Comments Locked

48 Comments

Back to Article

  • blanarahul - Thursday, June 20, 2013 - link

    First!
  • blanarahul - Thursday, June 20, 2013 - link

    I don't think the IVB heating issues will sorted out (I am looking at you Haswell).
  • Death666Angel - Thursday, June 20, 2013 - link

    There is a good chance they will stick with soldering for these enthusiast platforms. Anything less would get even more of a backlash than on the LGA1155/-50 sockets.
  • Rumpelstiltstein - Thursday, June 20, 2013 - link

    Why would they do that? They want you to spend more for the crappy bin 4820K over the top of its class 4770K. They gotta stick it to you somewhere with the 4770K in order to get you to do this.
  • extide - Thursday, June 20, 2013 - link

    If the 4820K is anything like the 3820, then it is not a binned and partly disabled chip.
  • dragonsqrrl - Monday, July 8, 2013 - link

    extide is right. the 3820 uses a different die than the 3930K or 3960/70X. It's a native quad-core die, as opposed to the fused-off octa-core die in its higher-end siblings.
  • piroroadkill - Thursday, June 20, 2013 - link

    Ivy Bridge and Haswell did something unexpected - they made Sandy Bridge-E look even better, with their stupid IHS design and limited improvements on the desktop.

    Ivy Bridge-E is a total waste of time, seeing as Haswell-E has already been announced, with 8 cores.
  • Kevin G - Thursday, June 20, 2013 - link

    Haswell-E is at least a full year away however.

    With Broadwell delays on the mobile side and the lack of desktop Broadwell chips, next year Intel will have its enthusiast and consumer segments back in sync for awhile.
  • HisDivineOrder - Thursday, June 20, 2013 - link

    Something tells me Broadwell delays will lead to Haswell-E delays, too.
  • Kevin G - Thursday, June 20, 2013 - link

    Why? Haswell-E is on the same 22 nm FinFET process as Ivy Bridge-E and vanilla Haswell.

    Broadwell is jumping to 14 nm and that is likely a big reason for its delay.
  • chizow - Thursday, June 20, 2013 - link

    Sorry, don't agree with that. Every time I looked at SB-E/X79 it just looked like dated tech. Much happier with Haswell, especially the modern Z87 chipset with full native SATA6G and 6 USB 3.0 ports.

    Intel just loves dragging their feet with these E platforms and I think you'll find fewer and fewer enthusiasts are going to be willing to wait for their releases unless they absolutely need the extra cores. 1.5 years and 1.5 gens behind the mainstream performance desktop platform is unacceptable, imo.
  • HisDivineOrder - Thursday, June 20, 2013 - link

    I agree. It's not like this is obscure tech no one will use. This is absolutely necessary technical advances that should have been easy for Intel to add into X79. Hell, I remember when X79 came out, everyone was complaining about how plain and boring it was. If I remember correctly, Intel had initially intended to offer more with it, but had to scale back because of some chipset snafu that forced them to remove features at the last moment.

    They were supposed to be added back in with the next chipset update, but no one--especially not then--expected to be waiting two years for a new update to the SB-E line or to get NO chipset update even when IB-E came...
  • Notmyusualid - Friday, June 28, 2013 - link

    ABSOLUTELY agree with you there pal.

    But at the end of all the testing, they still hold their own in gaming, and in most other applications / server uses they scream along with all those extra cores, and memory frequency.

    So I'm 60/40 against.... meaning I wouldn't build one unless I needed that memory performance, or the extra pcie channels...
  • bj_murphy - Tuesday, July 2, 2013 - link

    My thoughts exactly. What is the point of a so-called "high-end" platform that is consistently an entire generation behind the mainstream parts? Why don't Intel's releases follow a pattern more like Nvidia and AMD for video cards, where the high-end parts come first, followed by more mainstream parts down the road?
  • ludikraut - Thursday, June 20, 2013 - link

    Totally agree. With an 8-core Haswell-E in the pipeline, I see not reason to go to an Ivy Bridge E or any other processor. My good ole OCed I7-920 will just have to "limp" along for a while longer, LOL.

    l8r)
  • sna1970 - Thursday, June 20, 2013 - link

    Ivy Bridge - E , will have upto 12 cores. coming soon in September. havent you seen the Macpro announcement?
  • ElvenLemming - Thursday, June 20, 2013 - link

    That's Ivy Bridge-EP, the Xeon family. Ivy Bridge-E is the enthusiast line based on the Xeons, but with lower core count and higher clocks, hence still only 6 core at the high end.
  • Kevin G - Thursday, June 20, 2013 - link

    Considering that the new Mac Pro has been announced and will feature 12 cores in a single socket, these 6 core Ivy Bridge-E chips likely use a different die. It'd be wasteful to disable half the cores on a die to get a 6 core chip unless yields are abysmal. This 6 cores chips are likely native 6 core dies with the quad core being the harvested one.

    Perviously SandyBridge-E came in native 4 and 8 core dies with all 6 core chips being harvested from 8 core dies.
  • HisDivineOrder - Thursday, June 20, 2013 - link

    Couldn't the 12 core chips by Intel be a return to the integrating two separate dies into one chip a la Q6600-esque melding? That would let Intel keep their 8 core chips, die harvest to 6 and meld two of those into one chip. Sure, it's not pretty or particularly efficient, but it's cheaper than having a special 12 core die and a special 8 core die for different markets.

    It also promises to expand to 16 core dies in the future if they need to. Power sipping isn't as important to customers who need lots of Intel IPC-powerful cores. I didn't say it wasn't important, just not AS important.
  • Kevin G - Thursday, June 20, 2013 - link

    The dual die strategy wouldn't work as well now as you'd be cutting the number of memory channels in half. The Core 2 Quads were able to get away with a MCM as the memory controller was still a shared, separate chip as if it were two sockets.
  • Assimilator87 - Thursday, June 20, 2013 - link

    Each die has its own memory controller. That's why AMD's dual die G34 CPUs are quad channel, whereas the single die C32/AM3+ CPUs are dual channel.
  • Kevin G - Thursday, June 20, 2013 - link

    Socket 2011 supports quad channel memory. Putting two dies with quad channel memory support into that socket would mean that only 4 of the 8 channels possible would be usable. Similarly the number of PCI-e lanes would remain stack at 40 instead of the 84* possible.

    *DMI on the second die could be used as an additional 4 PCI-E lanes since wouldn't be connected to the chipset.
  • bsd228 - Thursday, June 27, 2013 - link

    The Mac Pro will be using the upcoming line (announced, but not shipping) of the Xeon E5-2600 v2.
  • frozentundra123456 - Thursday, June 20, 2013 - link

    Strange that the quad core has the same tdp as the hex cores, and is only very slightly higher clocked, and actually has lower turbo than the extreme edition.
  • ShieTar - Thursday, June 20, 2013 - link

    It's because Intel only has a few TDP classes. I think at this point in time, anything that might use more than 90W will be thrown into the 130W class. It does make sense though, its not like anybody will develop a case/cooling system which can handle only one of the three existing Ivy Bridge-E CPUs.
    If you are looking for a realistic power draw of the CPU, the TDP is not all that relevant for you anyways.
  • sna1970 - Thursday, June 20, 2013 - link

    it is about the quad channel memory controller.
  • sna1970 - Thursday, June 20, 2013 - link

    it is about the quad memory controller and the 40 lanes of PCIe 3.0
  • HisDivineOrder - Thursday, June 20, 2013 - link

    Wouldn't it be awesome if as an alternative to the 4820K they offered a 4830K that gave us a hexcore chip without hyperthreading for the same money as the 4820K?

    Alas, then they'd have to drop the price on the 4820K and we wouldn't want that.
  • HisDivineOrder - Thursday, June 20, 2013 - link

    If you believe rumors, Haswell-E is coming next year around the same time. SB-E, though, had an impressively long run (for those who invested), so I have my doubts that Intel will actually release it on time. If they DO manage to release Haswell-E next year on schedule, I think IB-E is an amazingly bad deal, due mostly to the chipset you're going to have use with it.

    Will Intel use fluxless solder or more of the air-gap el cheapo special they've been afflicting their post-SB CPU's with? The world wonders.

    If Intel truly wishes the enthusiast to stop using their mainstream chips (a la going from LGA to BGA), then Intel ought give thought to offering an E-series chip that AT LEAST matches the 2500k/3570k/4670k pricing since that's proven the sweet spot for ...how many years now? That would serve as balm to the horror story that will be "MOAR HASWELL!" next year if we can all just switch to the Enthusiast line.

    But right now, they want more for the CPU, more for the board, more for the memory... all so we can invest in old tech. If Haswell E rumors are to be believed, we'd also be paying more for antiquated tech not to last the year. Somehow, I think we'd probably get a year, but that's about all.

    Eh. I don't see why Intel can't just give Enthusiasts the same tech as the mainstream in the same year, even if not the same quarter. Seems like the least they could do. Oh, and reassure enthusiasts by guaranteeing them fluxless solder (addressing the complaint head on) if they spend up on an "Enthusiast" chip.
  • Notmyusualid - Friday, June 28, 2013 - link

    I know not what makes Intel tick, but my desktop shall await their next tock, if you get me...
  • cjs150 - Thursday, June 20, 2013 - link

    "I am hoping that the Ivy Bridge overheating issues are sorted with the IVB-E processors"

    Take off the IHS, then the rubbish thermal paste Intel uses. Replace with high quality thermal compound, re-attach IHS. Problem significantly reduced
  • Kevin G - Thursday, June 20, 2013 - link

    And in the process, your warranty is also voided.
  • ElvenLemming - Thursday, June 20, 2013 - link

    To be fair, technically any overclocking voids your warranty.
  • Razorbak86 - Friday, June 21, 2013 - link

    Personally, I don't really care about the warranty. I'm much more interested in the impact of better temperatures on overall performance. I de-lidded my 3770K two weeks ago, replacing the stock TIM with Coolaboratory Liquid Ultra, and I was stunned at how much the temperatures dropped. All four cores now run ~20C cooler than before on the 4.8 GHz offset overclock that I use 24/7, and I can now reach 5.0+ GHz without thermal throttling. I'm very happy with the results, and I would de-lid IVB again in a heartbeat. Once de-lidded, IVB performs exceptionally well in high overclocks.
  • ElvenLemming - Thursday, June 20, 2013 - link

    The TIM isn't the problem, it's actually quite high quality. The problem is the gap caused by the glue holding the IHS to the die. Delidding removes the glue which removes the gap.
  • apinkel - Thursday, June 20, 2013 - link

    It's good to see info on this upcoming release.

    On the x76 vs. x89 topic.

    I'm a software developer. In college I took a COBOL course. When I graduated everyone was saying that in 5, maybe 10, years COBOL would be obsolete. Here we are 25 years later and there are still a lot of fortune 500 companies who are running their businesses on effective, surprisingly flexible COBOL platforms.

    I've watched young leaders choose the latest and greatest software solutions instead of using the pre-existing COBOL platform. They believed the tool that made their life easier was in the companies best interest. They usually spent themselves into oblivion trying to integrate and manage their systems in the context of the existing platform.

    In looking back on Intel's past history, I trust that they have solid business reasons for choosing the release chipset that they did.
  • apinkel - Thursday, June 20, 2013 - link

    The big question is always what comes next and when. Until the day of it's actual release, (not just announced, but shipping) there's no way to know. It may be fun to discuss it because it's the stuff that makes this stuff interesting but no one has a crystal ball on these things.

    Just experiences, insights, decisions and actions. Kinda fun ain't it?
  • Torchholder - Thursday, June 20, 2013 - link

    No core count increase = not interested.
  • ShieTar - Friday, June 21, 2013 - link

    Most games still don't really use more than 6 cores. And anybody not into gaming is better of with a Xeon system. Single socket Xeon boards are often even cheaper than the Enthusiast-Boards.
  • Senti - Friday, June 21, 2013 - link

    Since when Xeons can be overclocked?
    And who cares about games? Most non-ancient CPUs will do just fine in them up to the crazy resolutions.
  • Hixbot - Monday, July 1, 2013 - link

    Yeah without a core count increase, we better see some impressive IPC gains over Sandy-e.
  • BlackDragon24 - Thursday, June 20, 2013 - link

    My 970 running at 4.0ghz is looking like it will last awhile....
  • coldpower27 - Friday, June 21, 2013 - link

    Yeah I still have the "old" Core i7 970. No need to consider upgrading till Haswell-E it seems.
  • Ktracho - Friday, June 21, 2013 - link

    What I'd like to know is to whom will Haswell-E appeal, given that other than graphics, Haswell is not much faster than Ivy Bridge? As for Ivy Bridge-E vs. Sandy Bridge-E, I imagine it will be primarily appealing for workstation/server applications, unless they update the chipset.
  • wallysb01 - Saturday, July 6, 2013 - link

    The "E" processors don't have integrated graphics. These are basically the Xeon E5-16xx-s. So the appeal is that none of the silicon is going to the GPU, just the CPU, allowing for more cores and better cores. Plus, Haswell E will probably see 8 or 10 cores. Ivy Bridge E5-2600s are going to get up to 12 now.
  • psyq321 - Thursday, July 4, 2013 - link

    It is a sad state of affairs that there is simply no application demand on the High-End Desktop (HEDT) so Intel can safely sell castrated and rejected 10-core Xeon E5 2600 v2 dies as 6-core Ivy Bridge E SKUs...

    With Sandy Bridge EP, the difference was not that big - 6 cores HEDT vs. 8 cores Xeon WS/Server.

    However, with Ivy Bridge EP, Xeon line will be getting up to 12 cores (15 with slightly different socket but still 2011 pins), while "high end desktop" still gets only 6 cores.

    I am sure if things were different in the application world and if we had today "killer apps" for desktop that are highly threaded, that we would be running 15 cores on LGA 2011 today (which is now exclusive for Ivy Bridge EX)

    Instead, we still have tons of single-threaded applications where it matters more to have less cores but with higher frequency. I suppose this is also why Intel is pursuing this strategy for the high-end desktop SKUs. If they went for more cores but with lower max frequency, some desktop benchmarks would not look as good.
  • adamdz - Sunday, July 21, 2013 - link

    I'm trying to find a good excuse to upgrade my 4.5GHz overclocked 2700K and I still haven't found any. This is really sad. We need some real competition from AMD here. The only compelling features of the Haswell desktop platform is the larger number of SATAIII ports and native USB3 support, but not the speed. And both E lines have outdated chipset:(
  • black_plague815 - Friday, September 6, 2013 - link

    Still running 980X @4.7ghz, still kills everything out except socket 2011 hex's, no reason for upgrade yet.

Log in

Don't have an account? Sign up now