i5 for?

      • Homeles
      • 6 years ago

      “100MHz! They’re still learning how to overclock, it looks like.”

      It’s about staying within their thermal targets.

    • MadManOriginal
    • 6 years ago

    The next upgrade I’m even considering is one with SATA-express.

    Even though people complain about small improvements, there is a benefit if you’re a tech geek and want some new stuff…it’s a chance to upgrade peripherals that you might not have. Mechanical keyboards, good mice, monitors, audio…all worthy and more long-term upgrades.

    • jdaven
    • 6 years ago

    I see ‘refresh’ Haswells holding major OEMs over for the rest of 2014 (including Apple) with the major mobile hardware upgrades falling in the GPU category with 20 nm mobile Maxwells.

    Then 2015 will be the year of 14 nm (16 nm?), 20 nm and 22 nm from the various manufacturers.

      • NeelyCam
      • 6 years ago

      I see Apple introducing next-gen MBA with Broadwell for holiday sales. 20nm Maxwell might appear in MBP in 2015 when TSMC gets yields to an appropriate level.

    • the
    • 6 years ago

    I was hoping that a part with 128 MB of L4 cache would appear. Unfortunately those parts are kept to expensive mobile systems while the desktop market is left to settle with a mediocre upgrade again.

    • Bauxite
    • 6 years ago

    4340 = 3.6ghz, 4350 = 3.6ghz, neither one has turbo so…

    Guess they bumped the igpu 50mhz and want more money for it?

    Witness more firepower from their fully operation monopoly!

      • Krogoth
      • 6 years ago

      Young fool, only now in the end you fully understand….

      Your feeble engineers and architects are no match for the power of the monopoly!

        • Neutronbeam
        • 6 years ago

        I am impressed by your response.

    • internetsandman
    • 6 years ago

    Now see, if there was an -E next to the codename, this would actuay be interesting news

      • Melvar
      • 6 years ago

      If they were releasing Haswell-E Celerons that really would be interesting.

        • willmore
        • 6 years ago

        *slow clap*

        • Ninjitsu
        • 6 years ago

        Yes, a 16 Silvermont core chip wouldn’t be a bad idea. 😀

    • Melvar
    • 6 years ago

    That 3.7GHz i3 would be interesting if it weren’t $166. The i3 doesn’t have turbo boost, so that’s the base clock rate. If the price wasn’t so high it would be a compelling alternative to the cheapest 3.0GHz i5, but not for only $14 less.

    • Duct Tape Dude
    • 6 years ago

    The 4790… I take it this is a locked CPU? It’s interesting that it is 4GHz stock. I’m curious if there will be a refreshed 4770k, and if so, is there any additional OCing headroom?

    I remember back in the Penryn days there was a noticeable difference between the E0 (SLGxx) steppings and the C0 (SLBxx) steppings. E0 ran at a lower voltage and was able to run cooler and clock higher.

    Is the same possible for this Haswell refresh?

      • nanoflower
      • 6 years ago

      Possible? Yes. Likely? Probably not because Intel seems to bin their CPUs more aggressively than they did in the past. Plus there is no over-clockable CPU in this new set.

      • crsh1976
      • 6 years ago

      We’ll only know for sure once Intel publishes the number, but Cyril mentions how those listed speeds are for turbo boost, not stock – the 4790 for instance would run at 3.6 GHz stock, 4 GHz turbo.

    • derFunkenstein
    • 6 years ago

    Wow, Intel puts a huge price premium on hyper threading. The lack of reviews around the web of these Pentium processors kind of obscured it for me because I’ve never really shopped for them. $43 (around 48%) more to go from the fastest Pentium to the slowest i3. 200MHz, too, I suppose, but it’s only around 10% to jump speed grades in the i3 line.

      • cosminmcm
      • 6 years ago

      I think HT is worth the money, especially on dual cores.

        • cygnus1
        • 6 years ago

        not trolling, but why?

          • jessterman21
          • 6 years ago

          Well for gaming, it expertly handles that light-duty third or fourth thread so common in modern games. It really only gets you about 25% more raw processing power, but going from a Core 2 Duo to my current i3 was night and day in DX11 games.

          Case in point: all the spikes in BF3 and Crysis 3 when you click on “Ivy Low” in the Trinity review: [url<]https://techreport.com/review/23662/amd-a10-5800k-and-a8-5600k-trinity-apus-reviewed/10)[/url<]

            • Godel
            • 6 years ago

            Yep, over on World Community Grid where grid computing uses every thread you’ve got, they reckon about a 25% to 30% gain from HT is reasonable.

          • cosminmcm
          • 6 years ago

          I had an old i3 (clarksdale) at work, and an Ivy Celeron at home. The i3 was clearly more responsive when doing heavy work, like opening my 30 hardware sites tabs at once in chrome while having a VM in the background. A little part of the difference was also from the higher frequency (3.6 vs 2.7 GHz), but most of the time all the 4 threads were working. And I had a SSD at home also, but the CPU felt underpowered even with that. That i3 felt faster than my Q9550 at stock, while the Celeron didn’t.

            • Welch
            • 6 years ago

            You can’t forget a large reason for the snappiness is the increased cache sizes. Even with the generational difference, the i3 Clarkdale has 4mb of L3 cache, whereas the Ivybridge Celeron only sports 2mb of L3. This makes a huge difference in windows I/O performance and we notice it as responsiveness. Of course the extra threads help in conjunction with it too 🙂

      • vargis14
      • 6 years ago

      HT on a dual core is a must these days……But I would like to see some benchmarks with a hot clocked Haswell dual core with HT on vs HT off to see t he actual gains from HT.

      I think it would be a interesting review as long as it included general CPU workloads along with gaming loads. Also with the Gaming tests SLI should be included in the review with at least GTX 760’s along with 770’s to really see how much of a bottleneck a dual core is running SLI. Also the SLI and single card performance with HT on Vs HT off would be interesting.

      In the Review the slowest non HT I5 quad core Haswell should be included in the review as a comparison of 4 slower real cores to 2 6-700mhz faster real cores with 2 HT cores riding shotgun.

      Thumbs up me if you think a review is in order

        • derFunkenstein
        • 6 years ago

        [quote<] Thumbs up me[/quote<] There has to be a better way to ask for + votes. And that way, I think, is to not ask for them. This isn't the YouTube comments section.

          • Grigory
          • 6 years ago

          Yeah, begging for up votes on TR is more like a gargantuan “Kick me!” sign.

          • flip-mode
          • 6 years ago

          Thumb me up if you agree with derFunk.

            • Bauxite
            • 6 years ago

            I deny your reverse reverse psychology.

            • derFunkenstein
            • 6 years ago

            I’m stuck. I agree with derFunk, but I want to be a cool kid and thumb you down. What to do?

            FWIW my complaint initially was with the wording of vargis14’s comment. “Thumb me up” just doesn’t work.

            • ssidbroadcast
            • 6 years ago

            I just gave Thumbs Up across the board for everyone. Now what say you?

            • flip-mode
            • 6 years ago

            I think it’s safe to say that the majority of you folks have a lousy sense of humor.

            • Chrispy_
            • 6 years ago

            Win a Free iPad if you downvote this comment.

    • Chrispy_
    • 6 years ago

    “Refresh”

    Also known as Same product, different packaging for cost reasons, add the bare minimum (100MHz) because there’s ZERO competition from AMD and you have a whole new marketing-friendly product stack for 2014.

      • stmok
      • 6 years ago

      [b<]Haswell Refresh[/b<] is a "delaying product" in order to buy Intel time for [b<]Broadwell[/b<]. They've had issues with the 14nm process node. (From what I've gathered, the desktop variant of [b<]Broadwell[/b<] has been pushed back to early 2015.) I personally despise the existence of such products. Especially when its not necessary in this particular case. (As you said, Intel isn't getting serious competition from AMD anyway). ...Don't even waste your money or time on these "refreshes".

        • chuckula
        • 6 years ago

        [quote<]..Don't even waste your money or time on these "refreshes".[/quote<] Uh... if I were upgrading from a Core 2 system then I don't think it would be a waste of time or money to look into these products. Don't make the false assumption that every consumer on earth is already running an overclocked Haswell rig.

          • Melvar
          • 6 years ago

          It wouldn’t be a waste, they just won’t make any noticeable difference vs the old Haswell chips. Even in benchmarks you’ll be hard pressed to notice a 2.5 to 3% difference without zooming in on the graph.

          • cmrcmk
          • 6 years ago

          As someone who just upgraded from a Core 2 system to an overclocked Haswell, I highly recommend the upgrade. I do have to admit a slight bit of buyers’ remorse when I read that the Haswell refresh is just around the corner. That rapid change of tech is why I love and loathe our hobby.

            • chuckula
            • 6 years ago

            Hrmm… downthumbed for giving an honest opinion based on facts. Yup, looks like the koolaid drinkers are lurking in the comments again.

            • derFunkenstein
            • 6 years ago

            Look at it this way – your new system (which you get to use now, as opposed to the near-future whenever this is released) is already such huge step up from your Core 2 system that I don’t think you’ll notice the extra 100MHz. IOW I don’t think it’s worth an extra wait for another slight bump when you’re already running older tech.

            • cmrcmk
            • 6 years ago

            Very true. The handful of benchmarks I ran showed a roughly 3x improvement over the old Core 2 Quad, so I’m not [i<]too[/i<] remorseful. 🙂

            • Chrispy_
            • 6 years ago

            There’s [b<]*always*[/b<] something better just around the corner, but holding off for [i<]The Next Big Thing[/i<] just means that: [list<][*<]You don't have shiny new toys [b<]*right now*[/b<] ![/*<][*<]You're paying top-dollar for your new toys when you do eventually get them[/*<][*<]You may pay to be a beta tester and discover horrible flaws[/*<][*<]OMG, last-gen stuff is reduced in clearance sales by [i<]how much?!?[/i<][/*<][/list<]

          • UnfriendlyFire
          • 6 years ago

          Do you need a Haswell now? Haswell refresh in the summer? Or Broadwell in 6-9 months?

        • Ninjitsu
        • 6 years ago

        [quote<]the desktop variant of Broadwell[/quote<] Is probably Skylake.

    • Ninjitsu
    • 6 years ago

    BUT WHAT IF THEY HAVE SOLDERED IHS-es?

    They won’t, i know. No K-series parts anyway.

      • Bauxite
      • 6 years ago

      Come to the dark side of socket 2011

        • Ninjitsu
        • 6 years ago

        My wallet is slapping me in the face at the mere suggestion. 😀

          • Bauxite
          • 6 years ago

          4820K was same price as 4770K at microcenter when I grabbed one.

          Motherboard with equivalent slot configuration was only $10 more due to $50 socket 2011 versus $30 socket 1150 bundle discount. Already had cheap fast ram from 2012 so was an easy decision.

          No native usb3 but the flash drives I have are the weakest link, 3rd party controller on board works fine. Only have one SSD in that system so limit of 2x6Gbps ports is no problem either. Having 24 more 3.0 pci lanes is far more useful to me, I have infiniband cards to feed and still room left over for a SAS card if more drives are needed.

            • Ninjitsu
            • 6 years ago

            I don’t live in the USA… 🙁

            $100 over the 4770K here, which is already selling for ~$340.

            I think my computing needs are met better by Z87, to be honest. The only platform I’ll be [i<]really[/i<] tempted to buy in the socket 2011 category is X89.

            • Bauxite
            • 6 years ago

            Pretty sure they will call it X99 for haswell-E w/ DDR4, given that haswell 1150 refresh PCH looks to be getting named {H|Z|B|Q}9{5|7}

            And while it will also be called socket 2011 (thanks intel…) it is different. Sandy-E/Ivy-E 2011 is not getting another PCH 🙁

    • DancinJack
    • 6 years ago

    Yeah, don’t care about the CPU’s. Gimme the chipset.

      • Anomymous Gerbil
      • 6 years ago

      Serious question: Given that motherboard manufacturers generally make up for the current chipset’s inadequacies (not enough USB3 or SATA 6GB/s, M.2 slots etc) by adding ancillary chips to the board, what are the expectations for the new 9-series chipsets? Is it just that mobos might be cheaper since there’l be less need to add the ancillary chips, or are there some functions that simply can’t be performed with the existing chipset + ancillary chips?

        • Deanjo
        • 6 years ago

        Besides chewing up additional PCI-e resources, many times those ancillary chipsets are not available to use until the OS is initialized. In the case of ancillary SATA, the need to initialize another controller drastically lengthens the time to boot (and usually not initialized in a uEFI setup pre-OS load). Then there is also the PITA factor of having to handle yet another set of drivers.

          • ronch
          • 6 years ago

          I myself would much rather have all functions integrated into the chipset. PCIe lanes are another reason, of course. And many ancillary chips for certain connectivity techs aren’t very good, an example of which is USB 3.0. Some manufacturers choose the cheapest option which isn’t really good as long as adding it makes the product box have another fancy logo on it.

            • Deanjo
            • 6 years ago

            I would rather have them integrated onto the chipset as well. I have several systems that have more drives than native SATA ports and having to wait for those ancillary and add-in card controllers to initialize add a good 45+ seconds to the boot times. I also hate finding USB products that only seem to work properly on native chipset implementations (audio / video USB devices are bad for this).

            • Bauxite
            • 6 years ago

            Chipset will not make more lanes from the cpu, you can shuffle around where you put your controllers but it all has to go there in the end.

            Ivy/Haswell only have 20 to use along with the PCH link which is basically another x4 2.0 (which gets shared up to x8) and most boards only expose the x16 3.0 lanes.

            • Deanjo
            • 6 years ago

            He’s not saying it will create more lanes, ancillary chipsets are usually sharing lanes with the pci-e slots.

            • Bauxite
            • 6 years ago

            Incorrect, except on the most terribly designed boards or some server configurations extra controllers use some of the eight x1 paths from the PCH which go back through a DMI link which is essentially x4 2.0, none are directly from the CPU.

            None of the good 3.0 lanes are being shared to begin with, so putting more features on the PCH won’t help.

            Hence why that bling-feature-checkbox board with 4 dual port usb 3.0 controllers and dual realjunk NICs still has a full x16 3.0 (or split to dual x8 etc) to offer the GPU. Thats 6 lanes of crap attached the PCH if you’re counting. (they even use realjunk’s full chips over pairing an intel PHY with the other half built into every PCH, its sad)

            • Bauxite
            • 6 years ago

            I should add that while it may free up some of the 8 lanes on the PCH, that is already fairly busy so if you put a high bandwidth card on a slot linked to it (e.g. 10GbE) it is sharing bandwidth with your sata 6Gbps and usb 5Gbps ports.

            Some people have learned the hard way that if a GPU is occupying the x16 direct lanes, it leaves very little for anything else.

            • Bauxite
            • 6 years ago

            Crappy block diagram of current ivy and haswell PCH bandwidth for clarity:

            GPU
            |
            |
            128Gbps x16 3.0
            |
            |
            CPU–20Gbps–extra x4 2.0 lanes only enabled by some boards*
            |
            |
            20Gbps “DMI” 2.0 (kids, meet Mr. Bottleneck)
            |
            |
            PCH
            ||||||||
            8 lanes for those extra controllers people are complaining about

            *probably not one you are using, typically Qxx or C2xx chipset models

            • Bauxite
            • 6 years ago

            UP Socket 2011 with Ivy-E, for reference:

            CPU–[20Gbps DMI 2.0]–[PCH]–8 lanes for low use devices
            |
            |
            320Gbps x40 3.0
            |
            |
            Slot configuration as desired by OEM

            If you use multiple pci express cards that need continuous independent bandwidth (e.g. not SLI GPUs for gaming) Socket 115x just doesn’t cut it.

            • the
            • 6 years ago

            It is also worth pointing out that even without any extra PCI-e devices, the chipset itself can saturate the DMI link with lots of heavy simultaneous IO. Such an IO scenario is quite rare for consumer workloads but it is rather straight forward to create a scenario where the lack of DMI bandwidth becomes a bottleneck.

            • Bauxite
            • 6 years ago

            3 good SSDs capable of saturating only half of the SATA ports on haswell PCH will just about do it all by themselves.

            If you care about device bandwidth, pony up for socket 2011, soon to have 3 flavors. (thanks intel…)

            • UnfriendlyFire
            • 6 years ago

            My Asus N61Jq laptop’s USB3 port isn’t compatible with USB2 connections. Flash drives? Nope. External hard drives? Nope. USB2 hub? Nope.

      • tanker27
      • 6 years ago

      So yeah, I’ve been questioning this for a bit. Back in the day when you upgraded more often than not you had to upgrade CPU and MOBO because of a socket change. The question is has Intel (and AMD) settled on one unified socket design. Like will the 1150 and 2011 live on for years to come are are they going to be dead with the next CPU iteration? And if not why cant they settled on a socket design. (probably ignorance on my part with these questions, but still…….)

        • DancinJack
        • 6 years ago

        Money. That’s why there are new sockets a lot of the time. Make more money if you have to buy a new mobo.

        Another reason is actual, physical limitations. Gotta have more pins to do more stuff in a lot of cases.

        • cmrcmk
        • 6 years ago

        Both companies have in the past kept the same socket/slot for several generations, LGA775 being a good example. But whenever they add or remove features from a chip that require a connection to an external device, they have to change the socket. Haswell (4th generation Core arch) just changed the desktop socket again because they brought the voltage regulator onboard the main CPU instead of having it on the motherboard. This actually reduced the number of connections they needed to the system, hence they moved from LGA1155 to LGA1150.

        Other things like adding more PCIe connections/lanes or more memory channels would also require a change in socket. Updating the same standards can also require a change since each new generation of memory seems to need a different number of physical connections, meaning a different socket for the accompanying CPU change. Basically, until Intel or AMD builds processors that have _everything_ integrated onboard with no changes in the external I/O available from them, there will always be socket changes.

          • tanker27
          • 6 years ago

          Sad because in my minds eye it just seems crazy to have to have a new socket each generation. But I guess because there really isnt a killer app the longevity of a current chip increases. Hence the reason why some of us run PCs for 5+ years.

      • derFunkenstein
      • 6 years ago

      [quote<]Gimme[/quote<] GIMME THE BEACH BOYS AND FREE MY SOUL I WANNA GET LOST IN YOUR ROCK AND ROLL AND DRIFT AWAYYYYYYY

        • Deanjo
        • 6 years ago

        Thought you would be singing Abba…

Pin It on Pinterest

Share This