Haswell compared to . . . everything

We’re hard at work on our in-depth Haswell review, but it’s not finished just yet. Since Intel’s press embargo has lifted, we thought we’d give you a taste of what’s in store. Here’s a quick look at how two Haswell processors—the Core i7-4770K and its Iris-infused i7-4950HQ sibling—compare to, well, everything.

Gulp.

The results are from only one benchmark, but they give us a sense of where Haswell sits in a sea of 36 processors dating back to 2005. We’ve come a long way since the Pentium Extreme Edition 840. That CPU is about eight times slower than the Core i7-4770K in Cinebench’s multithreaded test.

We’ll have much more to share with you soon. Stay tuned.

Comments closed
    • SonOfGodfrey
    • 8 years ago

    Not sure how this will turn out.

    Basically, Score/THz is proportional to IPC.

    [code<] Processor Max Clock Rate Cinebench Score Score/THz Core Name Atom D525 1.8 0.2 111.11 Atom Pentium EE 840 3.2 0.45 140.63 P4/Smithfield Core 2 Duo E6400 2.13 0.61 286.38 C2D/Conroe Core 2 Quad Q9400 2.66 0.82 308.27 C2D/Yorkfield Pentium G6950 2.8 0.91 325.00 Westmere/ClarkDale Core i7-875K 3.6 1.21 336.11 Nehalem/Lynnfield Core i3-560 3.33 1.12 336.34 Westmere/ClarkDale Core i5-760 3.33 1.12 336.34 Nehalem/Lynnfield Core i5-655K 3.4 1.16 341.18 Westmere/ClarkDale Core i7-980X 3.6 1.24 344.44 Westmere/Gulftown Core i7-990X 3.73 1.29 345.84 Westmere/Gulftown Core i7-970 3.46 1.2 346.82 Westmere/Gulftown Core i7-950 3.33 1.16 348.35 Nehalem/Bloomfield Core i5-2500K 3.7 1.49 402.70 Sandy Bridge Core i3-2100 3.1 1.25 403.23 Sandy Bridge Core i7-2700K 3.9 1.58 405.13 Sandy Bridge Core i7-3960X 3.9 1.58 405.13 Sandy Bridge-E Core i5-2400 3.4 1.38 405.88 Sandy Bridge Core i7-2600K 3.8 1.55 407.89 Sandy Bridge Core i7-3770K 3.9 1.66 425.64 Ivy Bridge Core i7-4770K 3.9 1.74 446.15 Haswell [/code<]

      • chuckula
      • 8 years ago

      That post is so ridonkulously awesome on so many levels that I have to just bow to its genius!

    • maxxcool
    • 8 years ago

    FWIW, Thank you TR for the trustworthy review as always.

    • chuckula
    • 8 years ago

    Quick update of importance to the HTPC users: [b<]The Haswell IGP has fixed the 23.976 Hz Bug[/b<] Anand did the tests: [url<]http://www.anandtech.com/show/7007/intels-haswell-an-htpc-perspective/4[/url<] [quote<]Even better news is that the set of display refresh rates obtained with the Haswell system is more accurate than anything I had obtained before with AMD or NVIDIA cards. The gallery below presents some of the other refresh rates that we tested out. madVR reports frame drops / repeats only once every 6 hours or more in the quiescent state.[/quote<]

    • insulin_junkie72
    • 8 years ago

    If you do a lot of video encoding with x264, 17% increase from IB / 28% from SB (according to tests by x264 devs) sounds pretty nice:

    [url<]http://forum.doom9.org/showthread.php?p=1631325#post1631325[/url<]

    • DragonDaddyBear
    • 8 years ago

    Feel free to ignore this mini rant…

    This is great, it really is. And it’s a far cry better than my Q6600. But maybe someone else can rationalize an upgrade excuse for me. I’m not kidding, either. I don’t do any mainstream gaming (my FLEX HD6450 does TF2 just fine enough to play it on Linux), browser works fine, and I’m rocking 3 monitors while working on an IT degree, etc. I’m having a hard time getting excited for these CPU’s when I really don’t have a need, even after ~7 years. (To be fair, it’s liquid cooled, I can OC again at any time, and I have a 1st gen OCZ SSD). Sure, if you did CS work, multimedia editing, playing recent games, etc, I could see a need. But most people just play media on their computers and use facebook. Heck, the X3 455 in the graph powers my MCE “server” for my 360.

    Just a thought, but I think the PC industry needs to find a good reason for the average person to need one of these if they want to get sales going in the right direction. But as it stands, iPads are enough for most people. Low power CPU’s, now that’s something we all upgrade at least every two years because we need to. Look how far cell phones have come in just the last few years. They offer a compelling reason to upgrade. And I can’t even OC anymore on a budget because they K CPU’s lock out all the IT goodies I like to play with (thank God AMD is catching up in performance, because they don’t do that).

    So, anyone want to help me sell upgrade to the wife? I’ve got nothing.

      • chuckula
      • 8 years ago

      [quote<] So, anyone want to help me sell upgrade to the wife? I've got nothing.[/quote<] Have her look at ultrabooks! Tell her they make her look thinner.. I've found that to be effective!

      • peartart
      • 8 years ago

      Birthday gift, which might mean starting to get your wife nicer presents.

      Either that or learn to be happy just having what you need, so you can feel superior to people who buy a new CPU and overclock it, even though their old CPU only hit 100% utilization once in a blue moon.

        • DragonDaddyBear
        • 8 years ago

        I think you all are getting my point. The idea of me upgrading was supposed to be sarcastic. And she has a xps14z. Who wouldn’t want some bigger, better, faster, etc. but why buy a sports car you drive 99% in traffic? I see a shrinking PC market and I think that is why. If I, a computer enthusiast, can find no reason to upgrade, why then would the average person upgrade? At the end of the day it’s a great CPU for those who need it but that demographic is shrinking.

      • paulWTAMU
      • 8 years ago

      why do you want to upgrade? My PC is a means to an end, and right now it works for that just fine. If you had a TV that looked great, was a good size, etc, would you feel compelled to buy a newer model of the same size and roughly the same image quality?

      • Klimax
      • 8 years ago

      I did upgrades on company PCs, because getting replacements wouldn’t be easy and I didn’t want to waste SSDs on them either. So when Ivy was out I took the jump. Replacement cycle is here about five to seven years. (Core i3 general machine, i5 for power where needed.)
      BTW: One of reasons was that I could lower power consumption. (It’s quite expensive here)

      As for machines at home: Mine primary is almost workstation, so different class and notebooks are replaced when failing or such. (generally two years most midrange, but mine is elitebook…)

    • Jakubgt
    • 8 years ago

    I’d just like to inform everyone that I just retired my old Q9450 with an i7-4770K haswell CPU. Man, what a performance boost, especially with this Samsung 840 Pro SSD. For those of you saying the results are disappointing and not worthwhile, please think again. For some of us this is a HUGE step forward in every direction.

    I’m loving every second of Haswell and glad I got it over IB.

    • kroker
    • 8 years ago

    So…

    1. Disappointing performance improvement vs Ivy Bridge if you don’t care about the IGPU
    2. Disappointing overclocking potential, and higher temperature (and power consumption) under load than Ivy Bridge. That small improvement in performance didn’t come for free regarding thermals and power consumption.
    3. Minor but annoying USB3 bug. All the more annoying that it was known many months prior to launch, but nothing was done about it. It’s like buying a brand new shiny car that already has a scratch – definitely first wold problems territory, but still annoying.
    4. DDR3 memory has increased in price by 50% or more (in my country at least)
    5. Next refresh was rumored to be BGA only? if true, not much hope for next year either.

    On the other hand, people with Sandy Bridge / Ivy Bridge can rest easy for quite some time to come knowing that their money was well spent.

    Honestly, I hoped for more.

      • chuckula
      • 8 years ago

      [quote<] 4. DDR3 memory has increased in price by 50% or more (in my country at least) [/quote<] Even if Haswell was slower than a P4 or 10x faster than Sandy-bridge E, it wouldn't affect the price of DDR3 (oh and, that goes for any competitors too).

        • kroker
        • 8 years ago

        I’m talking from the perspective of someone who might want to upgrade, that may include buying a new DDR3 memory kit. It’s not Intel’s fault, but it’s a factor affecting the buying decision.

        • Peldor
        • 8 years ago

        If Intel offers no meaningful improvements there are fewer purchases of new systems needing new RAM. You can debate how much demand falling would affect prices but surely it’s not zero.

      • ronch
      • 8 years ago

      AMD moves things backward with Bulldozer, people complain.

      Intel moves things forward a little bit with Haswell, people complain.

      Make up your mind people. Forward or backward?

        • kroker
        • 8 years ago

        Things haven’t really moved forward for desktops all that much, that was the point. This is a new architecture, this is supposed to bring the biggest performance improvements, and it didn’t. But it’s not only that, Intel could have improved things just by reverting back to fluxless solder instead of TIM for example, if they wanted to. But they simply didn’t want to.

          • ronch
          • 8 years ago

          Because they know we have no other choice apart from AMD’s current offerings.

          If AMD keeps messing up Intel will have a new slogan: “Take it or leave it.”

            • kroker
            • 8 years ago

            It’s simply sneaky malice on Intel’s part, you don’t have to blame it on anyone else.

            • ronch
            • 8 years ago

            Nope, not blaming AMD or anybody. Designing CPUs is rocket science, and AMD is doing what it could to compete. Unfortunately, it’s by no means a guarantee that they can outperform Intel again.

            I won’t say Intel is slacking off because designing CPUs is ANYTHING BUT slacking off. Perhaps it’s Intel not working as hard as they could. But if AMD does manage to pull a rabbit out of the hat again it will force Intel to work harder and move things forward more seriously.

            And ironically, despite the huge differences between Intel and AMD are in terms of resources, products, culture, etc., these two companies have a symbiotic relationship with each other. Intel can’t remain as one cohesive company without AMD, and AMD obviously can’t live without Intel either and continue living off x86.

      • vargis14
      • 8 years ago

      Pats tops of 2600k case, your a good girl your boob’s aren’t even starting to drop yet and I have been hanging on them for over 2 years.

      LOOKS at 560ti SLI setup ahh you two girls are ok too but you might want to get a supporting bra for the next year:)

    • willmore
    • 8 years ago

    I stared at the memory bandwidth chart at PCPER and all that comes to mind is “FFS, AMD, why are you unable to design a competetive memory controller?” The latency and bandwidth are both horrid. This isn’t new, either, they’ve been failing at this for years.

    Maybe there’s hope if they steal the GDDR5 controller from their graphics ASICs to use in Kabini or Vishera, the’ll improve relative to Intel.

      • sschaem
      • 8 years ago

      FX-8350 peak memory bandwidth : 19GBs
      i7-3770k : 28GBs

      EXACT same ram, exact same bus width, but AMD DDR3 memory controller deliver 50% less bandwidth

      No wonder AMD cpu do so poorly in bandwidth limited tasks, but kick ass in compute bound ones (x264, cinebench, ..)

      You cant combine a ‘broken’ DDR3 memory control + horrid L3/L2 cache latency and expect to build an overall winning CPU design.

        • willg
        • 8 years ago

        How much of that is down to pre-fetching though?

          • willmore
          • 8 years ago

          I’d have to say that its performance in many tasks are due to good prefetching and a decent L1/L2 (but not anything past that). Since Phenom, L3 and further has been death.

        • JustAnEngineer
        • 8 years ago

        19 compared to 28 is [b<]32%[/b<] less bandwidth, not 50%. Did the FDIV bug bite your calculation?

          • sschaem
          • 8 years ago

          a) If Intel mem control delivered 1/3 less bandwidth it would equal AMD
          b) AMD need to raise its memory controller by 50% to match Intel

          So let me rephrase instead of changing the math :

          Intel memory controller delivers 50% more bandwidth then AMDs

          19 * 150% = 28.5

      • tipoo
      • 8 years ago

      Err, GDDR5 controllers have notoriously high latency compared to DDR3 (with graphics, their target is obviously bandwidth, and latency suffers because of it). It’s possible they’ll improve it independent of that fact, but just going with GDDR5 controllers alone isn’t an indication of future better latency.

        • chuckula
        • 8 years ago

        A good GDDR5 controller will be fine for feeding the GPU since GPUs get along fine with very high-latency data access patterns (e.g. streaming lots of data in a linear manner). As you pointed out, GDDR5 (in the absence of separate DDR3 channels) will cause all kinds of problems is in the low-latency and random memory access patterns for CPUs.

        Remember RAMBUS back in the ’90s? Their selling points were (for the time) eye-popping theoretical bandwidth numbers, but the latency was such a disaster that regular DDR1 performed substantially better in the real world. There are several reasons why GDDR5 isn’t used everywhere in place of DDR3. The bandwidth is amazing (if you have super-fat bus widths) but there are tradeoffs to everything.

          • willg
          • 8 years ago

          Rumours now say no GDDR5 with Kaveri

          [url<]http://www.brightsideofnews.com/news/2013/6/1/amd-updates-roadmaps2c-shows-28nm-kaveri-for-socket-fm22b.aspx[/url<]

            • chuckula
            • 8 years ago

            Interesting rumor. If GDDR5 is off the table, then AMD may go toward embedded DRAM to keep Kaveri fed. I don’t foresee DDR4 being practical in the late 2013/2014 timeframe simply because the parts aren’t ready and will be quite expensive at first.

            • tipoo
            • 8 years ago

            I read on a comment which provided no source that the reason the Xbone went with eSRAM was that only certain fabs can make eDRAM, and that AMD does not have access to such fabs. Any truth to that? Perhaps they would use eSRAM instead? Which would be fine, even better in fact, since eDRAM needs to wait for refresh.

            • chuckula
            • 8 years ago

            An SRAM cache is another alternative. The trick is that while SRAM gives you excellent performance, its density (6 transitors/bit) lags DRAM (1 transistor + 1 capacitor/bit) by a large margin. Of course, AMD wouldn’t have to make the memory at the same fab where Kaveri is produced, but if they can’t get their hands on supply, then they might have to delay on-package memory for a future update.

          • willmore
          • 8 years ago

          [url<]http://www.pcper.com/image/view/26000?return=node%2F57420[/url<] We're talking over 2x to almost 3x latency increase for AMD. How much worst does GDDR5 need to be than DDR3 to not be *better* than the memory controller in current AMD chips? Normally, an arguement goes "Oh, GDDR5 is 20-30% slower than DDR3 for latency, it's just *so* slow." And that'd be fine. But, 2x-3x? Holy crap! You know what hides latency well? Good L1/L2 and many threads as well as a deep ROB--all things AMD has. If they didn't they'd already be dead, but they manage to beat Intel in a number of workloads. So, there is hope. “Is there not promise of rain?” Ling Tan asked suddenly out of darkness. There had been need of rain for many days. “Only a promise,” Lao Er said. -- Pearl S. Buck Edited: Grammer, spelling, etc..

        • willmore
        • 8 years ago

        Compared to what? A good DDR3 controller like Intel’s or a 3x slower one like AMD’s? Look at that latency chart at PCPER and try not to weep for AMD. It’s a slaughter and there is no excuse for it. AMD must do better. GDDR5 could only be lower latency than that and the 3x bandwidth improvement certainly would not hurt the GPU portion of the thing.

        Edited: missing article.

    • halbhh2
    • 8 years ago

    While a Haswell review is important, in fact most enthusiasts are instead probably wondering how the price/performance of the GTX770 compares to the 7970!

    So we are surfing to find *that* review, and will get to the Haswell review just whenever, in a week or 3. See?

    • ronch
    • 8 years ago

    Quite off topic, but lately I’ve been seeing a new Thumbdown Champ in TR’s article discussion areas. Not gonna name names.

      • NeelyCam
      • 8 years ago

      Abw is a pretty good troll

        • Anonymous Coward
        • 8 years ago

        I have to disagree about that.

          • chuckula
          • 8 years ago

          He lacks that certain.. [i<]je ne sais quoi[/i<]... Probably because all he can do are spew personal insults at his supposed enemies at Intel and anyone on this site who has a functioning brain. Oh, that includes many people who are strong advocates of AMD. If he had just a bit of sarcasm I could see it as satire, but from what I can tell he's completely serious and is probably incapable of mentally processing the fact that his insults are so infantile that it makes Intel look good....

            • abw
            • 8 years ago

            What i lacks mainly is your usual dishonnesty….

            • ronch
            • 8 years ago

            Dude, I think someone needs to put a large water cooling setup on your head. Oh wait, I think LN2 would be better. Or… ah.. maybe something else.

            • abw
            • 8 years ago

            When you declare war to someone only he has the power
            to decide the end of hostilities , that s what happened to
            this poor Chucktrolla….

            • ronch
            • 8 years ago

            I’m not declaring war. How about you?

            • abw
            • 8 years ago

            Was not talking of you but of the inexperienced Chucktrolla…

            • ronch
            • 8 years ago

            So you are declaring war.

            • abw
            • 8 years ago

            Not at all , i m retaliating to the Chucktroll initiated war….

            • ronch
            • 8 years ago

            You know, TR has a built-in spell checker. You can’t miss it, really.

            • abw
            • 8 years ago

            I m all for raw instantaneous spontaneity….

        • abw
        • 8 years ago

        No trolling but integrity, honnesty, straightness without mercy….

          • ronch
          • 8 years ago

          Ok. Uh huh.

          • maxxcool
          • 8 years ago

          to bad your on amd’s payroll…

        • maxxcool
        • 8 years ago

        Works hard for those bit coin deposits from amd…

      • abw
      • 8 years ago

      Are you trying to capitalize on the illiterates discontent.??..

        • ronch
        • 8 years ago

        [quote<]illiterates discontent.??..[/quote<] First, what the hell do you mean by this? Are you using Google Translate? Second, why so defensive? Do you think I'm referring to you? Third, I am in NO WAY attacking anybody. I'm just being observant here, my hot-headed friend.

          • abw
          • 8 years ago

          Intel gullibles gang hunger…..The kind of people whose first thought
          when faced with a lion is how to bring you down so they could escape
          as the cowards they indeed are , like all people who systematicaly
          side with whoever seems the bully even if he s an asshxole that gives
          them a slap from time to time when he wake up disgruntled….

    • jamsbong
    • 8 years ago

    Interesting to see that it overclocks worse that ivy bridge.

    techpowerup managed 4.2ghz only
    [url<]http://www.techpowerup.com/reviews/Intel/Core_i7-4770K_Haswell/3.html[/url<] xbilabs 4.4ghz but very very hot [url<]http://www.xbitlabs.com/articles/cpu/display/core-i7-4770k_12.html#sect0[/url<] some benchmarks of overclocked haswell vs ivy [url<]http://www.techpowerup.com/reviews/Intel/Core_i7-4770K_Haswell/5.html[/url<] note that the extra bus speed helped with the winrar compression speed. otherwise, 4.2ghz haswell = 4.5ghz ivy Ugh! gonna save my money for upgrades other than the CPU. hehe...

      • chuckula
      • 8 years ago

      Uh…. did you bother to read some of those links you submitted as being a “dissapointment” for Haswell?

      Edit: The first link with the “only 4.2GHz” was clearly referring to an attempt to only overclock using BCLK manipulation, which is not the only way (or even primary way) to overclock Haswell. They didn’t change the multiplier at all in that test!

      The third link clearly shows Haswell at 4.2GHz with a noticeable lead over Ivy at 4.5GHz. If this is what counts as “disappointment” for Intel, then I can think of a whole bunch of AMD engineers that are praying for steamroller to be a “disappointment” too…

    • Arclight
    • 8 years ago

    Look how close that Core i7 4950HQ (a mobile CPU) is to the Core i7 4770K (a desktop CPU).

    That’s pretty surprising. Not only that but in this bench it ranked 7th, better than CPUs like 2700K or FX 8350.

    • xeridea
    • 8 years ago

    Looks like they haven’t fixed temperature issues, still using $0.01 smear of mayonnaise inside chip. Less power, but runs hotter…. just like 2600k vs 3770k. It runs hotter than pretty much every other chip on the market, even the ones with considerably higher TDP, and not by a little, by a lot.

      • smilingcrow
      • 8 years ago

      “But at least it’s fresh puke”

      One for the Fawlty Towers fans. 🙂

    • abw
    • 8 years ago

    So USB3 wasnt an issue according to the usual intel apologists.??…

    Most sites downplayed the thing , of course , money talk…

    [url<]http://www.tweaktown.com/news/30730/haswell-and-usb-3-0-seeing-big-issues-with-usb-drives/index.html#SFefabMRoMesyMym.99[/url<] 14 USB3 devices out of 22 had serious issues... Edit : A saturday release...as if it should have been better kept unnoticed with people just living on the massive hype transfusions we experienced with this wonder....

      • Bensam123
      • 8 years ago

      Interesting… Most people were playing this down as it wasn’t going to be a issue…

        • abw
        • 8 years ago

        Telling about some sites integrity that advice buying this crap…

        • danny e.
        • 8 years ago

        It’s not much of an issue. A minor annoyance at worse.
        abw is either trolling or an idiot that didn’t read anything other than the headline of the link he-she posted.

        [quote<] Connected USB 3.0 devices based on one or more controllers - it's unclear exactly which one - [b<]will briefly disconnect and reconnect when a PC wakes up from standby (S3).[/b<] This can pose a problem if you had a file open that's located on a USB 3.0 storage device, the software will tell you that the file cannot be found after your computer comes out of standby mode. Intel sees the problem as annoying but not a deal-breaker, and we tend to agree. We're able to reproduce the problem, but can't really think of many scenarios in which it will be huge problem. All you have to do is to open the file again, that's it. No data will be lost due to this bug. [/quote<] [b<]The fact that TweakTown uses "Big Issues" in the headline proves they are not worth reading - since the article they link to as their "proof" states exactly the opposite.[/b<]

          • tipoo
          • 8 years ago

          My drives all disconnect when resuming from standby anyways, I’m not sure what’s unique about this?

            • abw
            • 8 years ago

            And my drives never disconnect, even with 10 years old laptops…

            But soon we ll be explained that disconnecting is progress….

            • maxxcool
            • 8 years ago

            yeah too bad your amd usb3 is slower than usb2 in most tasks that are not sequential.

            • derFunkenstein
            • 8 years ago

            Depends on the USB controller in the PC more than it does the drives. I can tell you from experience that if OS X continues to gripe when something gets disconnected, there will be lots of annoyed Mac owners every time a computer wakes from sleep.

            Still, as danny e. said, it’s an annoyance more than ZOMG BIG PROGLEMZ

          • abw
          • 8 years ago

          So intel see it as annoying but not being a deal breaker…

          The times have comes where the customers is treated
          as if he was the supplier , in a full reverse of the usual
          commercial behaviour , i ll tell you straight , intel consider
          consumers as a clueless herd that is to be happy
          to be milked up to the last drop…

      • abw
      • 8 years ago

      Where are all the suckers that slavishly parroted intels propaganda..???

      Mediocrity at its worse..

      Edit : Neelycam must be in complete disarray ,
      looking desesperatly for more benches , who knows ,
      perhaps there is one from a Groenland site that show
      unexpectedly low temps….

        • NeelyCam
        • 8 years ago

        No – NeelyCam has a life and sometimes does things during weekends that don’t involve browsing the web nonstop

          • abw
          • 8 years ago

          There are occurences that are not random variables….

            • NeelyCam
            • 8 years ago

            Your prose makes no sense

      • chuckula
      • 8 years ago

      [quote<] 14 USB3 devices out of 22 had serious issues...[/quote<] Uh no... 0 USB3 devices out of 22 had [b<]serious[/b<] issues. 14 of them had a minor annoying issue in weird situation. Whoopee doo. I've done tests just for fun, and on several machines at work (including both Intel and even one of your precious Trinity chips), the standard operating procedure seems to be that the USB drives get disconnected and reconnected after returning from a sleep state... which is exactly how to fix the supposedly "serious" issue here.

        • abw
        • 8 years ago

        lolz^10….here a quote from the site :

        “They’ve done some very, very extensive testing and confirmed there are USB 3.0 issues with the 8-series chipset. They tested 22 USB 3.0 drives, with 14 of them experiencing issues and were unable to stay connected when a Haswell system woke back up from standby.”

        As expected the intel stooges turn a blind eye , going as far as to explain
        to other stooges that this is the normal behaviour of USB3 devices….

        Pathetic…..but what to expect from frustrated half brains that think
        that downvoting someone will help to solve a technical issue….

          • swaaye
          • 8 years ago

          Maybe if you were to express yourself without emotional outburst and insults, people would read your posts more thoughtfully.

            • abw
            • 8 years ago

            The outbursts are for whom is of bad faith ,
            according to the downvotes poll about 20 peoples.

      • abw
      • 8 years ago

      The intel suckpuppeters in full panic mode….Truth hurt…

        • peartart
        • 8 years ago

        [url<]http://www.youtube.com/watch?v=9AlH2oYedfk[/url<]

          • abw
          • 8 years ago

          [url<]http://www.youtube.com/watch?v=jX70njhUKII[/url<]

      • trackerben
      • 8 years ago

      I’d like to know what’s your lead on certain points re intel.

      What do you think is the reason Intel is pulling out of the branded consumer board market?

      What’s Intel’s marketing rationale for applying Core/Pentium/Celeron tiering to their mobile cpus/apus?

        • abw
        • 8 years ago

        With Socs there s almost no added value in the MBs
        since everything is on the chip.

        The pentium/celery rebranding is to give a crappy product
        the aura of better perceived brand names , this is pure
        marketing spin with the obvious aim to mislead
        the milk cowery….

          • trackerben
          • 8 years ago

          Demand would have to be greater for x86 SoCs vs. x86 motherboards. Even if manufacturing costs are lower for x86 SoCs, without the volume these aren’t going far. Unless you are comparing all ARM & x86 & hybrid SoCs sold vs. all x86 boards sold?

          On the mobile rebranding, don’t you think Intel is doing this because they are focusing on the massive ultramobile/smartphone markets? Isn’t this Intel tiering marketing according to production because they see their ultra low TDP lines going mainstream in a bigger way than “traditional” cpus?

    • chuckula
    • 8 years ago

    In honor of Bensam123 being the first poster in this whole thread to mention core parking, I give you some Isaac Hayes:

    [quote<] Who's the x86 CPU that's a crunch machine to ALL the notebooks? HASWELL! You AMD right! Who is the core that would turbo-boost for his brother cores? HASWELL! Can you OC it? Who's got cores that don't cop out, when there's benchmarks all about? HASWELL! Right On. They say this chip Haswell is a power-efficient MUTHA... [b<][i<]PARK YO CORES![/i<][/b<] But I'm talkin' about Haswell! Then we can bench it! He's got complicated cores, but no one understands him like his uncore. Intel Haswell![/quote<]

    • tipoo
    • 8 years ago

    Woohoo, this was everything I suggested on that post asking for benchmark suggestions. Old processors going generations back, including C2D and C2Q which imho are important as they would make more likely upgrade candidates than Ivy or even Sandy.

    • chuckula
    • 8 years ago

    To all you Vader-Haters out there: You’re Right!

    Well… lemme put it to you this way. What is Haswell on the desktop? An overclocked notebook part? Yeah basically. That’s also why Haswell’s overclocking is about the same as Ivy’s (unfortunately). Before all you AMD fans start gloating about that, look at what it does to the overclocked server-grade FX-8350 while having a substantially smaller power envelope.

    Some people say that this is due to the lack of competition. Well, Intel is under intense competitive pressure and Haswell is designed for competition… in ultrabooks and tablets. The desktop isn’t the focus of Haswell, and it shows from the design of the architecture all the way down to the silicon process which is clearly tuned for low-power low voltage. While I was hoping for 5GHz overclocks to be much more achievable, it’s pretty clear that the silicon process has been tuned for lower-voltage lower-power performance. The ironic twist is that Haswell is actually a better low-voltage overclocker than Ivy [b<]but[/b<] when you pass a certain voltage level, the temperature spikes exponentially and you hit a thermal wall. If Intel really wanted to take the desktop seriously, they would either get out a 6-core version of the desktop chips using the same LGA-1150 platform or rework their process for higher-power consumption and higher-clocks (or do both). Unfortunately, that is very unlikely to happen since there just isn't the market demand for that type of chip to justify the costs.

      • Krogoth
      • 8 years ago

      Intel has a performance desktop platform. They want you to look at LGA2011 line-up which is going to get IB-E treatment in the foreseeable future.

      The “desktop” IB and Haswell chips are just silicon that fail ULV testing for portables, but otherwise work fine.

    • Bensam123
    • 8 years ago

    You’re such a tease Geoff. Way to drum up the hype for the article. 😛

    And all the steam goes out of the haswell balloon. SB-IB had a bigger performance increase and that was still close to nothing.

    Hopefully the OCing results will be more promising then this. Definitely looking forward to your new benchmarks.

    Curiously though, unless I’m mistaken I don’t see a i5-3570k, which is like ‘the’ most popular IB processor… The i5-2500k is listed, which is it’s predecessor… I know it’s sorta weird to scream ‘more more’ in this case, but that seems like something that should’ve been included if there was only like 8 processors, let alone a gajillion.

    Also, is core parking disabled?

      • derFunkenstein
      • 8 years ago

      If you’re filling balloons with steam you’ve got bigger problems than a disappointing desktop CPU.

        • Bensam123
        • 8 years ago

        I suppose so… then again people have been known to get steamy.

      • Krogoth
      • 8 years ago

      Heresy!

      Nobody here talks about core parking! 😉

        • Bensam123
        • 8 years ago

        Shh… Let the word spread before the inquisition comes out for heads.

      • chuckula
      • 8 years ago

      [quote<]Curiously though, unless I'm mistaken I don't see a i5-3570k, which is like 'the' most popular IB processor...[/quote<] According to you the 3570K has zero improvement over the 2500K so TR just did you a favor by combining the charts! In fact, since the 4770K has zero improvement over Ivy, Ivy has zero improvement over Sandy, Sandy has zero improvement over Nehalem, and Nehalem has zero improvement over the core 2, TR could just simplify things by comparing a Core 2 to the FX-8350!

      • Squeazle
      • 8 years ago

      It may not be huge, but take a look at the single core performance… it will be interesting to see that in with the efficiency.

        • Bensam123
        • 8 years ago

        Yeah, obviously this is a initial look. The new benchmarks and what not should be interesting regardless of how well HW performs in them (some of the new ideas were pretty exciting and refreshing). I’ll just hold my breath till we see more, but HW doesn’t look all that stellar (almost lackluster in some cases).

      • travbrad
      • 8 years ago

      [quote<]And all the steam goes out of the haswell balloon. SB-IB had a bigger performance increase and that was still close to nothing.[/quote<] Yep this is kind of what we all expected though. All indications have been that Haswell's main focus is on idle power consumption and GPU performance improvements. On one hand it's disappointing that we aren't seeing bigger performance improvements in desktop CPUs, but on the other hand I am feeling more and more like my 2500K for $150 in October 2011 was one of the best purchases I ever made. These Cinebench differences are about the biggest difference you'll see as well. In real world gaming there is virtually no difference.

    • Laykun
    • 8 years ago

    Looks like my good old 980X will be relevant for another year or two.

      • indeego
      • 8 years ago

      I have that processor too, but our (very expensive) processor gets trounced in other benches fairly regularly.

    • ermo
    • 8 years ago

    AMD has mentioned something along the lines of (up to) 30% performance improvements for SteamRoller vs. PileDriver in loads that lend themselves to parallelization. If we assume that this is within the same 125W power envelope and that AMD is not lowering their clock frequency, the optimistic, best-case scenario score reads 1.3 * 6.93 = 9.09.

    For a more conservative 15% performance improvement estimate, the math says 1.15 * 6.93 = 7.9695, which is ‘in the ballpark’, at least for rendering, which parallelizes well.

    The 30% improvement is probably not going to happen across the board, but interesting perspective nonetheless…

      • abw
      • 8 years ago

      30% is in respect of the first FX iteration , not PD ,
      in this latter case expect 20%.

    • kamikaziechameleon
    • 8 years ago

    So from what I see not a giant leap on performance in any sense of the word. Time to see if they deliver on the promise of power savings.

    • yammerpickle2
    • 8 years ago

    Meh. Unless they overclock really well this is really not much faster than my overclocked P-2600K from a few years ago.

      • travbrad
      • 8 years ago

      All of the overlclocking results I’ve seen so far seem to indicate it’s worse at overclocking than Sandy Bridge, mainly due to extremely high temperatures when you increase the frequency and voltage.

      Xbitlabs for example were seeing temps of around 90C with a 900mhz overclock (4.4ghz). My 2500K with a 1200mhz overclock (4.5ghz) runs at around 60-65C under full load…I would be scared of running a CPU anywhere near 90C.

    • Jakubgt
    • 8 years ago

    When will consumers be able to purchase these from retail stores? I can’t wait to finally retire my Q9450 and get my hands on Haswell

    • derFunkenstein
    • 8 years ago

    Wow, man. Just from what I’ve read around the internet it’s only worth upgrading if you have a Phenom II or a Nehalem (or older) system. If you’ve built with an Intel quad since 2011, keep on enjoying it.

      • chuckula
      • 8 years ago

      Pretty much, but then again, 2009 – 2013: 4 years, and Haswell has some noticeable performance advantages over first-gen Nehalems that were top of the line back then.

      386 – 486: 1985 – 1989 (4 years)
      486 – Pentium: 1989 – 1993 (4 years)
      Pentium – Pentium II: 1993 – 1997 (4 years)
      Pentium II – Pentium III: 1997 – 1999 (2 years, although the P3s were not a huge generational leap over P2s)
      Pentium III – Pentium 4: 1999 – 2001 (4 years from PII –> P4, and we all remember about Willamette’s questionable performance compared to later P3s)
      Pentium 4 – Core 2: 2001 – 2006 (5 years! Oh, and remember that the Core 2 seemed so much better simply because the Prescott P4s were that bad)
      Core 2 – Nehalem: 2006 – 2008 (only 2 years, but Nehalem was very-high end and Core 2 continued with strong popularity in the mainstream until the socket 1156 came out in the latter half of 2009).
      Core 2 – Lynnfield: 2006 – 2009 (3 years)
      Lynnfield – Sandy Bridge: 2009 – 2011 (2 years)
      Sandy Bridge – Ivy Bridge: 2011 – 2012 (1 year)
      Ivy Bridge – Haswell: 2012 – 2013 (1 year)

      Notice anything about the cadence of new product introductions? Should we really be shocked that Haswell is such a “disappointment” for a 1 year update over Ivy?

        • derFunkenstein
        • 8 years ago

        For a while there, though, because Intel and AMD were so embattled from 500-1000MHz we literally saw CPUs double in speed in the course of a year, and double again over the course of 2 more.

          • Klimax
          • 8 years ago

          Don’t think this is related to competition, but to increasing complexity. (Like GPUs) We’d need new algorithms to have significant jump.

            • derFunkenstein
            • 8 years ago

            His point was that we should expect small progress and over the course of 4 years it aggregates into worthwhile improvement. It wasn’t that long ago that this wasn’t the case and it was due to (roughly) linear scaling with the clock speed. Look at the jump from Core 2 to Nehalem, even.

            • Klimax
            • 8 years ago

            Because arch still had a lot of to improve. After jumps there is not much left. Intel engineer even said more or less so. (Reddit AMA) And when asked about popf he asked back, what he can make slower. By that we are definitely in trade-off phase, until we get some major change.

            ETA: Same will hit ARM, btw.

      • paulWTAMU
      • 8 years ago

      My Phenom II X4 still does OK for me even…

    • donkeycrock
    • 8 years ago

    I thought you guys were going to stop doing synthetic bench marks.

    • clone
    • 8 years ago

    Xbitlabs has an early review including power consumption and some overclocking.

    the conclusion was that Haswell is destined to fight in the mobile arena and that it’s come at the expense of desktop in a variety of ways including excess heat and power consumption.

      • abw
      • 8 years ago

      Lol….fans are also heating , even boiling….

      But , but….where is IntelCam..???….

      • smilingcrow
      • 8 years ago

      It seems to be another disappointing chip for over-clockers but for mainstream users it is clearly a better platform overall.
      I do wonder if Intel are deliberately using cheap TIM to encourage people to move to LGA1366! Or is there a different cause this time? TDP!

        • Krogoth
        • 8 years ago

        Intel is trying to save some more on production costs. TIM is only problematic if you trying to push the silicon on air (4.8Ghz or more).

        At that point, most extreme overclocker types prefer to remove the IHS and rely on direct contact.

          • smilingcrow
          • 8 years ago

          Aren’t the savings trivial though? If they wanted to push the K series as really good over-clockers they wouldn’t save a small amount on TIM for the K series and they charge a premium anyway.
          As for the extreme brigade that remove the IHS I doubt Intel even consider them and neither do 99% of the market.

            • Krogoth
            • 8 years ago

            Blame bean counters on this one. The will find a way to cut corners and costs as long as the product works without too much fuss.

            • smilingcrow
            • 8 years ago

            But for a K series chip there will be a fuss; hard to say if affects sales.

            • Krogoth
            • 8 years ago

            You realize that binned happens after silicon and packaging are put together? Intel is using TIM for their entire line-up to save production costs (even the SB-E chips and soon IB-E chips have it).

            K series is silicon that fails that ULV testing, but has generous overclocking headroom when you throw volts at it. Intel alters the microcode to unlock the chip, but disables VT-T support (marketing reasons to justify existence of Extreme Edition chips).

            The whole TIM thing only affects extreme overclockers who don’t have the stomach to remove IHS and do a direct contact.

            • Derfer
            • 8 years ago

            E chips do not have TIM as their TDP is above 95w. Everything above that line is soldered.

            • smilingcrow
            • 8 years ago

            No I didn’t know so thanks for the info.

          • abw
          • 8 years ago

          It can be for whatever reason but surely not costs…

          What is the saving .?..at 1cts/CPU that would save them
          3 millions $ while it render the CPUs slightly less reliable.

          Better explanation is that this will direct the enthusiasts
          to buy the old inventories , SB/IBs , that overclock better….

            • Krogoth
            • 8 years ago

            CPUs are more than reliable. It is only the extreme overclockers who are raising a storm over this.

            • abw
            • 8 years ago

            You re assuming short term while i m talking years..

            Reliability over years decrease exponentialy with increased
            usage temps….

            • clone
            • 8 years ago

            motherboards tend to fail a lot sooner than cpu’s even when running at stock speeds.

            not since the Prescott era has power or heat been a significant issue and remember AMD’s FX cpu’s are fine as are the abundance of 125 watt TDP cpu’s that have been sold in the past.

            • abw
            • 8 years ago

            Agree on the weak reliability of MBs but CPUs as FX or other
            expensives series are bought by people who take care of their
            PC heatspreaders and casing contrary to the vast majority
            of users , just take a look inside PCs that were used
            just three or four years , likely that the CPU run very hot
            thanks to vast amount of dust stocked everywhere in the case…

            • clone
            • 8 years ago

            abw AMD and Intel design their product to work under the lower end of the love meter and to be honest if you believe that more than 5% of FX 8xxx owners “take care” of their cpu’s you are kidding yourself, I’d wager far more than 5% are actively overclocking their cpu’s and running them too hot so long as they are stable.

            the cpu’s cost less than $200 and that’s the only appealing thing about them, to even hint that more than 5% of AMD FX 8xxx owners are wasting time every 6 months cleaning out the inside of their pc’s is highly optimistic.

            computers are an appliance, pre 2000 they were worthy of consideration but in the age of the $500 pc….. they are meant for abuse and condemnation if they don’t survive it.

            • abw
            • 8 years ago

            That does make sense for the higher incomes areas but 500$
            is a hefty sum in quite a large part of the world , often way
            more than a monthly salary.

            • clone
            • 8 years ago

            high end doesn’t sell well in areas where $500 is the monthly income.

        • briskly
        • 8 years ago

        Most of the problem doesn’t seem to be that Intel used a cheap TIM, it’s that used any grease at all, and that they used an adhesive to stick the heatspreader to the chip. Scrapping off the goop and reapplying anything would probably work fine, so long as the silicon makes good contact with the heat spreader.

        See [url=http://forums.anandtech.com/showpost.php?p=34053183&postcount=570<]here[/url<].

    • puppetworx
    • 8 years ago

    The award for Best Tease on the Work-Safe Internet goes to… 😉

    Can’t wait to read the full review now!

    • internetsandman
    • 8 years ago

    Is there gonna be an updated version of Sandy-Bridge-E coming out soon, perhaps Ivy/Haswell based? I’m on a 2600K right now and while Haswell is a fair bit faster in this one benchmark it’s not enough to warrant an upgrade for me

    I know this is just one benchmark out of many, and I have high hopes for the architecture but if the comparisons in other benchmarks mimic this then Id rather jump to the higher end E versions of an older architecture rather than this

      • smilingcrow
      • 8 years ago

      Ivy Bridge E is due Q3 or Q4 2013.

        • HisDivineOrder
        • 8 years ago

        Soon, they’ll have to be THREE generations ahead before they decide to update that platform with yesteryear’s architecture…

    • abw
    • 8 years ago

    As i expected it , decent at low power usage and disapointing
    as a DT part.

    The usual fans must feel robbed of their 20-30%
    expected and ultra hyped IPC gains…

      • smilingcrow
      • 8 years ago

      20 – 30% IPC gains!
      Give me a pint of what you are drinking although on second thoughts 200 – 300 ml should be enough for me; I still want to be able to crawl into bed.

        • abw
        • 8 years ago

        Now that the real numbers are here everybody is amnesic….

        Really , no need of any brewage…..

          • smilingcrow
          • 8 years ago

          Well I’m going to have a brew anyway. 🙂
          I’m curious where those projected 20 -30% IPC gains are from as that would be exceptional and something that I and many others would have been very sceptical about. I was thinking that on average the gains would be in single digits.

            • abw
            • 8 years ago

            You are thinking single digits since some chinese site did leak
            the perfs as well as THG with intel s secret approval , as a mean
            to cool down high expectations that would have been detrimental
            if not reduced before launch.

            Browse the sites in late 2011 and early 2012 , including TR ,
            even intel was still touting 15% on early 2012….

            • smilingcrow
            • 8 years ago

            I was thinking single digits because that is normal for Intel these days and has been for a while now; above that would be related to new instructions etc.

            I don’t track predictions from 18 months prior to release even from the manufacturers as even they don’t really know at that point. Predictions from 3rd parties are worthless to me.
            But you said 20 to 30% and now you are talking 15%!

            • abw
            • 8 years ago

            “Theses days” as you call them started once IB was launched
            and didnt provide the expected gains over SB so the crowd,
            that once thought about 10% for SB–>IB , started to realize
            that the far more huge gains in the waiting of HW could
            possibly be of the same vein , although mandatory better.

            The 20-30% figure , thanks to AVX2…., looked to be overly optimistic
            so all the estimations started to slip down but still , no one one year
            ago was expecting less than 20%.

            • smilingcrow
            • 8 years ago

            I haven’t expected more than a 10% IPC gain since Conroe (was that in 96!) as it has been clear that it is not easy to do more than that any more.
            I don’t consider new instructions such as AVX2 as being a part of IPC gains as they typically only impact a small range of applications.

            ‘no one a year ago was expecting less than 20%.’

            It must be good to be able to read everybody’s mind or is it that ability that is causing you to be so confused?

            • abw
            • 8 years ago

            I beg to differ , new ISAs as AVX , FMA increase IPC.

            It happens that the previous ISA extensions was mainly AVX1 wich is dedicated
            to FP computations and can double the throughput in FP , but most of the usual
            softs use integer and logical ops processed in integer pipelines execution units ,
            i.e ALUs , theses are the most visible perfs improvements at the average user level ,
            so integer IPC matters a lot and that s where all the intel fan crowd got it wrong ,
            they thought that since AVX2 can double the integer IPC they would see
            dramatic perfs improvements in usual softs , but alas , only part of a code can
            be paralelised , moreover with integer ops , so the improvement could only
            be incremental , provided there s softs that make use of this new instructions,
            that is to say , none.

            Now the crowd is in full denial mode in respect of its past expectations….

            • smilingcrow
            • 8 years ago

            Sane people are only interested in real world IPC gains and leave the fantasy IPC gains to the nutters.

            • abw
            • 8 years ago

            Sane people and hardware enthusiast do not match well ,
            and even not at all when it s intels pigeons….

            • smilingcrow
            • 8 years ago

            “intels pigeons.”

            Since the Grouse season is still over 2 months away I’ll stick with pigeon pie for now; you enjoy your humbly oblivious pie.

            • abw
            • 8 years ago

            Well , theses are stool rather than pie dedicated…

            As for the pie i still never tasted this quite savage plate but who knows.?..

            • smilingcrow
            • 8 years ago

            Stool pie sounds right up your street. You talk a lot of ‘stool’ anyway.

            • abw
            • 8 years ago

            There s lot of pigeons by there…Btw , in french a pigeon
            is someone gullible who is easily ripped off by unscrupulous
            and dishonnest merchants , this apply perfectly to the usual
            intel siding consumer.

            • smilingcrow
            • 8 years ago

            What a load of merde.

            • abw
            • 8 years ago

            Suits perfectly the hype we endured for two years…

            • maxxcool
            • 8 years ago

            to bad at 4.5 ghz were at 2.0x on the single thread performance index and your … oh yeah under the table.

            • abw
            • 8 years ago

            You re coming late to the party..

            But maxxxxwxxxxxxxxxx , since you are kkkool
            you can take all the remaining boones…….

            • maxxcool
            • 8 years ago

            the bones of amd?, sure, my dog will gnaw on them’

            • abw
            • 8 years ago

            I hope you get a few upvotes for your imagination skills………..

            • green
            • 8 years ago

            [quote<]Browse the sites in late 2011 and early 2012 , including TR , even intel was still touting 15% on early 2012....[/quote<] 15%? compared to what? the latest chip out in late 2011 and early 2012 was sandy bridge you could say ivy bridge was early 2012, but given a late april launch that [b<]really[/b<] stretches the definition of early don't the marketers do this every time? X is Y% better where no compartor is given for the baseline of improvement been happening for decades now either way looks like i'll be skipping yet another generation of chips

        • chuckula
        • 8 years ago

        I seem to recall that I predicted abw’s behavior a few weeks ago.

        Not one person at Intel ever claimed IPC gains like what abw is spouting.
        Not one of the supposed Intel fanboys on this site ever predicted gains like that.
        But there is one and only one person who did… abw… so the only logical result is that abw must be the fan who feels robbed!

          • abw
          • 8 years ago

          We all know that you are unmatchable for predictions….

          An advice , although you re already following it :
          To be always right , predict all and its contrary…

      • abw
      • 8 years ago

      The downvotes are a direct measurement of the fans frustration…

      And it is high , even higher that the IPC expectations , neverless
      all the suckers will again be ripped off 350$ , they just cant
      resist their slave driver….

        • Klimax
        • 8 years ago

        In this case it’s measure of your wrongness and fanboyism, because you’re pulling hell lot out of air.

        Oh, and hell of projection.

        BTW: Intel never promised for Haswell larger increase then low tens in best cases.
        (And experts were in the same camp – Like David Kanter of Real World Tech)

          • abw
          • 8 years ago

          Actualy hateful suckers loves to suck and hate..

          Since they are unable to be critical with their ripping off master
          they will forcibly be agressive with whoever remind them their mediocrity..

          Hypewell was highly proped up by the enthusiast crowd , it was assumed
          to be a giant leap but the numbers are just what they are and this wonder
          happen to be barely better than a lemon in respect of previous gen.

            • Klimax
            • 8 years ago

            Hypewell???

            May be in your mind. Show the hype, because currently you’re spouting not even wrong nonsense.

            • abw
            • 8 years ago

            As said , everybody became amnesic once the early numbers were shown.

            When SB was released people expected the same bump
            with IB and an even bigger one with Hasfail , just go through
            the enthusiast sites.

            Things started to cool down once everybody saw IB small
            bump that was telling for future itarations.

            • Klimax
            • 8 years ago

            That’s hell of dodge and goal post moving. And still nothing to show…

            • abw
            • 8 years ago

            Isnt it an Hashell thread.?..

            • derFunkenstein
            • 8 years ago

            Get your troll-names right. It’s HASBEEN!

            • abw
            • 8 years ago

            Given the issues , isnt it rather HUSBEEN 3rd of the name..?…

            • derFunkenstein
            • 8 years ago

            Is…is that a stab at humor? I can’t tell.

            • abw
            • 8 years ago

            Read USB3 issue….

            • NeelyCam
            • 8 years ago

            Still weak

            • abw
            • 8 years ago

            Hasweak..??..

            • maxxcool
            • 8 years ago

            hahaha oh snakey… really .. name calling ?

            • Klimax
            • 8 years ago

            Yes. Still didn’t provide anything you were asked.

        • ronch
        • 8 years ago

        Oh… you sure can’t get it, can you?

          • abw
          • 8 years ago

          Being downvoted by the intel milk cowery is raging…..

          If only it was their PCs that are voting , at least they
          have a non zero IQ….oh , wait…..

      • ModernPrimitive
      • 8 years ago

      I don’t recall ever reading that Haswell would deliver large gains in IPC. Everything I have read pointed towards better IGP and lower power and it seems to deliver very well from what is shown so far. Intel is smart in catering to mobile devices. AMD has moved in that direction also. I can’t help but wonder how many resources Intel would have left towards a more powerful desktop part had AMD remained more competitive. Another 6 or 8 core chip sans IGP would have been nice. Nicer yet would have been a scenario where both companies had a high end part for us to choose from…… I’m not one to yell the PC is dead, but the landscape has changed dramatically the past couple of years.

        • abw
        • 8 years ago

        It s still an outdated product in respect of the current market,
        set apart its IGP it doesnt adress the current demand ,
        it was simply designed well before the current market
        evolution was really sighted.

        It may well satisfy the enthusiasts but Bail trail/atom
        will be far more in line with the times , we are witnessing
        a violent change in the market and i must admit that few
        saw it coming as rapidly as it s occuring.

        • Krogoth
        • 8 years ago

        There’s already silicon that has more cores and cache while axing the IGP. It is called LGA2011…

          • abw
          • 8 years ago

          Right but it s not even marginal , perhaps 0.1% of the market
          the best weeks….

          Edit : by current market i mean the current tendency
          that is geared toward cheap devices , PCs included.

      • maxxcool
      • 8 years ago

      we all new this was intel’s piledriver update…

    • sschaem
    • 8 years ago

    So in this absolute best case senario the flagship haswell is not even 2x faster then my q6600…

    Now that’s unfair because I’m comparing overclock to stock, but still. 6 years have past between the two model.

      • accord1999
      • 8 years ago

      Well, if you overclocked Haswell, then it would easily be more than twice as fast, just as my 2600K@4.8 was more than twice as fast as my old Q9550@3.8.

      And the last two years have been slow in progress, but the previous 4 years were impressive in terms of performance and performance/watt increases for CPUs.

      • Kaleid
      • 8 years ago

      I gotta agree, this is quite unexciting. Imagine if GPU speed increased this slowly..

    • smilingcrow
    • 8 years ago

    What a great chart. I’d like to see a version scaled to show single threaded performance per GHz.

      • Anonymous Coward
      • 8 years ago

      Yeah, this would be awesome.

      • Bensam123
      • 8 years ago

      Why would that matter when ghz really doesn’t matter at the end of the day and hasn’t for years?

      I suppose Intel would win it every single time, but besides that, what would it show us? It’s not like all the chips operate at the same frequency.

        • smilingcrow
        • 8 years ago

        “It’s not like all the chips operate at the same frequency.”

        Thanks Colombo; one for the older readers. 😉

          • Bensam123
          • 8 years ago

          I think you missed the point unless your OPing post was a joke I took the wrong way. :l

          If a chip is rated to run at 4ghz and another 3ghz and they have the same performance, why would it matter if there is a difference in frequency? They’re simply designed to run at different speeds to get performance… Hypothetically this would show us that the 3ghz chip run at 4ghz would be faster, but the point is it isn’t capable of doing so, that’s why it runs at 3ghz and not 4ghz.

          This is like saying ‘we should reorder chips in order of ghz instead of performance!’, which also would have no meaning.

            • smilingcrow
            • 8 years ago

            It’s purely an academic interest in what the IPC per GHz is.

            • Firestarter
            • 8 years ago

            not only that, but it also helps to compare overclocked configurations

            • Bensam123
            • 8 years ago

            IF they all OC’d the same amount, all else being equal (which it isn’t).

            • Firestarter
            • 8 years ago

            Well arguably the overclock ceiling is very similar for Sandy Bridge, Ivy Bridge and Haswell, 3 generations that won’t do 5ghz+ without significant cooling and tinkering.

    • ssidbroadcast
    • 8 years ago

    Where would an i7-3517U ultrabook cpu fit in there?

    • DeadOfKnight
    • 8 years ago

    +10% IPC, +15% multi-threaded vs SB and half that vs IB. Not bad, but nothing to write home about. The only real reason to upgrade from Sandy Bridge is if you need the updates that come with the Z87 chipset. However, this would be a decent upgrade for the first generation Core-i processors and a great boost in performance for anything older.

      • chuckula
      • 8 years ago

      Oh, I wouldn’t recommend upgrading to Haswell from Sandy/Ivy unless you are really doing it for the IGP or have an extreme need for a very specific Haswell feature like AVX2 (which will become more useful over the next year as software gets updated).

      I’m about to upgrade to Haswell from a Core 2 E8400… I’m pretty sure I’ll move up that TR graph by a pretty big leap when the time comes.

        • DeadOfKnight
        • 8 years ago

        What troubles me is that these improvements are clearly on a curve, with diminishing returns. We’re slowing down dramatically. The good news is if you upgrade now you likely won’t have to again for a very long time.

          • chuckula
          • 8 years ago

          For traditional software that isn’t multithreaded and isn’t taking advantage of things like AVX you have a point.

          However… for other types of software including OpenCL, there are some pretty major performance gains. Take a look at the 47/55 watt Iris Pro parts in OpenCL that Anand tested: [url<]http://anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/17[/url<]

            • DeadOfKnight
            • 8 years ago

            Even multithreading is showing diminishing returns, but you’re right. The software hasn’t yet caught up with the capabilities of these chips, particularly when it comes to utilizing new instruction sets. I’m no programmer, but I have heard much excitement about these more “latent” improvements in other circles. We have yet to see what actually comes of it, but I remain hopeful.

            • chuckula
            • 8 years ago

            One thing to keep in mind: Haswell’s peak floating point throughput is [b<]quadruple[/b<] that of Nehalem on a clock-for-clock basis... but there's a catch. To get all that throughput your code needs to be able to take advantage of AVX2. That's why all these benchmarks we are seeing with only moderate performance improvements are only part of the picture. The legacy applications are still important and they still need to have good performance, but there is definitely potential for better performance with the right optimizations.

            • DeadOfKnight
            • 8 years ago

            I wonder if there is a way to “trick” legacy software that is no longer being updated into using these new optimizations, perhaps through some layer of virtualization.

            • chuckula
            • 8 years ago

            With the right software… maybe (with a huge number of caveats there). Applications that run with just-in-time compiled code might see a boost if the JIT compiler gets smarter about emitting the right instructions. GCC and other major compilers can already use AVX2, although to really get the benefit you often need to have critical sections coded in assembly since the C/C++ compilers have to make certain performance-harming assumptions about higher-level functions to maintain memory protection when you have direct memory access and aliasing*.

            * Actually, that’s one reason that FORTRAN is still quite popular in the HPC world: The lack of pointers** means that there are far fewer weird side-effects that the compiler has to worry about and optimizations like auto-vectorization become a lot easier.

            ** Yes, I am aware that the newest versions of FORTRAN add pointers or pointer-like constructs, but most HPC code isn’t using those and can benefit from the simplified memory model.

            • abw
            • 8 years ago

            Yeah , the right softwares a la sysmark….

            • chuckula
            • 8 years ago

            abw, if you are going to downmod my non-troll posts, at least try to show some intelligence. Lemme guess, for all your supposed brilliance, you couldn’t code up a script to automatically post those trite little snarks that you pass off as “comments” if an invading German was holding a gun to your head.

            • abw
            • 8 years ago

            You have to admit that asking for specifical softs is somewhat
            tricking the tests.

            Softs are already hugely intel optimized , so even more optimisations
            just to show that it could be better in some rare and unlikely instances.?.

            • abw
            • 8 years ago

            Out of track , as usual , AVX2 is for integer , FP is adressed
            by FMA extension wich should theoricaly double the throughput
            on parts of a code….

            I guess that we couldnt expect more accuracy , that is , none ,
            from our favourites intel subsided trolls…

            • chuckula
            • 8 years ago

            Moron… FMA is included as part of the overall AVX2 update. You’ll note that FMA instructions use the VEC prefix and are executed through the AVX registers. Just because part of AVX2 includes integer instructions doesn’t mean that other parts of AVX2 can’t include more floating point. Oh, and I can be more pedantic than you are: AVX2 also includes bit manipulation and gather instructions that aren’t strictly integer math operations either! Ooh look!

            • abw
            • 8 years ago

            You are the moron , FMA is for FP , VEC prefix is just the instructions
            coding scheme that is common for AVX AVX1.1 AVX2 and eventual other
            new instruction as bitmanip or popcount etc….

            The AVX2 instructions that are not integer are counted as logical operations
            and are still executed by the integer pipelines..

            Go buy yourself a brain , btw , i m now completely sure ,
            if it was still infinitesimaly doubtfull , that you re not Neelycam ,
            at least , he , he has some EE and necssary theorical knowledge ,
            it s just that he use his skills to mislead the public.

            • abw
            • 8 years ago

            On second thought , those intel suckers , i understand
            now how they can be so easily fooled as to upgrade
            for a half percent IPC gain , perfect milk cows….

            • mczak
            • 8 years ago

            No, you’re wrong, technically FMA is not part of AVX2. It has its own cpuid feature flag (fma). Same story with the new bit manipulation instructions (bmi2). While they may be lumped in together with AVX2 for marketing reasons, it just isn’t true that they are part of it.
            Gather though is indeed part of AVX2, I’m not entirely sure why it didn’t get its own feature flag as it seems quite different in nature to the other AVX2 instructions.

            • abw
            • 8 years ago

            They are clueless , they see their PC as a sexual organ ,
            that s their only reason to hang on in this tech site , TR
            should add some porn for their usage….

            • chuckula
            • 8 years ago

            I upthumbed you just for posting something intelligent.

            • abw
            • 8 years ago

            After downvoting 100x for the same “something intelligent”…

            • abw
            • 8 years ago

            Lolz, 16 half neuroned illiterate suckers exposing their mediocrity…

            • maxxcool
            • 8 years ago

            Says the guy defending a (fake) 8 core cpu that cant even beat a quad core cpu with hypertreading disabled and now has the slowest single thread performance behind its own jaguar cores in the tiny little niche of a tablet market.

    • indeego
    • 8 years ago

    Geez. Barely worth bothering. This is it?

    /Collective shrug, waits for next year.

      • ULYXX
      • 8 years ago

      I think everyone knew it was going to be a small increase in performance. It’s the power usage packages the news will be about. Hopefully the overclocking too.

        • willmore
        • 8 years ago

        According to PCPER, power usage at load is up as is idle power use. Not a good start. 🙁

          • ULYXX
          • 8 years ago

          Interesting. I am looking forward to TR’s review even more now to see what they come up with.

          • smilingcrow
          • 8 years ago

          All of the 3 or so tests I looked at showed idle being ~12W lower and load being higher.

          • derFunkenstein
          • 8 years ago

          That’s really odd. Anandtech is showing it about 15W or os lower.

            • NeelyCam
            • 8 years ago

            Power management settings? Non-compliant PSU?

            • chuckula
            • 8 years ago

            It’s entirely possible that different motherboard & PSU combinations could sway the difference either way. From what TR and other sites have seen, Haswell has lower power usage (especially at idle) than Ivy Bridge, but on the desktop it’s not a gigantic margin. Because of that, outside factors could counterbalance comparatively minor reductions in power consumption.

      • Krogoth
      • 8 years ago

      Join the Krogoth-side, it is your destiny!

        • flip-mode
        • 8 years ago

        Krogoth-side = destiny. Enough said.

        Nailed it.

          • Krogoth
          • 8 years ago

          Damage never told you who your father was.

          Filp-Mode, I’m your father…..

        • dashbarron
        • 8 years ago

        By far the best thread post.

      • flip-mode
      • 8 years ago

      The view is great from my X4 955.

        • ermo
        • 8 years ago

        How are you running it? At P0 = 3.6 GHz and lower than stock voltage?

          • flip-mode
          • 8 years ago

          Stock. My sample takes additional voltage to run at 3.6 and is difficult to get stable at 3.8.

      • Bensam123
      • 8 years ago

      Shhh… such treachery should not be mentioned in the presence of ‘others’. I’m surprised you somehow managed to maintain a +2.

    • Wirko
    • 8 years ago

    [quote<]compare to, well, everything.[/quote<] Thanks a lot, and thanks a lot again!

      • ronch
      • 8 years ago

      Everything? I can’t see my old NEC V20 in there.

        • maxxcool
        • 8 years ago

        Please press “play” on tape.

    • smilingcrow
    • 8 years ago

    The BGA desktop parts interest me so I wonder if anyone will sell a Mobo+CPU package? I’d be happy with an i5 with Iris Pro.
    The only issue is that the lack of heat-spreader causes issues with heat-sink installation but couldn’t they just include a shim or something similar?

    • vargis14
    • 8 years ago

    I really hope your testing on windows 7 so the windows 8 difference does not cloud the performance differences from sandy and early ivy bridge benchmarks.

    I feel that is very important.

    I never post this much but its so hot out and my pool is freezing to the point your genitalia shrinks to a alarmingly small size:P

    • smilingcrow
    • 8 years ago

    “A new addition is the ability to adjust BCLK to one of three pre-defined straps (100/125/167MHz). The BCLK adjustment gives you a little more flexibility when overclocking, but you still need a K-SKU to take advantage of the options.”

    As reported by Anandtech and that’s a bummer.

      • chuckula
      • 8 years ago

      Is that really true that you can only do bclk overclocking with a k-series part? If so, I am disappoint. I’m very interested in bclk options.

      [Edit: Unfortunately it does seem to be the case. Boo.]

        • smilingcrow
        • 8 years ago

        Just quoting Anand as I figured people were hoping that regular non K chips also supported the three clock straps.

          • chuckula
          • 8 years ago

          I’d like hard confirmation either way. I was/am planning to get a 4770 and do some bclk overclocking. Non-k Ivy brige parts can do it too although the lack of different frequency straps means it isn’t all that useful….

          [Edit: Holy-crap, the AMD fanboys are even down-modding posts that could expose disadvantages for Intel’s new products.]

            • NeelyCam
            • 8 years ago

            AMD fanbois still exist? I thought they all killed themselves after the Bulldozer fiasco

            • smilingcrow
            • 8 years ago

            Some bought a bad batch of cyanide and are still stuck in their bunker:

            [url<]http://www.youtube.com/watch?v=SArxcnpXStE[/url<]

      • vargis14
      • 8 years ago

      That is a bummer…I hope asrock and asus etc, can do a work around for no K CPUs

        • smilingcrow
        • 8 years ago

        I doubt it but I like your optimism.

    • flip-mode
    • 8 years ago

    Someone point out where the i5 3570 is on that chart. I can’t see it.

      • vargis14
      • 8 years ago

      Its invisible but i imagine its between the 2500k and 2600k:)

        • DPete27
        • 8 years ago

        Yeah, lets hope at least one IVB i3 and i5 in the review. Sure, prospective upgraders aren’t likely to make a jump that short, so I’m glad the Sandy i3 and i5 were included, but those are still very common processors today….can’t satisfy everyone though.

          • Star Brood
          • 8 years ago

          Just multiply performance by about 0.08 and that’s the improvement between SB and IB.

            • green
            • 8 years ago

            [Edit: brain not working]

      • Krogoth
      • 8 years ago

      Damage most likely doesn’t have a unit on hand for testing.

      Just look at i5-2500K, give it an extra 1-3% boost and that’s where the i5-3570K would fall.

        • abw
        • 8 years ago

        Was it….damaged.??..

          • Meadows
          • 8 years ago

          That was terrible.

            • abw
            • 8 years ago

            Indeed , but how to explain the absence of this popular CPU ..?.

            • Krogoth
            • 8 years ago

            Damage already has IB architecture covered in the form of I7-3770K.

            • HisDivineOrder
            • 8 years ago

            Personally, I think the 3570k is more representative of what most people have and the more interesting comparison. So if anything, let’s have the 3570K and let the users who want to figure out the 3770k +/- 1-3% instead.

            I think there are a lot more users willing to forego that $100+ for hyperthreading, so this is far more relevant to the users who’d read the review.

            • swaaye
            • 8 years ago

            I imagine a comparison of 4770k vs. 3570k would bring lots of questions as to where 3770k is, along with a extra large load of conspiracy theories.

            • ronch
            • 8 years ago

            Yeah, I can’t think of ANY OTHER REASON why TR couldn’t include the 3570K. I just can’t!!

            • abw
            • 8 years ago

            Congratulations, you have freed yourself…

            Welcome in the brainfull community….

          • ronch
          • 8 years ago

          I busted my spleen rolling on the floor laughing. Sorry kids, won’t be able to post cute little comments for the next 5 seconds.

          Or not.

      • MadManOriginal
      • 8 years ago

      I looked up and down the chart 3 times trying to find it before I realized it’s not on there. The i5 3570(K) are almost certainly the most popular Ivy Bridge enthusiast CPUs because of price/performance. They aren’t *that* much more than the lesser i5’s but cost 1/3 less than the 3770(K) which only has Hyperthreading as a major difference.

      • Damage
      • 8 years ago

      It’s on…. the other chart.

      Charts. With an S.

      Hold tight!

    • vargis14
    • 8 years ago

    From what i have been reading Idle power draw is amazing. I am looking forward to seeing how low it goes when not using a 1200watt overkill PSU and using say a 500watt platinum rated psu.

      • smilingcrow
      • 8 years ago

      I have to eat my hat (an original Panama no less) as the power savings for Haswell on the desktop are better than the 5W I predicted. Anandtech have the i7 at 11.5W less than Ivy Bridge.
      Excuse me while I pour a Guinness to help the hat slide down.

        • ermo
        • 8 years ago

        “YouTube clip or it didn’t happen.”

        Enjoy your meal. ^^;

          • smilingcrow
          • 8 years ago

          Damn, I ate my camcorder to.

    • chuckula
    • 8 years ago

    A couple of interestly takeaways just from this chart:

    1. Look at the multi-threaded scores for the 4770K that are almost identical to the 6-Core 970 from the Gulftown era. While Gulftown might sound positively ancient, the chip actually launched in Q3 2010. So in less than 3 years we now have what are effectively quad-core notebook parts putting out scores that are comparable to what are effectively hex-core server parts from not all that long ago.

    2. Per-core score improvement from 3770K to 4770K: 0.08 points. Per-core score improvement from the FX-8150 to the FX-8350: 0.08 points.

      • spuppy
      • 8 years ago

      Power consumption improvement from FX-8150 to FX-8350: 10-15%

      Power consumption improvement from 3770K to 4770? -5%

      That’s a minus in case you can’t see it 😉

        • smilingcrow
        • 8 years ago

        Considering how low the idle consumption is it seems odd that load has increased. Maybe having the VRMs onboard negatively effects load consumption!
        The plus side is that performance per watt has still increased slightly.

    • Sargent Duck
    • 8 years ago

    Hmmm….my E350 HTPC is looking a little lacking…

      • mcnabney
      • 8 years ago

      My e350 is doing just fine under WHS

      • vargis14
      • 8 years ago

      Awe its ok for 1080p vid and web based games but besides that your being a bit optimistic saying its a little lacking:)

      When intel if they do release dual core i3 haswell with the 4600 graphics or the magical iris IGP you might want one of those.

      Or possibly a low wattage richland.

      Temash just does not have the graphics power i was hoping for.

      Whoa just seen the iris whoop the a10 5800k trinity…….I am impressed. A dream for notebooks/ultrabooks.

    • nanoflower
    • 8 years ago

    Glad to see all the work you put into this being appreciated. I’m looking forward to the complete review.

    You can see that the slowdown in single threaded performance increase. Five years and the performance of a Pentium G6950 was at 0.91 while a 4770K is at 1.74. Not even a doubling of performance over five years.

      • Star Brood
      • 8 years ago

      [url<]http://ark.intel.com/products/43230/Intel-Pentium-Processor-G6950-3M-Cache-2_80-GHz[/url<] That's like 2 years old. Nonetheless, Core2 single threaded performance is only about half of what we get from the latest generation...

    • chuckula
    • 8 years ago

    Anand has his first set of benchmarks up, along with some interesting info about die sizes:
    [url<]http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/5[/url<] A 4-core desktop die is only 177 mm^2 (up 15 mm^2 from Ivy Bridge). The large majority of that die size increase is almost certainly in the GPU. It's kind of not a huge surprise that we are seeing much bigger gains in the GPU vs. the CPU when you look at the resource distribution for these chips. What's a little more interesting is that Anand got out a review of the higher-end GT3e parts here: [url<]http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/5[/url<] Using the first link above, the die size for the biggest 4 core + GT3 chips is estimated to be 264 mm^2. Let's do some rough math: GT2 + 4 cores = 177 mm^2 GT3 + 4 cores = 264 mm^2* * This might be a high-estimate since Anand did a direct measurement of a chip covered in epoxy. Good range of estimates are from 220 - 250 mm^2. Size of one "slice" (20 EUs in Intel speak): 87 mm^2 Total GPU size in GT3 parts: ~165 mm^2. This is the first ever Intel design where potentially a majority of the die area is being dedicated to the GPU in the GT3 parts. It will be interesting to see how the GPU does now that it actually is assigned resources that are commensurate with the resources that the GPUs in Llano/Trinity have had for the last two years.

    • My Johnson
    • 8 years ago

    Still no love for for the lower end FX’s (4300 and 6300?)

      • chuckula
      • 8 years ago

      At some point TR will benchmark core i3s that would be more in the ballpark for those parts. In multithreaded benchmarks you can extrapolate their performance pretty easily using the 8350 as a benchmark (or go back to the original reviews).

    • chuckula
    • 8 years ago

    Two words about what you are doing with these reviews: AWE SUM!

    • dragontamer5788
    • 8 years ago

    How is the i7-4950HQ slower than the i7-3770K ?? That makes no sense…

    I’d like to see where this leads to in the full discussion.

      • JumpingJack
      • 8 years ago

      One is a mobile part, the other is a desktop part. The 4770K is the top end Haswell desktop part as 3770K is the top end Ivy Bridge desktop part, hence, that is the comparison to make. The 4950HQ will be the high end notebook part, with the extra graphics oomph. Due to lower power constraints the 4950HQ has a lower base and turbo clock speed, thus the the appearance that the 4950HQ is slower.

      What makes not sense is the naming convention in general.

      • vargis14
      • 8 years ago

      Not that much slower at all….pretty much like having a 3770k in your laptop. Pretty awesome.

      Now all i need is TR to donate the test notebook to me when they are done with it. Pretty please with sugar on top:)

        • dragontamer5788
        • 8 years ago

        Ah, I didn’t realize it was a laptop part.

        Then again, AnandTech is reporting that its a 47W part: [url<]http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested[/url<] There are desktop chips that have lower TDP than the iris Pro / i7-4950hq... so yeah... its pretty much like having a desktop chip in your laptop >_<. Still, the performance is impressive. I'd like to see Techreport's "frame latency" measurements as well, and see how it compares.

      • Krogoth
      • 8 years ago

      It is because has a lower multiplier for its turbo-clocking due to thermal reasons. It is marketed as a laptop, DTR-chip where the platform cannot afford the have same thermal dissipation that normally find on desktops.

    • glacius555
    • 8 years ago

    Enthusiast logic:

    If Asus PQ321 monitor will cost under a grand, I may as well upgrade from Core i5-750. Oh, and maybe replace my HD6950. And get an SSD drive. And a new keyboard. And more RAM than my 4 GB.

      • chuckula
      • 8 years ago

      [quote<]If Asus PQ321 monitor will cost under a grand[/quote<] Sorry ain't gonna happen at launch (as much as I wish it were true).

        • mcnabney
        • 8 years ago

        I would be amazed if it was under three grand. I’m not buying until it is well under one.

          • kuraegomon
          • 8 years ago

          I.e. never. You won’t see a 31.5-inch 10-bit 60 Hz@4K monitor under $1K under any circumstances in the next 10 years. I cheerfully paid $1300 for my Dell 3007WFP when that was a bargain, and I’ve never regretted that purchase. And 6 years later, I’ve still never seen a new 30-inch 4MP monitor (non-Korean division) sell under $1K. So what are the odds that a _4K_ monitor ever will? Yep. Zero.

            • drfish
            • 8 years ago

            Rumor is ~$5k. I think that sounds about right.

            • sschaem
            • 8 years ago

            That for true 4k medical grade display….

            ASUS monitor is far, far from that. At that price they won’t sell even one.

            edit: asus label those as “Gaming monitors” … anyone thinking a second those will be priced at $5K , the same as medical grade monitor, is beyond delusional.

            Remember 2560×1400 IPS monitors are now ~$380

            • sschaem
            • 8 years ago

            Tell that to intel.
            Their chart show 3840×2160 to be mainstream for ultrabook and desktop in 2015.

            We already have 3840×2160 50″ monitor for under $1400

            It’s not hard to imagine that when 4k tv flood the market that intel will be right.

            3840×2160 is the new 1080p

            It won’t take long for 32″ QHD monitor to fall below $1500, and by 2015 below $800

            • Firestarter
            • 8 years ago

            [quote<]by 2015 below $800[/quote<] i want to believe

            • sschaem
            • 8 years ago

            I think its totally possible when you consider that IPS 2560×1440 monitor are now $390
            [url<]http://www.monoprice.com/products/product.asp?p_id=10509&seq=1&format=2[/url<] Have faith 🙂

            • Kurotetsu
            • 8 years ago

            Its under $380 if you’re comfortable with obscure Asian brandnames (which hasn’t been a huge issue, admittedly), but the big brands like Dell, HP, and ASUS still aren’t moving from the $700-$800 price points for 30-inch @ 2560×1600 and ~$600 for 27-inch @ 2560×1440.

            If they (Dell, HP, ASUS, etc) DO decide to skip those and go straight to affordable 3840×2160 monitors, I’d very VERY surprised (and delighted).

            • mcnabney
            • 8 years ago

            Actually, that is totally untrue and the $1200 Seiki 4K display is proof of it. The Chinese are going to destroy Japan and Korea by selling great equipment at a tiny fraction of the price. Remember those $350 2560×1600 IPS displays from Korea that turned out to be pretty decent hardware? China is going to completely upset the fat-margin side of the display market. I just hope they hurry up with it. I’ll be happy to learn a bunch of new Chinese brands if it means I don’t have to pay thousands for high DPI equipment from Japan and the Koreans.

            • anotherengineer
            • 8 years ago

            Those $350 displays which people are calling ‘Korean’ are shipped from there. Upon disassembly you will see that it is an LG (Korean company) display panel that is made in China.

            And in fact almost all LCD display panels are now made in China.

            And Japan hmmmm I guess some Sharp and Panasonic panels are still made there, but probably end up in TV’s, or other devices other than PC monitors.

            So a Chinese knock-off IPS would probably never show up in North America since LG owns the patents on this and would seek a legal injunction.

            China is of right now basically a sweatshop for the worlds’ large corporations.

    • vargis14
    • 8 years ago

    Ill be keeping my 2600k…but for everyone’s sake i really hope they solder the IHS onto the silicone instead of using the goop they do on ivy chips.

    • pack66
    • 8 years ago

    Can’t wait to see how this family of chips overclocks.

      • vargis14
      • 8 years ago

      From what i have seen with a h100i and just under 1.2 volts in a 20c room it hit 90c at 4.6ghz.
      But at 4.6ghz the 4770k is faster then a 4.8ghz 3770k in cinebench.

      Here is a link to Overclock3d video review. I enjoy his reviews.
      [url<]http://www.youtube.com/watch?feature=player_embedded&v=xrLtjgbga5g[/url<]

        • Krogoth
        • 8 years ago

        I’m more curious to see if Intel has gone back to thermal epoxy with Haswell.

        It looks like its overclocking headroom is similar to “good”-“golden egg” samplings of Ivy Bridge.

          • DeadOfKnight
          • 8 years ago

          Well if you really want to overclock then you’re best going for the i5, unless you really need hyperthreading, to have some more thermal headroom. I think you might also be able to disable the iGPU to get even more(?).

          • vargis14
          • 8 years ago

          I can pretty much guarantee Intel is using the same paste as they did for Ivy bridge since pretty much all the reviews that include overclocking are pushing temps the same or higher then Ivy Bridge.

          I see de-liding just like has been done with Ivy for the hard core Overclockers that do not use phase or any other extreme cooling.

          I have seen many Ivy de-lid reviews where they use a Liquid Metal TIM like Coollaboratorys Liquid Ultra between the IHS and the CPU chip itself and get a good 15c cooler temps on average, along with a few more hundred MHZ out of a cpu that is hitting its ThrottleBack temp.

          The difference between soldering and paste between the IHS and CPU is Huge….A soldered IHS effectively in a way pulling heat out of the silicone since it fills every imperfection and sort of increases the die surface area since it goes down the side of the actual cpu die a wee bit but does not block the heat transfer like a too thick layer of paste does. A single mm of fluxless solder surrounding the cpu die increases surface area a good bit since it is pulling heat from the side of the cpu die also. I think that is why removing the IHS and mounting the cooler directly on the cpu die did not make much of a difference as the liquid metal does. If that makes any sense at all.

          • abw
          • 8 years ago

          wrong chain…..

      • My Johnson
      • 8 years ago

      Anandtech has between 4.3 to 4.7 on air. Intel is leaving very little on the table as it gets more efficient.

      • sschaem
      • 8 years ago

      Not to well from what we see around thw web. ~4.2 ghz depending on your luck.
      It seem mostly due to the build in voltage regulators.
      Well, 22nm is 22nm, power and heat seem just a little worse then the i7-3770k at equal clock speed.

    • 5150
    • 8 years ago

    This chart just awesome. Can’t wait to read this thing!

      • Deanjo
      • 8 years ago

      Hopefully they include a handbrake test (quality based, not useless two pass) and have runs utilizing no openCL, openCL and QuickSync encoding.

        • JrdBeau
        • 8 years ago

        I couldn’t agree more. I’d like to upgrade from my i7-920 but not have to delve into the high rent district of LGA2011.

          • mcnabney
          • 8 years ago

          Just what is it that you feel the i7-920 is sputtering on? My core2 Duo chugging along at 3ghz is only just now feeling pokey.

            • chuckula
            • 8 years ago

            It all depends… my core2 E8400 at 3.6GHz is just fine for daily desktop use and generation of trolls for TR.
            However… for more intensive computing operations including larger compile jobs and number crunching, it is clearly showing its age.
            You have a good point that many computing tasks don’t need a 4770K though, which explains the popularity of lower-power components now.

            • JrdBeau
            • 8 years ago

            It’s not so much that it’s feeling pokey (although it does with HD encoding) but it’s more that things are starting to fail in the system and I’d like to take advantage of some newer technologies.

            • mcnabney
            • 8 years ago

            What did you buy that is failing that quickly? Seriously, my Apple 2e and Amiga 2000 still function! I’ve lost hard drives, optical drives, and one time a power supply crapped out on me, but something is seriously wrong if you can’t get more than 2-3 years out of a computer before it starts to fail.

    • Deanjo
    • 8 years ago

    Looks like the typical modest generational jump in performance. About the same jump we’ve seen going from SB to IB.

      • chuckula
      • 8 years ago

      You might want to check out the OpenCL compute page over at Anandtech: [url<]http://anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/17[/url<] The comparison between 47/55 watt GT3 Haswell parts and the 100 watt A10-5800K is interesting and shows that Intel has potential to compete in this area.

        • Deanjo
        • 8 years ago

        Ya I did check them out but if it isn’t on a desktop processor it really is of no interest to me. Even then utilizing openCL on a add on card is going to offer more flexibility and more then likely better performance and can be scaled by adding more cards if needed. I still see openCL support on a processor more of a “proof of concept” then anything else. I’m not going to have my laptops crunching away at openCL/Cuda code with the exception of maybe using it for light development purposes.

        Plus unless they finally give us proper GPGPU support in linux it is a non factor.

          • chuckula
          • 8 years ago

          I agree that the IGP is never going to go up against Titan. It’s more of a relative thing compared to other solutions in the same category. The Iris Pro graphics are just the very first iteration of what we will see leading up to Sky Lake where there is a great deal more integration between CPU/GPU. It’s basically the same thing as the “HSA” architecture that AMD has pushed.

          As for Linux support, neither Intel nor AMD have done a good job with OpenCL and CUDA is really the way to go. Ironically, Intel is pushing Linux in HPC in a huge way with the Xeon Phi, but the benefits of that work have not translated into desktop processors (maybe they will in the future though).

            • Deanjo
            • 8 years ago

            Well they can push openCL with Phi as intel has had cpu openCL support in linux for a while and it isn’t a really big challenge. I can see it being useful in a handful of areas in the regular desktop/laptop for file compression and video encoding were it would help out somewhat for marginal processors.

        • Mr. Eco
        • 8 years ago

        Then look at the prices, i7-4950HQ is $657. It does beat Trinity, but at that price the CPU/GPU costs more than a complete medium range laptop.

        [Edit]That was to chuckula.

    • ronch
    • 8 years ago

    I wonder just what Jim and the boys are doing to keep up.

    • esc_in_ks
    • 8 years ago

    I’ts interesting to compare the 3770k and 4770k, of course.

    The single threaded performance difference is small, but the multithreaded performance difference is much larger. I presume the reason for this are the new Transactional Synchronization Extensions (TSX) which seek to exploit “hidden” possibilities for concurrency without acquiring locks.

    I look forward to seeing more results, especially in the multi-threaded realm.

    • pedro
    • 8 years ago

    Bring it on boys!

Pin It on Pinterest

Share This