Intel removes modest ‘free’ overclocking from standard Haswell CPUs

Intel’s K-series processors offer fully unlocked multipliers that place few restrictions on overclocking. With Sandy Bridge and its Ivy Bridge successor, Intel has also allowed limited overclocking on regular Core i5/i7 CPUs that aren’t part of the K series.

On those regular-model chips, one may increase the maximum Turbo multipliers by four "bins" above stock, effectively delivering up to a 400MHz overclock. That boost won’t get you into the near-5GHz territory attainable with fully unlocked versions, but it’s a nice freebie that most CPUs and coolers can tolerate.

When we were discussing Haswell overclocking on this week’s podcast, we weren’t sure whether Intel had extended limited multiplier control to non-K Haswell products. We checked with the firm and are sad to report that this feature has been removed.

We’ve already reported in our 4770K review that, although Haswell offers more flexibility the form of additional base clock straps, access to those straps is disabled in non-K parts.

As a result, with Haswell, overclocking support is now almost entirely confined to K-series CPUs.

Intel says it’s targeting non-K chips at "the business and consumer market where overclocking is generally not performed."

Based on what we’ve seen in our labs and heard from industry sources, K-series Haswell CPUs have less clock headroom than their Ivy and Sandy Bridge predecessors. The Core i7-4770K being tortured on my test rack peaks at 4.5-4.7GHz, depending on the motherboard, and it requires potent cooling at those speeds. A more modest 400MHz bump shouldn’t require an exotic cooler, but getting it will require paying the premium for a K-series processor.

Buying a K-series Haswell processor will set you back an extra $20-$30 over the equivalent standard model. The Core i7-4770K is priced $30 higher than the Core i7-4770, while the Core i5-4670K is $20 more than the Core i5-4670.

Paying the extra for a K-series product also means giving up support for one of Haswell’s key features, the TSX extensions that enable transactional memory. Intel has stripped out the VT-d device virtualization and vPro management features in the K series, as well.

In the end, enthusiasts face a rather unfortunate set of choices in Intel’s Haswell-based product offerings. We can’t help but think this situation wouldn’t exist if AMD were putting more competitive pressure on Intel.

Comments closed
    • GatoRat
    • 8 years ago

    I’d wager that buyers of well over 99.9% of non-K series Intel CPUs have no intention of overclocking. Supporting that tiny group cost a disproportionate amount of money that is very likely non-trivial and doesn’t have a hope of being made up for in sales. This drives the price up for everyone and creates more breaking points for overly zealous IT people who overclock what they shouldn’t (further driving costs up.)

    • BaronMatrix
    • 8 years ago

    Awwww…. Poor babies… Cue the excuses…

      • clone
      • 8 years ago

      tone it down ace, demonstrating such a low level of maturity speaks badly of the author only.

    • ronch
    • 8 years ago

    I’m not really affected by this since i (embarrassingly, perhaps) mostly stick with AMD. If anything, Intel’s arrogance when it comes to taking away certain features that only serve to irk their loyal fanbois only makes AMD’s value proposition stronger.

    And anything that can help balance an oligopoly can’t be all that bad, right?

      • Ryhadar
      • 8 years ago

      I prefer AMD but I have a no exceptions policy when it comes to the performance PCs in my home (i.e. my gaming desktop) so I’ve gone Intel for newer builds — though half of the PCs in my home run AMD CPUs still.

      I like being able to overclock my chips later down the line when they get old but if Intel continues to get rid of features from their K series CPU in the future it’s going to muddy the waters a little bit. Plus, if AMD gets their act together and motherboard makers start releasing some decent mATX offerings I wouldn’t hesitate to switch to AMD for my gaming rig.

      Luckily with my 2600K the only thing I’m losing out on is VT-d (which is something I will never, ever use).

    • hippie69
    • 8 years ago

    I did this for my Ivy Build. Get an E3 Xeon. You get Hypertheading, VT-d, and for Haswell based ones TXT plus it cheaper than simlar clocked i7s. I have E3-1245 v2 running in Asus Sabertooth stable for a year+ now. I’m just a hair under the performance of an i7-3770k.

    Look into the Xeon E3-1240 V3 as an example. 🙂

    Word of warning though. Putting a Xeon in a non “workstation” labeled board is unsupported by motherboard makers, but in practice should work from when I did my research last year.

      • Krogoth
      • 8 years ago

      The only caveat is that you need to get a motherboard with ECC support if you intend on using unbuffered ECC module with that Socket 1155 Xeon. 😉

        • hippie69
        • 8 years ago

        You don’t have to run ECC memory. It’s happy purring along with those Samsung DDR3 sticks that were great overclockers.

        [url<]http://imgur.com/a4fgNzA[/url<]

    • credible
    • 8 years ago

    Well if you needed any extra confirmation that there is no longer any competition, here you have it.

    • RachelGat7
    • 8 years ago
      • ronch
      • 8 years ago

      WOW! REALLY??? OMG THAT’S REALLY REALLY AMAZING!!!!!! SIGN ME UP!!!!!!!

    • gc9
    • 8 years ago

    Doesn’t this mean that TSX and VT-D hardware are on a critical path and are the first features to become unstable at high clock rates? So disabling means that those features cannot be used, so the chip will remain stable at higher clock rates than if those features were enabled?

    Many social chatters seem to be blaming the psychological intent of the marketing strategy, but couldn’t the cause be the limitations of the engineering design or fabrication process? TSX in particular is a first product implementation, so if it is in a critical path one might well expect that it may glitch at high clock rates. If so, maybe the restriction will be relaxed in a future stepping a year from now or so.

    • LoneWolf15
    • 8 years ago

    The much bigger thing that ticks me off is removing features like VT-d in K-series CPUs. Yeah, because enthusiasts would never run virtual machines…wait.

      • Kougar
      • 8 years ago

      Pretty much. Kernel overhead comprises more than half of the total virtual machine overhead on my 4770K, and I cannot help but wonder if VT-d would have minimized it appreciably.

      The chip runs so hot under OC that even with exotic cooling it just wasn’t worth it. If they had left the turbo mulitpliers intact I would have been happy with a vanilla 4770 running at 3.9Ghz, which is exactly why they did this in the first place.

      • maxxcool
      • 8 years ago

      And we do not run multi-threaded apps either… (tsx loss)

    • Evil_Sheep
    • 8 years ago

    Honestly…it’s debatable if this will have much of an impact on people. The main takeaway is that this is effectively the nail in the coffin for overclocking. A decades-long, proud tradition of nerds squeezing out every last Mhz from their CPU’s is coming to an end, along with any kind of control or customization of your hardware, and soon to be followed by the desktop itself. RIP.

    • Stickmansam
    • 8 years ago

    Bit irked by this. I had originally been considering getting a 3470 or a 3570 and the “free” turbo thing felt good to me as sort of a free speed boost and not having to pay the K premium. I did eventually settle on a 3570k though.

    Intel has been cutting back mainstream OC for quite a while with a unified bclk, no more FSB (Loved to OC with that), no i3k, and now no “OC” for any turbo enabled CPU’s

    I am also guessing this mean no OC for the Xeons? I had also been considering at the time getting a 1230v2 or a 1240v2 as they were cheaper than the 3770/3770k and could be OC to 4.2ghz and 4.3ghz respectively using the turbo method.

    Guess that means I can’t use Xeons as i7 replacements anymore nor any hope of a speed boost on the cheap for my future upgrade 🙁

    • moose17145
    • 8 years ago

    I would like to be the one to say that AMDs chips are not entirely uncompetitive. Yes they struggle wiyh gaming relative to Intel. And Intel does beat AMD soundly in media encoding tasks. But who cares on that one? Intel has always beat AMD on that front. Even during the pentium 4 vs athlon 64 days Intel was still beating AMD on that front. It has been Intels playground on that front for a while now.

    But for certain work related tasks the AMD chips do just fine. Some of TRs own tests do show the 8350 keeping up with and even beating the haswell chips. For example one of my buddies just built an ESXI box on the cheap as a home media server using an AMD FX 8350. And we were both shocked at how well the chip handles multiple VMs running at the same time, even when under a fairly heavy load. Their chips are hardly as impotent as everyone seems to make them out to be.

    • Nutmeg
    • 8 years ago

    I’m more annoyed by how Intel is wasting die space on integrated graphics even on high-end desktop chips where they will never not be paired with a graphics card. As they start to hit the nanometre limit, transistor budgets will stop increasing every step down, and they’re wasting space on the chip for integrated graphics that no-one running a desktop PC asked for or wants? Gee thanks Intel.

      • OneArmedScissor
      • 8 years ago

      Die, factually incorrect myth, die. They will always sell the Xeons as enthusiast parts.

      And guess what? They’re bigger [b<]without[/b<] the GPU! Regular Sandy Bridge quad-core: 216 mm² Sandy Bridge E quad-core: 294 mm² GPUs are copied and pasted circuits which are easily die harvested. It does not kill yields or put the price through the roof. Look at what they do with the desktop parts - disable part of the GPU! If there comes a day and age where they are stuck with silicon, and it cannot be shrunk down anymore, they will continue to do what has already been done for years: Add more circuits, at the same node, because the price of manufacturing will continue to drop. Power gating already allows "oversized" chips to fit into reasonable TDPs, negating the "transistor budget." Jaguar would actually melt if the entire chip were powered on at once, and those are tiny!

        • chuckula
        • 8 years ago

        I agree with most of your post but: [quote<]Look at what they do with the desktop parts - disable part of the GPU![/quote<] Not correct unless you are talking about future ultra-budget parts where they will turn off 10 EUs. The desktop Haswell die only has 20EUs. The Iris-pro Skus include 40 EUs and are a different (larger) die.

          • OneArmedScissor
          • 8 years ago

          Well, all I’ve ever seen is 177 mm² for any Haswell quad-core. I figure they just lop half of the GPU off to make life easier, much as they still disable 1/4th of the L3 cache whether it’s an i5, i3, or Pentium.

          They definitely build the dual-cores with at least two completely different dies, sometimes three if OEMs want more of the super borked version bad enough.

      • ColeLT1
      • 8 years ago

      I agree somewhat, but it’s not a waste of space when the iGPU can be used to offload the cpu, like floating point calculations.

    • Diplomacy42
    • 8 years ago

    “We can’t help but think this situation wouldn’t exist if AMD were putting more competitive pressure on Intel.”

    the funny thing is that this whole article is complaining about ultra niche features, meanwhile Intel has all but done away with their flagship service that EVERYONE should care about: Hyper-threading. HT should be on every chip above an atom class, instead, it has been all but removed from the i3 and i5 lineups and isn’t even guaranteed in the i7.

    meanwhile TR is worried about 400mhz? okay then.

      • bittermann
      • 8 years ago

      I agree…hyperthreading does have its use now days. Never did like that trick they used disabling it on the i3/i5’s after the original 5xx/6xx series.

    • Stian Aarskaug
    • 8 years ago

    [quote<]In the end, enthusiasts face a rather unfortunate set of choices in Intel's Haswell-based product offerings. We can't help but think this situation wouldn't exist if AMD were putting more competitive pressure on Intel.[/quote<]Amen! The enthusiast platform is still old (SB-E), outdated (lacking features), Intel let us overclock less and less, removing features, using crappy TIM. I hope AMD come back with some competition, we need that badly! Haswell is great and all, impressive low power consumption, but: Performance increasement from last architecture is not that big, in some applications there is no difference, in some ~15 %. They are removing features from K-models, still using bad TIM, we have to wait for the Haswell-E. I remember back in the days when we had Yorkfield and Wolfdale. Overclocking was a joy and I was just happy to pay them the price for my CPU. The enthusiast platform was good till X58 too, but after that Intel just stopped focusing on us enthusiasts. Sandy Bridge came, we had to wait for Sandy Bridge-E, which sadly was a dissapointment. The lowest priced CPU was also locked, there was no SAS support and so on. Now Ivy-days are gone and with an new architecture we're still stuck at Sandy Bridge-E with LGA 2011... AMD, we need you! Please be aggressive about Steamroller. Don't be shy, we can handle high TDP, after all, we're enthusiasts. It would just be fun to have an unleashed toy to play with! Sadly the AMD-motherboards are limited compared to Z87. I am considering holding my breath until next LGA 2011 update. But seriously, I won't wait til late 2014... Somebody should put Intel in the corner of shame.

    • tfp
    • 8 years ago

    When Intel comes out with a Haswell Xeon, will any of them be unlocked? I can’t remember what they did in the past.

      • Stian Aarskaug
      • 8 years ago

      All Xeon will be locked as far as I know. Xeon is more focused on “stable 24/7 operations”.

        • jihadjoe
        • 8 years ago

        +1, there hasn’t been an unlocked Xeon for quite a long while now.

      • Krogoth
      • 8 years ago

      Xeons have always been half-lock, but their memory controllers have ECC support. They have all of the other features enabled. Extreme Edition chips are fully unlock and have all of the CPU features enable. The caveat is that they do not have ECC support. Engineer samples of Xeons are the only pieces of silicon that are fully unlocked and contain everything. 😉

        • ermo
        • 8 years ago

        Do you know if the features are disabled in silicon or microcode?

        I really wish someone would unlock the low-end-ish Xeons (the ones priced around the K-models) and have all the features work on the Z7x/Z8x chipsets (IB and Haswell, respectively)

      • Kougar
      • 8 years ago

      Haswell Xeons will be linked to Haswell-E, which won’t even debut until 1H’15. That’s a long ways off. IB-E Xeons will launch at the end of this year…

    • OneArmedScissor
    • 8 years ago

    Lame and confusing, but I’ll take this way over unmarked FSB walls that just wouldn’t boot if you crossed the mystery number.

    The problem isn’t lack of competition as much as Intel’s bureaucracy. They have X number of fixed price tiers and ever fewer ways to fill them. Consider what else has changed since they stopped bus overclocking:

    The clock speed gap is gone. Even some inexpensive CPUs without turbo are nearly 4 GHz, right out of the box.

    Yes, you could overclock those in the past, but, again, look at what changed:

    What that meant was turning your artificially crippled CPU up to the speed of one that cost more. It did not mean you buy the 2 GHz Core 2 and turn it up to 4 GHz when the highest binned parts were only 3 GHz!

    If the problem were a lack of competition, we’d see the return of $600 or $1,000 mainstream parts and branding by clock speed.

    We’re spoiled. My initial reaction was the same as many people – that something has been taken away. But in retrospect, I’m not sure where we developed the idea that you buy an inexpensive chip and turn it up to 5 GHz. That never actually happened!

      • tfp
      • 8 years ago

      I have that FSB wall on my Q9400, so annoying.

        • OneArmedScissor
        • 8 years ago

        I only even know about it because of one time when my friend and I both bought the exact same model CPU, at the same time…and mine had a super low FSB wall, but his didn’t.

        It turned out they had changed it in some minor stepping and it drove me mad trying to figure that out.

        No more of that, please!

    • maxxcool
    • 8 years ago

    Ok, I missed that part? TSX and VT-D do not function on k-series, but work on non-k series… huh.

    welp, guess what. time to go darkside I guess. As everyone knows I am a HUGE INTEL FAN. I hate AMD’s fake core bull%^&* and I despise their raid implementation….

    But I build vmboxes for myself. not alot. but enough to now have to buy AMD only.

    Well played intel…

      • strange_brew
      • 8 years ago

      So you build VMboxes that require overclocking? That seems … odd. Also, VT-d is only useful in very specific scenarios where you assign specific HW to a specific VM.

      I do agree that not having TSX available across the entire product line is very disheartening. I think this is the first time Intel has extended the ISA conditionally based on SKU.

      In general I think it is a good idea to always have a golden SKU with all features enabled. That used to be the K models. Now, nothing.

        • PixelArmy
        • 8 years ago

        I think it’s a good idea for a fully featured SKU, but let’s not pretend that the K model filled that role.

        They’ve been omitting VT-d since the original K, the i7-875K…
        As far as I can tell, the only K model with VT-d is the Sandy Bridge-E one, the i7-3930K.
        (ark.intel.com)

        • maxxcool
        • 8 years ago

        I do overclock, everything (even my car)
        I do need VT-d for esxi
        TXS .. meh for now… but in the near future 100% yes

        But the idea is if I buy a k, I expect it all on by default. K means better, more fetaures. Or well it used to… oh well. AMD esxi boxes it is then.

    • dashbarron
    • 8 years ago

    I miss the days of pressing the turbo button on the front of the case and enjoying a raw speed boost of 30MHz

    • anotherengineer
    • 8 years ago

    “We can’t help but think this situation wouldn’t exist if AMD were putting more competitive pressure on Intel.”

    Don’t blame AMD, people vote with their wallet and AMD doesn’t have the cash to compete.

    Furthermore for an “enthusiast” whose hobby is PC’s if they need all those options and want the speed, they have the chips, but they will cost ya.

    If no one bought ‘k’ series chips and told Intel it’s because of the options that are missing, maybe they would add them to the ‘k’ series.

    However on that note, kinda sucks for the consumer on Intel’s decision to semi-castrate the ‘k’ series chips though.

    • Willzzz
    • 8 years ago

    You still get multi-core enhancement though, which is still a free 200mhz boost when using all 4 cores.

    • DeadOfKnight
    • 8 years ago

    I never thought I’d say this, but if you can wait another year then you might be well served by watching closely for the Steamroller value plot. AMD won’t be catching up on raw performance, but performance per dollar is a different story. If you look at [url=https://techreport.com/r.x/cpu-gaming-2012/value.gif<]this plot[/url<], they could make a significant leap around the $180 mark if you factor in overclockability.

      • pureevilmatt
      • 8 years ago

      Edited 11 time(s). You may have OCD.

        • chuckula
        • 8 years ago

        I’ve been known to edit posts a couple of times here and there, but DoK just took the crown.

          • DeadOfKnight
          • 8 years ago

          What? There’s no preview button 😛

            • cybot_x1024
            • 8 years ago

            [i<]" Edited 42 time(s). "[/i<] Now you're just showing off

    • flip-mode
    • 8 years ago

    Geoff, I’d love to see you delid your CPU just as our colleague chuckla did, and give a report on what kind of benefits you get from it.

    • chuckula
    • 8 years ago

    Somewhat irked at Intel over this and.. most importantly.. their decision to abandon soldered IHS components which appears to be why getting to near 5GHz is quite difficult with these chips. I delidded mine, but I did that more for the fact that one day I’ll be able to say I did it rather than it being a particularly good idea.

    Haswell is a complex and sometimes contradictory beast. Intel got the hard stuff right but then screwed up the easy stuff. You can see the mega-corporate nature of the design since the engineering side put in some very useful and powerful features on that silicon, but then the marketing side screwed over a lot of consumers who would like to actually use those features.

    The trick is, as others have pointed out, would I go with AMD to save less than 10% on my new build? Nope. They aren’t offering anything better right now either and frankly AMD has its own corporate issues that get papered over because they are the perceived “little guy” so everybody cuts them slack in the process.

      • kroker
      • 8 years ago

      This is what we get when Intel has complete control of overclocking. They didn’t abandon overclocking outright because of the backlash it would have caused in the vocal overclocking community, instead they just slowly make it more and more inconvenient in the K-chips and erode it away in the non-K chips. Just like boiling a frog.

      • Mr. Eco
      • 8 years ago

      IMO, the only issue with AMD chips is the heat. They are fast enough, very good integrated graphics, good idle power consumption. They all have got the features (AES, VT-d) that Intel excludes from i3 and even from some i5 and i7 processors. The price is much better, and that matters a lot for a low or medium class computer.
      What is not good is the power consumption under load.

        • chuckula
        • 8 years ago

        [quote<]They are fast enough, very good integrated graphics, good idle power consumption.[/quote<] Idle power consumption is OK. The chips that have "very good integrated graphics" are frankly [b<]not[/b<] good enough at CPU performance for my desktop needs. They're fine for running Office on a desktop, but I wouldn't be building my own systems if that was all I care about. The chips that are "good enough" for desktop use (the higher-end of the FX series) have no integrated graphics (which is actually a plus), but their performance still isn't stellar and the FX platform is rapidly showing its age. I'm aware of the "at least it's cheap!" argument, but when I built my new system the total overall price savings of going to AMD would have been less than 10% for more than a 10% deficit in performance, plus being saddled with an OLD platform that would require a major update in less than 2 years anyway. When AMD can get my SSDs, RAM, PSU, GPU, and case discounted by a large margin when I buy their CPU, then maybe it would be worth it, but I doubt that is happening any time soon.

          • M3gatron
          • 8 years ago

          Not good enough for desktop needs???
          What needs?
          A core 2 duo is enough for most desktop needs and a 4.1ghz quad core isn’t. Year right.

            • flip-mode
            • 8 years ago

            AMD’s FX is a server processor at heart. Having 8 threads isn’t exactly highest priority for a desktop. Workstations and servers – sure. There are many instances where a dual core i3 3225 turns in better performance than an FX 8350. That should never happen in my opinion. Stuff like that makes it very easy to argue that AMD’s FX isn’t well suited to desktop needs, at least not compared to any of the several alternative options in Intel’s chip lineup at the same price or less. The FX chips certainly don’t suffer a conclusive loss, but neither do they deliver any kind of clean win. That’s the epitome of a compromise. The FX is a compromised product.

            • OneArmedScissor
            • 8 years ago

            It’s a compromise for servers, too. They went from 12 K10 cores to 8 Bulldozer modules – but without the clock speed hike of desktops.

            AMD just plain blew it with the modules thing. They admit it by reneging on the shared decoder with Steamroller and have publicly stated there is still a bug in the L2 cache which increases its latency.

            I’m sure AMD would have liked to start out with a 32nm “Phenom III” while they worked out the kinks in the obviously flawed Bulldozer, but it just wasn’t meant to be. Their hand was forced when something went horribly wrong with the yields in Llano, caused specifically by the K10 cores. Weird science!

            • flip-mode
            • 8 years ago

            Yup. Never know. Steamroller could be an updated Thuban! Bah, that’s too good to wish for.

            • ermo
            • 8 years ago

            Do you have a good link to the L2 cache bug thing for those of us who are curious about it? I didn’t know that they had actually acknowledged it publically…

            I’m also quite amazed that they didn’t take the time to fix it for Piledriver if this is indeed true. For something like the shared decoder, I can see how it would make more sense to wait until Steamroller with that.

        • flip-mode
        • 8 years ago

        And single thread performance.

          • heinsj24
          • 8 years ago

          Doesn’t matter to me if apps are single-threaded or not – the OS alone makes it a multithreaded world.

            • flip-mode
            • 8 years ago

            A multithreaded OS does not magically convert single-threaded loads to multi-threaded loads. So the point remains: if any of your applications are highly dependent on single-thread performance, AMD’s FX processors are pretty much immediately unattractive.

      • ronch
      • 8 years ago

      In my case, I shoot AMD regardless of how much I prefer their chips if they do something that I think is stupid or questionable. How I wish Cyrix is still around so I could have a third (or should I say, second) option.

      • sircharles32
      • 8 years ago

      So, all of your computers are about $3,000?
      Because on mine, it’s more like 30%, which is NOT insignificant.

        • chuckula
        • 8 years ago

        Really? For a $1000 computer I get about 21% if you are being generous to AMD:

        1. $160 savings for an FX-8350 vs. a 4770K (note: this number drops massively if you just get the 4670K instead).
        2. $50 for the motherboard, which is again being generous to AMD since there are plenty of full-ATX Z-87 motherboards from quality manufactures that cost about the same as quality (note the term quality there) motherboards from AMD.

        That’s 21% using optimistic numbers for AMD. You can get the number bigger if you want, but now you are going WAY down the performance ladder for AMD while artificially requiring purchase of the highest-end Intel parts when they clearly make lower-end parts as well.

        Now let’s assume your 30% figure is true. Well if you look at the benchmarks, it turns out that the supposedly overpriced Intel parts are delivering 30% higher performance in a wide range of workloads too…

          • NeelyCam
          • 8 years ago

          This is what I’ve been saying about cpu prices for a long time – people should focus on system prices instead ot CPU prices. When you move from a cheap but slow cpu to a more expensive but faster one, you can improve the overall system performance significantly with a pretty small relative system cost increase. In a way, the better cpu lets you extract more value from the other components in the system you have to pay for anyway.

          This is why I’m against those price/performance CPU scatter plots – they are misleading unless you only look at CPUs at the same performance level. Same thing with graphics cards, although there the two big competitors are on more equal performance footing these days.

          What I’d like to see is a performance/price scatter plot for [i<]systems[/i<] in the next TR system guide.

            • sircharles32
            • 8 years ago

            Not exactly true.

            You’ll have to sacrifice on other components in order to squeeze in the more expensive CPU/Motherboard within a given budget.

            If you take gaming for example, the video card is much more important than the CPU. If you go with the lower cost AMD systems, you can squeeze in a more powerful graphics adapter, which will allow you better performance in the games. If you dropped the $$$ on the CPU/motherboard, then you’re dealing with significantly less capital to spend on the graphics card.

            Another example, would be if you wanted to go down the SSD route. You can drop the extra cash into the CPU/MB, or spend it on the SSD, which will most likely be something you will actually notice, in day to day computing.

            Dealing with a budget, it’s a balancing act. It’s up to the consumer to decide what is more important.

            • flip-mode
            • 8 years ago

            There’s no silver bullet. I only plan to upgrade my core components – not the whole box.

            Then with the IB launch you had a bunch of people with SB already who really wanted to know the scatter plot for the CPU and CPU only!

      • Krogoth
      • 8 years ago

      5Ghz is difficult to obtain with 22nm and it doesn’t take well to overvolting. ~5Ghz range requires exotic cooling even with a direct die to HSF contact. IHS issue is massively overplayed. It only makes it more difficult to manage thermals with an aggressive overvolt and overclock, but 22nm process becomes as much as a problem at those speeds.

      I think the real issue is overclockers got spoiled on the Intel with Conroe through Sandy Bridge. The silicon from these generation had excellent headroom and require little effort to obtain a healthy overclock. Ivy Bridge changed this convention through more problematic 22nm process not scaling as well as its previous fab processes and the aforementioned IHS issue.

        • NeelyCam
        • 8 years ago

        Not hitting 5GHz on Intel’s 22nm chips is not because of the process but the architecture – SB/IB/HW were designed for high ICP. In contrast, Vishera was designed for high speed.

        If you make a Prescott on 22nm, I wouldn’t be surprised if it was running at 7-8GHz at reasonable supply voltages. It would still consume a ton of power, though – not a good choice for servers etc. Where efficiency is king.

        AMD brilliantly (or desperately?) revived the gigaherzt race because they have a speedster architecture, and laymen don’t know that clock speed is only half of the equation. It’s easy to market a 5GHz CPU when the competitor doesn’t have any. This is similar to what Intel was doing when Athlon was released – their marketing focused on clockspeed and managed to trick consumers into buying even though Athlon had better performance.

          • OneArmedScissor
          • 8 years ago

          After Vishera, we can safely say that there is no architectural clock speed advantage. They are just underclocking the L3 and memory controller, and sometimes to no core clock benefit!

          Compare apples to apples for 32nm and 95w TDP:

          FX-8300: Base 3.3 GHz. Turbo 3.9 GHz. L3 / IMC 2.2 GHz.

          i7 2700K: Base 3.5 GHz. Turbo 3.9 GHz. L3 / IMC 3.5 to 3.9 GHz.

          If you look at the die harvested Vishera parts, it’s even worse because there are massive amounts of disabled cache and the L3 / IMC is further reduced to 2 GHz!

          • Krogoth
          • 8 years ago

          The fab process is the problem. You cannot ignore that laws of the physics my friend. AMD’s “Extreme Edition” Piledivers consume a ton of power to get that ~5ghz turbo clock at load. I’m willing to beat that units consume 250W to get it. Not even the Extreme Edition Smithfield consumed that power back in the day. 32nm Sandy Bridge and Clarkdales aren’t that far behind when they are pushed that hard. Gulftowns are in the same ballpark.

          22nm process on Ivy Bridge only yielded a modest reduction in power consumption over Sandy Bridge. Ivy Bridge sucks up power and throws out a ton of heat when you aggressively overclock and overvolt it. Haswell doesn’t fare much better. Delidding IHS yields only ~5-10C drop on loaded temperatures. It may get you an additional 100Mhz or as far as another 300Mhz if you are lucky before hitting the thermal wall.

          Intel doesn’t need to push out a 5Ghz chip, because the performance race isn’t about “MEGAHURTZ” anymore. It is about having more threads. Intel is the clear winner here if you look at 6-core and 8-core SB-E chips (12 and 16 threads). In professional applications these beasts have no equal from AMD which is why they command north of $500 all way to $2000 range.

          AMD’s marketing diviosn is just being facetious with the return of thinking more “MEGAHURTZ” marketing will work in this day and age. The only people who get Extreme Edition Piledivers are die-hard AMD fanboys that have far more money then sense (a smaller minority then the tiny Intel Extreme Edition crowd).

          • M3gatron
          • 8 years ago

          Yes man the problem is Intel cpus are reaching their limit. In the last 2 years we had only modest improvements. What are you going to do when yo cant improve the IPC anymore?? Raise the clock.
          AMD cpus already run at high frequencies so they only have to improve the IPC. And they have a lot of room for that.

            • chuckula
            • 8 years ago

            [quote<]AMD cpus already run at high frequencies so they only have to improve the IPC.[/quote<] LMAO... that's like saying that my little 4-banger already runs at high RPMs so the only thing I need to do to make it faster than a V-12 is to massively improve the horsepower.. what could be easier!

        • heinsj24
        • 8 years ago

        I was spoiled with the second stepping of Netburst. Nothing like a guaranteed 33% overclock out of the box with 3GHz parts like the P 530J, P 630, or the P-D 930.

    • DarkUltra
    • 8 years ago

    [quote<]The Core i7-4770K being tortured on my test rack peaks at 4.5-4.7GHz[/quote<] Aida64 cpu stress test and Unigine Nature benchmark is not sufficient to test stability. [quote<]honestly I do not truly find value in this Aida64 "stress test." The fact of the matter is that even though Aida64 shows my system to be "stable," this is not true. All you have to do is start Handbrake, and use it to encode a Bluray rip at 1080P and the system will hard-lock. Your mileage may vary, but I am going to stay old school in terms of how we decide if hardware is "stable."[/quote<] [url<]http://m.hardocp.com/article/2013/06/01/intel_haswell_i74770k_ipc_overclocking_review/7[/url<] old school he means prime95 blend + occt gpu There are also reports that the chips provided by intel overclocks much, much higher than regulars you find in stores and that Asus had to recalibrate their auto oc feature to regular retail versions. I hope Ivy Bridge E will truly be a worthwile upgrade to my sb-e...

    • yeeeeman
    • 8 years ago

    Well, AMD is not in a very bad position. The fact that a customer, in all his minds can’t discern the fact that a 180 bucks CPU which is the 8350 is a better choice than a 300 bucks cpu from Intel is sad. I always thought that the industry can throw extremely expensive products on the market and still have “customers”.

    That’s the situation with Intel and nVidia now.

    Nvidia, with the GTX780 is 15% better than the 7970. It costs 650 bucks compared to 450 for the 7970. But again, the customers still think that this 15% difference will make a matter when your card is under 25fps. Well, 15% difference at 25fps means 3.75fps. Do you really think that would make a difference in the visual experience you get in your game?

    With the CPUs is the same thing. The 3770K is faster in general, but we all do multitask. So the gap between the FX8350 and this expensive **** from Intel is almost null.

    You decide, who is guilty? AMD or the customers?

      • FubbHead
      • 8 years ago

      Brand recognition.. Probably the same reason why people buy Apple. 😉

      • Diplomacy42
      • 8 years ago

      so, you had the opportunity to use any of a dozen direct comparisons and take actual numbers and thought “nah, I’ll multiply 25 * .15 and pretend that this number has some bearing on reality.”

      gotcha

      • jihadjoe
      • 8 years ago

      Nvidia, with the GTX770 is equal than the 7970. It costs 400 bucks compared to 450 for the 7970.

        • l33t-g4m3r
        • 8 years ago

        Lies. I just bought a [url=http://www.newegg.com/Product/Product.aspx?Item=N82E16814127732<]MSI 7970 BE[/url<] (Ghz) for 399, and the regular version is $20 cheaper. The 770 is selling in the 450 range. Edit: There apparently are some 770's selling for 399 now, but there wasn't any when I bought my card. Also, I'd like to point out the compute performance deficiency of Kepler, and anyone who buys/bought a 680/770 over a cheaper or similar priced AMD card is an idiot. The 7950 is also a mid-range beast. If you buy nvidia, it's purely because of brand loyalty and nothing else.

          • MadManOriginal
          • 8 years ago

          Up to your old tricks again I see, using the worst possible pricing to try to make your point. There are numerous GTX 770 selling for $400: [url<]http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007709%20600451269%20600451260&IsNodeId=1&name=GeForce%20GTX%20770[/url<]

            • l33t-g4m3r
            • 8 years ago

            Uh, NO. I said there were some 770’s selling for $400. HOWEVER, there were NO 770’s selling for 400 when I first bought my 7970, and the 7970 is absolutely NOT selling for 450, unless you’re counting special overclocked editions.

            Even if both cards are selling for the same price, the 7970 is still a better / more rounded card. Not to mention all the next gen consoles will be optimizing for this chip. The 7970 is future proof, the 770 is not.

            If anything, the 7950 is the real card to get, provided you get a model that has boost clock and comes with a cooler good enough to get Ghz speeds. Nvidia doesn’t really have an answer to that card.

            • MadManOriginal
            • 8 years ago

            Editing after being called out makes it all good? 🙂

            You do get some credit for not editing out what you did first write at least, which was this:

            [quote<]The 770 is selling in the 450 range.[/quote<] I guess I'm just stupid, but the word IS threw me off, and you didn't say GTX 770s are selling for $400 until after I pointed it out. Rather than get all internet-ragey about it, why not just explain that when you bought your card (probably when the GTX 770 had *just* come out*) that wasn't the case?

            • l33t-g4m3r
            • 8 years ago

            Yes it does, because 1: I edited it BEFORE you posted. 2: It was 100% true at the time I bought my 7970. There were NO 770’s selling for 399. I decided to check this, and it had changed, so I edited my post appropriately. BEFORE YOU POSTED, BTW. Your comment is completely unwarranted, other than being sheer trolling. You’re grasping at staws, but I suppose that is to be expected from someone who has nothing better to do and is pro-nvidia.

            Shame on me for thinking people on the internet have any sense of decency. I should have just COMPLETELY REWRITTEN my original post. Then we wouldn’t even be having this discussion. Because you posted AFTER I edited it. Not to mention you don’t really have any point whatsoever.

            I’m the only one here that has an actual point. There is no price difference between either card now, it’s literally the same. The only reason someone would pick the 770 over the 7970 is brand loyalty, nothing more, and I know from hardware statistics there is an overabundance of nvidia fanboys.

            Nvidia is selling compute crippled chips, so they can up-sell cards like the Titan. (which still doesn’t beat a 7970 in compute) Why people would support that, I don’t know., I’d rather have a 3GB / 384bit 7950 instead of a bus crippled and compute limited 660Ti any day of the week, and same goes for the 7970 vs 770. Only the most hardcore mindnumb fanboy would buy a mid-range nvidia card at this point in time, when there is much better alternatives.

            I suppose the 770 is pretty equal on most fronts to the 7970, I just don’t like it’s compute performance, lesser memory, and the fact that amd has an edge with next gen consoles. Other than that, it is the best card nvidia has offered in this price range for a long time, and IMO it’s what the 670 should have been. Price and performance-wise.

            • flip-mode
            • 8 years ago

            You’re so charming and personable that I forget you’re full of poop.

            • James296
            • 8 years ago

            a bulls***er calling out a bulls***er, now that’s funny

            • PixelArmy
            • 8 years ago

            GTX 770 price has held constant since launch.
            [url<]http://www.priceneverlies.com/go3c/search.ldo?k=gtx+770[/url<] Furthermore, as far as I can tell, there were no outcries in the forums that would indicate prices being over the $400 MSRP, nor was there the supply shortages that generally cause this situation. Unless you are not in the US, the only plausible (though highly unlikely scenario) is that you were looking at the factory OC versions and that those were the only ones available.

          • Bensam123
          • 8 years ago

          The GE edition is just branding, you can unlock normal 7970s to GE speed with firmware upgrades (AMD provides and even encouraged when it was released). So you can basically get a GE for 7970 prices, a lot of people seem to be overlooking this in their comparison.

          GE release was supposed to completely replace normal 7970s, but manufacturers instead decided to segment it and it’s still sticking around.

      • flip-mode
      • 8 years ago

      You’re looking pretty guilty to me, sir. Either epic trolling or shameless shilling.

      • BaronMatrix
      • 8 years ago

      The sorry wimpy OEMs… They complain about margins and can’t figure out how to get more chips out of AMD…

      Who won’t usually ream them on pricing…

      ULV UltraBook chips require appendages…

      Suckers…

      • clone
      • 8 years ago

      the few ppl buying GTX 780’s (ultra high end makes up a fraction of the market) want more than an HD 7970 has to offer and they are willing to pay more, that’s not consumer ignorance that’s consumer preference.

      as for CPU’s, Intel’s K’s are interesting even after factoring in price, when I bought my FX 8320 it was the Intel K’s that I considered.

      as for “who is guilty” I’d argue you should say “who should I thank” with the answer being AMD not the customer because it was their mistakes and missteps that have forced them to compete on price for which I am grateful.

      had AMD not screwed up in so many ways we’d still be paying $400 for today’s version of the X2 3800 + and $1400 for the modern version of AMD’s FX 62.

    • DeadOfKnight
    • 8 years ago

    That’s gay, and not in a good way.

      • ronch
      • 8 years ago

      Yeah. As Neely said recently, the AMD FX is a MAN’s CPU.

      Get AMD FX and prove that you’re a REAL MAN!

      LOL

    • HisDivineOrder
    • 8 years ago

    [quote<]In the end, enthusiasts face a rather unfortunate set of choices in Intel's Haswell-based product offerings. We can't help but think this situation wouldn't exist if AMD were putting more competitive pressure on Intel.[/quote<] Omg. Where's the hate response that I get when I point out AMD's complete and total failure to compete as the single most important reason nVidia and Intel are MILKING their customer base without much meaningful performance improvement from a refresh (except nVidia's giving you lower prices for the same performance, which is... something)? Where is it? Suddenly, if TR says it, omg, it's true. When I said it months ago before it was slapping people in the face, it was all vitriol and hate and, "Omg, he's evil, he's anti-AMD." No. It's just fact. AMD is letting the whole industry down by NOT competing across all their markets. They're just sitting on their hands, becoming less and less relevant as time goes on, and the whole PC hobbyist industry will collapse into the aether unless they start getting active again. I can imagine a world where AMD does not exist and it is a world where nVidia releases a new Geforce product every two years with 10% performance gain over the last, higher prices than the last, and where Intel releases a new CPU every two years while doing tocks (or is it ticks, refreshes) for mobile products, eventually moving hobbyists to E-line products altogether that are marked up 2-5x the price for technology that's 1-3 years older than the main line. It won't matter much, though, because performance will have plateaued in favor of more and more power efficiency improvements and IGP improvements that do nothing for the enthusiast. Then one day, similar to how Intel ended their motherboard design team in favor of NUC, a cold day probably in November or early December when news and talk will be flying here and there so much they think it'll get lost in the din, Intel will quietly announce the dissolution of all LGA-spec motherboard chipsets and CPU's. Not long after, nVidia will announce the next Geforce product will be the last user-installed GPU. By then, AMD will be a distant memory, a subsidiary now of Qualcomm to manage their patents for GPU's and... well, yeah, GPU's. Within a few months of Intel and nVidia's announcements, the scraps of AMD within Qualcomm will be folded into Qualcomm proper and the AMD branding will be ended, similar to how AMD ended the ATI branding. PC gaming will live on in SOC's and pre-built units, but the hobbyists who have built the best damn PC's for decades now will be told, "Go buy a NUC." That's what I see in Galadriel's mirror if AMD doesn't get up off their encountered suits and do something.

      • trieste1s
      • 8 years ago

      Blaming AMD for what should be initiative on a consumer’s part?

      You vote Intel with your pocket but slag AMD for their inability to put their hands in the aforementioned pocket, and then whine all day long about Intel’s direction with Haswell.

      Nice.

      • MEATLOAF2
      • 8 years ago

      I agree with some of it, but Bensam123 has a great point, complaining about AMD’s irrelevance while (I’m just assuming here) buying Intel parts is not helping the situation. If you want to make a difference then buy AMD when they have a decent offering. They can’t make better products unless they have money for R&D, production etc. and Intel has a massive lead already anyway. If they have any chance at all to close the gap even a little, it’s right now. It’s a catch 22, they can’t make good products if they can’t sell the ones they have now, but nobody wants to buy the ones they have now because they aren’t as good as Intel.

      As for their GPUs, they aren’t doing so bad performance wise, they have good products in that market. I can see that falling apart if they can’t get their CPU side of things going well however.

      If the market goes the way you say it will (pretty likely in my opinion, at least some of it anyway), the fault will not be on Intel or Nvidia, it will be on us, and us alone. You should be thanking the people that have been buying their stuff up until this point though, they have been taking one for the team.

        • Klimax
        • 8 years ago

        Sorry, but we won’t buy bad product, just to help somebody and reward them for bad decisions.

        This is market, not charity…

          • Spunjji
          • 8 years ago

          That’s not the point being made here. It’s that to buy Intel whilst castigating AMD for not being able to afford to develop products you want is a stupid thing to do. If you genuinely believe in a balanced market you have to play a role in that; if you don’t then buy whatever you want, that’s your choice.

      • Antimatter
      • 8 years ago

      Why would AMD willfully cede sales to its competitors? The reality is that it’s not easy designing CPUs and GPUs particularly with the limited resources AMD has taking on juggernauts of the industry, even more so when they play dirty. I suspect AMD is doing their best to compete, all we can hope is that their best is good enough.

        • nanoflower
        • 8 years ago

        I’m not sure that it’s even the issue of designing CPUs. How well can AMD compete when part of Intel’s advantage is the process advantage they have with their own foundries? I wonder just what sort of competitive product could AMD come up with if they had unlimited funding but still had to rely on TSMC and Global Foundries to actually produce the chips. Could they make a CPU that provides great performance and a low TDP like Haswell?

          • faramir
          • 8 years ago

          GlobalFoundries is well versed in 32 nm manufacturing now, wouldn’t you agree ? Yet AMD still doesn’t make CPUs that could match Sandy Bridge (also 32 nm) performance, despite running at much higher clocks.

          Hopefully Steamroller will change that as I will be upgrading my computer and if top Kaveri A10 APU (at 100W) brings performance of Sandy Bridge i5 with equivalent GPU performance of Radeon HD5770 glued to it, they will definitely get my money.

      • M3gatron
      • 8 years ago

      Imagine an Intel executive reading this comment. So Intel offers less and less to their costumers and who do they blame? AMD.
      No competition my ass people like this are responsible for Intel’s behavior. I hope AMD goes bankrupt so you would be able to buy a 200$ celeron.

      • Klimax
      • 8 years ago

      You can pin on failure of competition only product line-up. Performance is different case. Completely different case. I tried to point it out, but people failed to understand.

      We are hitting limits of math and algorithms. The only remaining way to significantly upgrade performance is through great increase in complexity of chips or by inventing new ways to do things like new algorithms for basic math (integer/floating point multiplication and division).

      We are already way past diminishing returns from investment.

      No matter what, everybody hits that ceiling. Intel, AMD, ARM, everybody. The only question is who and when and how they will try to break that. (Example could be quantum computing or new approaches)

      • Spunjji
      • 8 years ago

      I think you’ll find it has a lot to do with a little thing called tone.

      • cynan
      • 8 years ago

      When in the last 5 years (since June, 2008 when the HD 4000 series came out) has AMD not been competitive for discrete GPUs? Even if their CPU woes doe end up final dragging them down, is it not reasonable to expect another company to come in and buy what’s left of ATI and carry on with what is essentially the most promising component of AMD’s current portfolio?

      And who cares that AMD sold ATI’s then mobile GPU tech to Qualcomm. That was a long time ago now. Maybe it was a shortsighted move, but there’s no reason to think that AMD is out of contention here. Starting from scratch making more efficient GCN architecture for mobile use (as Nvidia is doing – Tegra 5 graphics will be based on Kepler architecture) is not that unfeasible. AMD has experience with this already in making some of its lower power APUs, not to mention the chips in the new consoles.

      • NeelyCam
      • 8 years ago

      Your TL;DR post made me think of something. Intel doesn’t compete in high-end discrete GPUs – that’s all nVidia/AMD. Intel has said that they are open for providing foundry services for companies that don’t compete with Intel.

      So, ARM-based systems are out, but wouldn’t it be possible that Intel could agree to fab high-end GPUs to either nVidia or AMD? Whoever gets an exclusive contract with Intel could gain a major advantage over the other that’s stuck on TSMC..?

      Plausible..?

        • PixelArmy
        • 8 years ago

        Would be a lawsuit waiting to happen…

        AMD w/ Intel would be a conflict of interest. Imagine if something went wrong with Intel fabbing AMDs chips.
        nvidia w/ Intel at some point would probably be accused of price fixing/dumping (I’m assuming Intel fabs would result in more cost efficient yields and higher performing parts).

    • Bensam123
    • 8 years ago

    “We can’t help but think this situation wouldn’t exist if AMD were putting more competitive pressure on Intel.”

    Pretty much… When they get too far ahead, they get lazy or they try to exploit what they have. Can’t wait to see what Steamroller has in store.

    At least non-K users can still increase their bclk by +3-5 for a good 60mhz increase and risk of data corruption. XD

      • 0g1
      • 8 years ago

      Once Steamroller comes out — early next year — Intel will probably unleash their Socket 2011 Ivy Bridge 12 core part.

        • faramir
        • 8 years ago

        Ivy Bridge-E will be 4-6 cores. Perhaps you’re confusing cores with threads (hyper-threading) ?

          • mczak
          • 8 years ago

          Ivy-Bridge-E will have 12 cores (maximum) there’s no doubt about it. Now if you talk about the _desktop_ versions of Ivy-Bridge-E, yes they will be limited to 6 cores. And frankly I don’t think Steamroller will put enough pressure on intel so they feel like releasing a 8 core desktop version but nothing would stop them from doing so (after all they already have the exact same chips sold as xeons).

            • cynan
            • 8 years ago

            The desktop version [i<]is[/i<] IB-E. Those other CPUs are called Xeons...

            • madmilk
            • 8 years ago

            Well, they’re actually called Ivy Bridge-EP, if you’re going by that naming system.

      • DavidC1
      • 8 years ago

      Nonsense.

      Maybe you should move on to the fact that most of the world is about being more mobile than ever, and Desktops are decreasing every day. Sure, some people will continue to use it, but from a revenue point of view, less and less people will upgrade and keep their systems longer and longer.

      AMD does have a reason to continue Desktop development for a bit longer, because their revenue is greater than Desktops compared to Laptops. But for Intel, Desktops are less than 30%, with bulk of their microprocessor revenue from Servers and Mobile increasing fast.

      Whether with Ultrabooks, or Tablets and Smartphones, Intel’s non-server PC market is laser focused there. And they need to, because ARM and the companies that use them are their new competition. But keep imagining ever dwindling company called AMD is their biggest threat.

        • DeadOfKnight
        • 8 years ago

        Don’t forget that AMD is now partnered with ARM and promises new products in the future.

          • Klimax
          • 8 years ago

          ARM section of markets is already full of sharks and bodies. AMD is just another.

            • DavidC1
            • 8 years ago

            Exactly.

            AMD is not only low on resources, both money and talent, but spreading itself even thinner by going for ARM. Good luck on that working out.

            • NeelyCam
            • 8 years ago

            Agreed. Overall, I think Rory Read has managed a very difficult situation exceedingly well, but spending resources on ARM is an unusual misstep.

            • Bensam123
            • 8 years ago

            Perhaps I see a partnership different then you do…

            Here I thought the idea was to share technology and expertise, not compete with each other. I don’t think AMD has any plans to compete directly with ARM, especially while being partnered with them.

        • Bensam123
        • 8 years ago

        Desktops aren’t decreasing everyday, someone else tried to make this point too… Perhaps mobile devices are increasing, in terms of cell phones and maybe tablets, but they aren’t outright replacing desktops anymore. Those that did have already. Most people have a laptop and a desktop for various tasks. Corporations still use desktops, I still use a desktop and have a laptop.

        Someone also pointed out people wont upgrade their current system because haswell really make no difference performance wise, so there is no reason to upgrade their system. Why would you upgrade to a component with almost identical performance specs?

        If Intel was being smart about this, they would consider spinning off their CPUs into a mobile and desktop variants and tailor each accordingly, perhaps cross breeding them when they hit some good R&D that would help both out. Instead this time around we got mobile chips for the desktop.

        Completely shunning one part of the market isn’t a good way to entertain them.

    • travbrad
    • 8 years ago

    So I guess there will be a lot more K processors in the next system guide? :p

      • DeadOfKnight
      • 8 years ago

      Maybe, maybe not. The review samples seem to do an OK job at overclocking, but if you read customer reviews you’ll find that most of them can’t get over 4.3 GHz on air stable. The “K” SKU still leaves much to be desired. Honestly, I think the only good reason to update the guide to Haswell is the outstanding motherboard selection.

    • ALiLPinkMonster
    • 8 years ago

    How much of that heat output does the IGP account for? If it’s any significant amount (and if Intel wasn’t nice enough to disable it when it’s not being used), then they should sell -K parts without the IGP. Not factory rejects with a disabled component, but fully functional chips. If priced right, they could sell really well. Not that they have any reason to TRY and sell chips, but it would be a nice addition to the lineup.

      • DavidC1
      • 8 years ago

      Uhh, zero?

      Modern CPUs have advanced power management features like Power Gating, where it nearly completely turns off individual cores. There would be no reason for an iGPU to contribute to heat.

      In fact, having a turned off die might actually help dissipate the heat for the cores, because they are cool.

        • ALiLPinkMonster
        • 8 years ago

        I figured as much. I still think it could be a popular product if the lack of an IGP would lower the price enough.

          • DavidC1
          • 8 years ago

          That’s also wrong. iGPU development costs are paid off by people wanting those setups. Think of if Intel did not create the integrated GPU at all. Then people would be buying AMD, meaning its losses in sales for Intel.

      • M3gatron
      • 8 years ago

      AMD managed to improve their cpu and igp speeds with the Richland APU without making is run hotter or consume more power.
      Are you saying Intel wasn’t able to do this with their 22nm process?

        • DavidC1
        • 8 years ago

        That’s funny, because whatever gains Intel had going from 3770K to 4770K, Richland had fraction of that.

        [url<]http://hexus.net/tech/reviews/cpu/56657-amd-a10-6800k-32nm-richland/?page=3[/url<] That's true with iGPU as well. It's funny how the % C-P-U gains 4770K had over 3770K were better than Richland had under i-G-P-U gain. That's pretty pathetic. Also, while the CPU is showing higher power use, overall power went down in lot of cases.

          • pumero
          • 8 years ago

          You can’t compare the the transition from Ivy to Haswell with Trinity and Richland.
          Haswell has architectural changes over Ivy, while Richland is nothing more than a “fully unlocked” version of the one year old Trinity core.

          The finer power states and the FMA3 support were probably already inside Trinity right from the beginning and were just “unlocked” for Richland, so that they can state that they changed anything apart from process optimization.
          AMD knew that Kaveri would be ready for release just 18-24 months after Trinity and with those little changes they can stay on the yearly refresh cadence more or less dictated by Intel. With Richland they got a fresh round of reviews, while nobody would have looked at Trinity again with the release of Haswell.

          The transition to Kaveri will be the one to look at and only then you can compare the gains between those two cores versus the gains Intel had with Haswell.

            • M3gatron
            • 8 years ago

            I was comparing the improvements.
            Richland has better clocks for the igp and cpu then his predecessor but the temps and consumption haven’t changed. Plus they improved the overclocking ability of the chip. That is done on the same architecture and manufacturing process.
            Intel didn’t manage with Haswell to improve temps, clocks, consumption is improved only in some areas and situations in others is even worse.
            And Trinity cpus were not locked in any way that is just your assumption. GF’s 32nm has improved that is why the managed to raise the clocks with Richland and that is why they are able to release a 5GHz CPU.
            I just assumed that Intel’s 22nm will do the same and Haswell will be like Sandy in terms of OC and temperature. But is not.

            • DavidC1
            • 8 years ago

            You do see the huge gains, just not in the places you want them to be.

            Intel’s focus is on the Ultrabooks, Servers, and Tablets.

            I wouldn’t be surprised if Haswell’s lack of gains on the Desktops are a direct consequence of Intel engineers focusing it for best perf/power. That happened even on the process side where 22nm offers little gains in drive current with higher voltages.

            10 years ago with Pentium M, Intel started making chips based on Standard Voltage laptop chips, the Desktop chips being a derivative of it. Now, 4th Gen core takes the optimization point to ULT and Y class chips, meaning Desktop chips are even further away from it. The voltage and power Desktop performance is needed isn’t as optimal as the ULT chips are at.

            • M3gatron
            • 8 years ago

            You do understand that when I want to buy a desktop cpu I don’t care that is not as good as it should because it’s designed for mobile.
            And I not that impressed with mobile Haswell either.
            They managed to improve idle consumption mostly. If you plug your laptop into the wall you don’t see any performance boost or lower power consumption.
            And I only saw a poor test as they didn’t compare similar laptops and they haven’t analyzed the efficiency of the device. A newer laptop should be more power efficient than an old one even if it uses almost the same configuration.

          • M3gatron
          • 8 years ago

          The 6800k is just a 5800k with 300mhz more plus it overclocks higher. Some tests show bigger improvements some none although you did link me to a page with synthetic benchmarks only. That review is poor quality.
          You can also easily clock the 6800k at 4.4ghz with minimal power and temp impact and have a similar improvement Haswell had with a new architecture over the last gen.

    • ronch
    • 8 years ago

    Intel sure is making it difficult for their fanbois by using their performance and power efficiency advantage. More performance and more efficiency in exchange for higher CPU and platform costs and castrated features. You gotta love Intel for that. Reminds me of China. China’s on a roll for the past few years but uses their economic might to bully little countries around Asia. (Before the downthumbs roll, verify it first and see if no one in Asia’s getting pissed about it.) Yeah, the US does that too, but China does it a little bit… ‘differently’.

    On the good side, this improves AMD’s value proposition. All FX CPUs unlocked across the board, no disabled features or instruction set extensions.. they make [u<][b<]REAL CPUs for REAL MEN[/b<][/u<]. Somehow you gotta wonder whether Intel is deliberately limiting their products' appeal to give AMD a breather and at the same time compensating for it by having higher ASPs.

      • heinsj24
      • 8 years ago

      I don’t think Intel cares anymore about AMD. And they certainly never cared about enthusiasts – that’s why they make their most expensive processors within a product family the “Enthusiast” cpu. Gone are the days of taking a second rate processor and overclocking it to outperform the flagship model for a given Intel socket. Now you’re buying the flagship and losing features to boot.

      This is one of the reason I stick with AMD – I don’t have to buy the top-of-the-line processor for performance or features.

        • DavidC1
        • 8 years ago

        Maybe you should think of the fact that computers are cheaper than EVER to buy. Remember how on the same socket the X6800 was at $1000, and $300 CPUs were 2 grades lower?

        Also, there’s something called “inflation”.

        • ronch
        • 8 years ago

        Yup. Totally happy with my FX-8350. A very solid CPU for just $180.

    • mcnabney
    • 8 years ago

    I’m getting less and less thrilled by Haswell every day now.

    And AMD would likely have been a far more formidable competitor if they actually got to make some money and sell some chips back in the pre-Core2 days when they had far superior kit. We know now that they were illegally locked out of major sellers. That few billion settlement that came years later didn’t even come close to replacing the huge loss of development revenue and momentum they had going.

      • NeelyCam
      • 8 years ago

      [quote<]We know now that they were illegally locked out of major sellers.[/quote<] There have been claims by AMD about this, but we don't know if it's true.

        • willmore
        • 8 years ago

        You’re right, dell paid the SEC $100 Million in fines (but admitted no wrong doing) over nothing, nothing at all, they just thought the SEC needed some money and, being good citizens, they thought they should chip in a little.

        • mcnabney
        • 8 years ago

        Yeah, this is pretty much common knowledge. It was suspected a decade ago, but it has been confirmed for a long time now.

      • ronch
      • 8 years ago

      This is one of the reasons why I’m part of the rebellion. AMD is about rebellion. It’s about freedom.

      /sarcasm with half-truths

      • Game_boy
      • 8 years ago

      AMD were supply constrained. They could not have made or sold more chips even in a fair market.

        • swaaye
        • 8 years ago

        I remember reading this was a reason they didn’t get in the original Xbox as initially rumored. It also happened with their Llano APU and Apple apparently.

        • Antimatter
        • 8 years ago

        That may be true but didn’t the licensing agreement with Intel also prevent AMD from manufacturing CPUs elsewhere?

          • OneArmedScissor
          • 8 years ago

          If that were the case, it would be their own fault for signing the contract.

          But then we get more into politically fabricated legal issues of the patent system, rather than whether or not Intel intentionally wronged someone.

        • ronch
        • 8 years ago

        Hector talked about ‘sure deals’ disappearing overnight. It’s like, an OEM has issued a purchase order one day and abruptly cancels it the following day. Something like that. Anybody who becomes the victim of such bewildering and suspicious practices never forgets it. It’s similar to how Microsoft was originally gonna use the Athlon in its first Xbox (the REAL Xbox One) then Intel swoops in and steals the deal by giving Microsoft a ‘presentation’.

        • Spunjji
        • 8 years ago

        Up-voted, because to an extent this is true – it was one of several factors that stunted their growth in the desktop sector. It wasn’t the key factor in the crucially profitable server sector, though.

          • OneArmedScissor
          • 8 years ago

          It also kept them out of laptops just about altogether, which, in the long run, probably hurt them most.

          I remember it being easy enough to find OEM Athlon 64 desktops at the time they were supply constrained, but for a laptop, you pretty much had to go to one of the custom places and pay out the wazzoo.

      • Phr3dly
      • 8 years ago

      AMD’s profligate spending under Sanders may have also contributed to their demise.

      • Bensam123
      • 8 years ago

      Or brand name recognition and fan base…

      • Deanjo
      • 8 years ago

      Sorry but AMD started losing big wind out of their sails when Ruiz snubbed the desktop enthusiasts and put everyone on concentrating making headway into the server realm where they had a bit do success for the short term. That was the beginning of the downfall. All of that effort put AMD relying on their old K8’s for far too long.

    • derFunkenstein
    • 8 years ago

    Booooo!

    But then again, how often did someone buy a non-K CPU and take advantage of that? How often were non-K CPUs paired up with cheaper chipsets? I guess it matters how many people are affected. I’m not entirely sure that we’re losing too much here.

      • mcnabney
      • 8 years ago

      People that want to overclock are going to be crippled in the future as uses for TSX actually get written into software. So in theory the non-k models should have longer legs.

      • ronch
      • 8 years ago

      I don’t have a problem with non-K chips being fully locked. Heck, one deserves every bit of it by not getting a K CPU. What irks me is the fact that TSX is disabled on K CPUs. Intel sure has lots of levers to push and pull on each of their chips.

        • kroker
        • 8 years ago

        I do have a problem with non-K chips being fully locked. Considering that you can’t really overclock K chips that much anyway without expensive cooling setups and/or delidding, and you also lose some some features like TSX and VT-d, and that they’re also a bit more expesive than non-K chips, I would have been much more tempted to buy a non-K chip if I could get at least an extra 10-15% of performance when needed without the need for exotic cooling setups and insane practices like delidding a chip that costs half my monthly wage (yes, I live in a developing country with high VAT). I could never have gotten as much value for money from my current Q6600 if I couldn’t overclock it.

          • cosminmcm
          • 8 years ago

          I agree, I wanted to buy an i7-4770S for an itx build and was hoping for that little overclocking, but now I am really disappointed. I will still buy it, but not with the same joy that I had before this.

            • BIF
            • 8 years ago

            [quote<]"I will still buy it, but not with the same joy that I had before this." [/quote<] But with the same money. That's all Intel cares about, really; and with no real competition, they get what they love most. Your money and AMD's market share. Pretty soon, Intel could just sell you a spider-webby 486/DX and call it a day. And you'll buy it. And love it. With a joyless kind of love, of course...

            • clone
            • 8 years ago

            “I will still buy it, but not with the same joy that I had before this.”

            why would you buy a non K Haswell if you were hoping for a little overclocking capability, is that 7watt power savings at stock clock that important as opposed to grabbing an Ivy or a K series?

            your interest in even a mild additional overclock would seem to refute that.

          • Diplomacy42
          • 8 years ago

          so you’d rather they come out with a 2.8ghz chip that is overclockable for 400mhz, rather than get a chip that is fully optimized for the speed and power its using?

        • Krogoth
        • 8 years ago

        TSX and VT-D were disabled on K series, because Intel wants to justify the Extreme Edition which has all of the bells and whistles.

        Besides, you aren’t going miss either feature since TSX and VT-D are only useful for professionals and enterprise types where overclocking is a big no.

          • ronch
          • 8 years ago

          TSX is one of the exciting new features of Haswell. Including it wouldn’t hurt, the same way AMD includes all features across all its FX models. There are still folks who may want to play with some features. Take virtualization, for example. Most folks probably won’t use it; it’s usually used in servers, but it wouldn’t hurt to have it, would it?

          I guess this is the product of a strong Intel and a weak AMD. Intel can afford to throw a few levers here and there while AMD needs to sweeten the deal by being generous with the extras.

            • Krogoth
            • 8 years ago

            TSX only yields tangible benefits with applications that are massively threaded a.k.a servers/workstation. Customers of these markets going to get Xeon verison of Haswell or regular desktop version if budget is a concern.

            It is kinda sad that fanboys are grasping straws on features that yield little or no benefit to them.

            • ronch
            • 8 years ago

            It would still be nice to have. I’m using an FX-8350 and I’m happy to know it has all features turned on. I know I won’t easily find software that uses FMA 3 or 4, AES-NI, XOP, AVX, F16C, or even use virtualization, but it’s all there.

            • Waco
            • 8 years ago

            Or, you know, future games (assuming anyone jumps on it).

            • chuckula
            • 8 years ago

            What’s hilarious is that TSX is available on zero AMD CPUs and isn’t even announced as a feature for Steamroller. If Intel had bothered to put TSX into every single new chip it makes, then TSX would just be another “meh” feature that doesn’t help in the real world.

            BUT… TSX became this amazing feature the nanosecond that Intel decided not to put it into every single chip from low-end Atoms all the way up the line. From what I can tell, all Intel has to do to make anything for its chips amazing is to segment the market. At that point, it suddenly becomes insanely unfair that your overclocking game machine doesn’t have features for lock elision on server workloads.

            • clone
            • 8 years ago

            to say AMD doesn’t have TSX is to say Intel doesn’t have ASF because arguably both companies have chosen different paths.

            the problem with this being one or both features could go ignored much like Nvidia’s PhsyX.

      • DPete27
      • 8 years ago

      I have to regretfully agree with you on that. Even here on the TR forums where there is a much higher concentration of “techincally savvy” visitors, I don’t think many people really knew that [url=https://techreport.com/forums/viewtopic.php?f=4&t=87608<]overclocking non-K Intel CPUs[/url<] was even possible. With the limited overclockability of Haswell and the introduction of BCLK strapping, I was ready to declare that [url=https://techreport.com/forums/viewtopic.php?f=2&t=88214<]there's little motivation to buy K-series Intel CPUs anymore.[/url<] Intel obviously saw this potential so they took away everything....Dreams crushed.

      • ochentay4
      • 8 years ago

      I do. I have a 3770 and I enjoyed paying it really cheap (3770K was terribly overpriced at the time). I have it at 3.9 and have been using a cheap air cooler. Also, I use VT-d. Locking people out of features sucks really bad specially when its arbitrary. I want a full featured 4770K with TSX extensions VT-d and vPro. Or an overclockable 4770.

Pin It on Pinterest

Share This