Rumor: Core i9-9900K could hit 4.7 GHz on all cores and use a soldered IHS

Not that long after Intel apparently tipped off the internet about the existence of its ninth-gen Core CPUs, there are some more rumors flying about the company's latest developments in the desktop CPU space. Grab your salt shaker and proceed. According to Computerbase, a now-defunct thread in Chinese forum Coolaler contained specifications on rumored upcoming Core i9-9900K chips, Core i7-9700K, and Core i5-9600K chips. More recently, German site Golem.de claims to have obtained information stating that the two higher-end models will use solder instead of thermal pads under their heatspreaders. Let's look at the rumored specs first.

  Core i9-9900K Core i7-9700K Core i5-9600K
Cores 8 8 6
Threads 16 8 6
Base clock 3.6 GHz 3.6 GHz 3.7 GHz
Turbo clock 5 GHz 4.9 GHz 4.6 GHz
Cache size 16 MB 12 MB 9 MB
One-core turbo 5.0 GHz 4.9 GHz 4.6 GHz
Two-core turbo 5.0 GHz 4.8 GHz 4.5 GHz
Four-core turbo 4.8 GHz 4.7 GHz 4.4 GHz
Six-core turbo 4.7 GHz 4.6 GHz 4.3 GHz
Eight-core turbo 4.7 GHz 4.6 GHz
TDP 95 W 95 W 95 W

Source: Coolaler forum

Those could be some mighty chips indeed, and the Core i9-9900K in particular bears some resemblance to a BFG with that 4.7-GHz all-turbo clock. While the specifications above are nothing but hearsay at this point, they look at least reasonable enough in light of the Core i7-8700K's six cores, twelve threads, and 4.7-GHz single-core turbo speed. The oddball chip in the purported lineup looks to be the Core i7-9700K. Breaking with the Core i7 tradition, this CPU would boast eight cores but no Hyper-Threading support, along with a reduced amount of cache and slightly slower clocks compared to the Core i9-9900K big honcho. Overclock3D further reports that chip already showed up in the SiSoftware Database, too.

The rumored move back to soldered integrated heatspreaders (IHS) for the Core i7-9700K and Core i9-9900K is a welcome one. Intel hasn't used solder in its desktop Core-series chips ever since Sandy Bridge seven years ago, and enthusiasts everywhere are certainly looking upon this development with interest. The soldered IHS could improve the rumored chips' thermals and out-of-the-box overclocking potential, as well as obviate the need for delidding for those looking for an extra clock jolt. A quick search reveals scores of threads claiming that delidding a modern-day high-end CPU can net temperature improvements of as much as 20° C. Our own Zak Killian has joined in on the delidding fun, too, and he says it's as cool as people say.

Golem.de goes on to claim that the new chips will land in stores in September and that they'll drop into current-gen boards with an LGA 1151 v2 socket. If this information is accurate, it looks like enthusiasts won't have long to wait. Thanks to HotHardware and VideoCardz for the tips.

Comments closed
    • Sahrin
    • 1 year ago

    Wow..if the 8-core CCX rumors are true Intel is so fucked.

      • chuckula
      • 1 year ago

      Ready for it with 8 core Ice lake parts that have integrated graphics too.

        • Goty
        • 1 year ago

        … in 2020?

    • Jigar
    • 1 year ago

    Straight from Anandtech –

    TDP is calculated at Base clock

    [url<]https://twitter.com/IanCutress/status/1016268206447751168[/url<]

    • elites2012
    • 1 year ago

    time to revamp the chips from ground up. yall still using old coding. AMD has changed things from ground up. price it just right and people may buy it. price it wrong and only ur oem companies will look at it.

      • Ninjitsu
      • 1 year ago

      [quote<]yall still using old coding[/quote<] lol

    • Shinare
    • 1 year ago

    Here’s a big thumbs up for the use of the word “obviate”.

    Well produced, sir!

    • Puiucs
    • 1 year ago

    Yeah no. This smells really fake.

    • Noinoi
    • 1 year ago

    Some very naive maths here, but it seems like the 9700K would be a really nice upgrade compared to simply hopping into the Coffee Lake.

    I use an i5-4590 with MCE turned on, and I’m probably going to do the same to the future system, so let’s start with the clock speed.

    4.9 GHz ÷ 3.7 GHz = a 32.4% uplift by clock speed alone.

    Then you double the amount of cores (8 ÷ 4 = 2).

    A 164.8% aggregate multi-core performance increase and 32.4% single makes my i5 look really bad. I should run a timing test on my Unity compiles to see if the CPU is holding up and causing it to commit slowly.

    That soldered IHS, if true, would also mean that my existing H60 is likely to be fine despite the increase in heat output.

    It’s a shame that I’d have to sell off my motherboard and RAM no matter what upgrade path I take, though, be it Intel or AMD…

    • Jigar
    • 1 year ago

    So many people here think that Intel chips are using 95W @ 4.7GHZ, its just sad. Intel calculates TDP at base clock i.e 3.6GHZ.

    Straight from Anandtech –

    TDP is calculated at Base clock

    [url<]https://twitter.com/IanCutress/status/1016268206447751168[/url<]

      • Noinoi
      • 1 year ago

      Considering that not all CPU workloads are equally as straining, I expect the CPU to be somewhere in between the base clock and the specified all-core turbo in practice when not running a power virus. AVX2 burn is a very real thing but most workloads don’t use it, and it also happens that stress-tests usually use it.

        • DPete27
        • 1 year ago

        On the desktop? My i5-3570K always maintains its all-core turbo, even when running stress test.

        In laptops your statement is probably correct, but only for the best cooled products.

      • Klimax
      • 1 year ago

      Got evidence?

        • Spunjji
        • 1 year ago

        Circumstantial evidence would be the behaviour of their mobile chips, which bust a hole in their TDP as soon as you ask them to do anything remotely interesting.

        *Edit* – see the power values for the 6-core i9 in the new Macbook Pro. Intel’s specifications allow it to suck down a solid 125W at max-turbo, with a steady-state turbo limit of 100W. There’s no way on this fine green Earth that a desktop-grade chip on the same manufacturing process with 2 more cores and 4 more threads will somehow require less power to run at a higher speed.

        • Ninjitsu
        • 1 year ago

        I remember Turbo Boost going past TDP.

        • Jigar
        • 1 year ago

        Couple of days i was speaking with Ian Cutress from Anandtech on twitter, i have posted the link to that twitter.

      • Chrispy_
      • 1 year ago

      All I can tell you about Intel’s TDP is that our UPS claims a stack of eight Broadwell-E 6850K machines, plus a rackmount KVM console consumes 625W at idle and 1950W under load, to an accuracy of +/- 25W.

      The only thing changing is the CPU load, so the delta between idle and load of a single 140W TDP 6850K is 166W and you can be 100% sure it’s not using zero Watts when idle.

      I’d guess that the idle is 15-20W so so the all-core turbo of the “140W chip” is ~185W multiplied by around 0.92 for PSU efficiency loss, so about 170W.

        • thecoldanddarkone
        • 1 year ago

        What about VRM’s and Fans?

          • chuckula
          • 1 year ago

          And that assuming even an extremely optimistic 95% efficiency on all power supplies there’s

          (1950 – 625) * 0.05 = 66 watts gone just in the power supply consumption.

          Make it a more realistic but still strong 90% efficiency and that’s over 130 watts.

            • Chrispy_
            • 1 year ago

            Indeed. They’re platinum supplies at 92% or better and I’ve already accounted for that, bringing the estimate from 185W to 170W per CPU.

            All fans are constant (it’s in a noisy datacenter room) except CPU fans and we’re talking about a single 120mm fan difference between low RPM and high RPM. It’s going to be +/- 1W at most.

            I hadn’t considered VRMs. What sort of efficiency loss do they have? Is it really a [s<]40W[/s<] 30W loss to drive a [s<]130W[/s<] 140W chip? I though VRMs were in the low-teens for high-end boards with ridiculous overclocks. [i<]edit - brainfart on Haswell-E TDP of 140W[/i<]

          • Chrispy_
          • 1 year ago

          See my reply to Chuckula. I hadn’t accounted for VRMs and I’m not pretending I know.

          Would VRM’s lose 30W for a 140W chip?

            • liquidsquid
            • 1 year ago

            Multi-phase regulators for chip cores are insanely efficient. They have to be as to not add too much heat to the immediate area around the CPU. Figure better than 98% efficient. Likely 99%+. That is why there are all of those inductors and solid polymer caps there to aid in performance.

        • derFunkenstein
        • 1 year ago

        What’s the efficiency on the power supplies, though? Some of that energy is lost. Plus, some of that energy is consumed by memory and storage.

          • Chrispy_
          • 1 year ago

          I already covered PSU efficiency at 0.92 (80+ Platinum).
          DIMM slots count as CPU TDP because the memory controller is part of the CPU.
          Storage is an SDCard or onboard GBE NIC for each node. We’re talking [i<]milliWatts[/i<] here, so negligible.

            • smilingcrow
            • 1 year ago

            “DIMM slots count as CPU TDP because the memory controller is part of the CPU.”
            But what does the actual RAM consume under load as that is not part of the CPU’s TDP?

            Plus any storage in use and system and other chipsets will use more power when in use.

            • Waco
            • 1 year ago

            DIMM slots do *not* count towards the CPU TDP, they are fed by the mainboard…

    • ronch
    • 1 year ago

    CPU wars are here again.

    • Unknown-Error
    • 1 year ago

    4.7 GHz on “All-eight-cores” @ just 95W? YIKES! Now that IS impressive.

      • Jigar
      • 1 year ago

      Correction, its 95W @ 3.6 GHZ, Intel doesn’t tell you how much their chips sucks when they are using turbo speed.

    • synthtel2
    • 1 year ago

    4.7 on 8C in a 95W TDP sounds like it’ll either need 14+++ or be very tough to maintain (or maybe both). That is a particularly large gap between base and all-core turbo, hinting at the latter.

    Disabling SMT before cores seems like an increasingly odd choice at this level. A lot of that is probably just that the workloads that still give me trouble are low-IPC ones, but even if 6C12T and 8C8T were exactly equal in net performance, 6C12T would have advantages in power management and binning.

    I had been suspecting that Intel had dropped soldered IHSes for reliability reasons, and that Zen wasn’t having that problem due to its lower power density. (It could have been worth some engineering effort for the sake of reduced power consumption, even if it ended up limiting something else a bit; from what I’ve seen, GloFo 14’s leakage-vs-temperature curve is fairly steep.) If Intel plays their typical TDP-disrespecting game with a 4.7@8C part with a soldered IHS, that’ll definitely be interesting.

      • Redocbew
      • 1 year ago

      If there is an i7 with a soldered IHS in the works here, then I think it’s a pretty safe bet it’s being done for a reason other than “become some people want it”. I would guess it’s just one thing in their bag of tricks that allows them to hit their target TDP.

        • synthtel2
        • 1 year ago

        I’m also assuming there’s a reason they’ve been sticking with the non-solder option on their top parts so far. Cost alone wouldn’t seem to explain that one.

        No doubt, solder will improve a chip’s performance in some obvious ways. The question is whether Intel’s willing to sacrifice anything less-obvious to get those obvious gains, and given the state of their lineup, I wouldn’t put it past them.

        • Spunjji
        • 1 year ago

        Soldering the IHS won’t do squat for the TDP, though. Well, it’ll help fractionally under maximum load, but realistically speaking TDP is about how much power goes *into* the CPU, not how efficiently you can get that power out of it. Using toothpaste TIM could actually help a CPU stick to its TDP by forcing it to throttle… but that’s a bit like slowing your car down by slashing the tyres.

          • synthtel2
          • 1 year ago

          Leakage current increases significantly at higher temperatures, even if you’re holding clocks and voltages the same.

            • techguy
            • 1 year ago

            This is true, and the principle can be observed via overclocking endeavors. Better cooling yields lower power consumption, and higher clocks (potentially).

      • mczak
      • 1 year ago

      There is imho no way it’ll reach 4.7Ghz on all cores with any kind of stressful load at 95W.
      My guess though would be that it’ll simply ignore the TDP limit. Just like the i7-8700k and i7-8700 pretty much do for the most part already on most boards. (The i7-8700 is supposed to have 65W TDP, but in practice it’ll reach pretty much exactly the same frequencies as the i7-8700k, and consequently draw the same 100+W too – this is apparently considered acceptable behavior by intel, only a few boards actually have a bios which will enforce the TDP limit.)
      The soldered IHS should easily give the ability to keep it at reasonable temperatures even when it’s drawing 150+W (which was a problem with the i7-8700k).

      • psuedonymous
      • 1 year ago

      [quote<]Disabling SMT before cores seems like an increasingly odd choice at this level.[/quote<] It seems like an obvious choice when chasing clock speed. Assume a workload that scales perfectly with hyperthreading (i.e. highly threaded, little interdependence). If you have 4 cores and hyperthreading, you are doing X watts of work spread over Y die area. If you have 8 cores and no hyperthreading, you are doing X watts of work spread over 2Y die area. Lower areal heating means per-core temperatures will be comparatively lower.

        • synthtel2
        • 1 year ago

        Given 2018’s power management tech, we always have to assume that a core can be fully active (and burning full power) regardless of whether it’s being fed by one thread or two. SMT will increase average-case power consumption a bit, but not the worst-case that clocks have to be planned around. If it’s bumping up against turbo limits, 6C12T then continues to clock a bit higher than 8C8T for the same reasons 6C turbo is higher than 8C within one CPU’s clock tables.

        If it’s power-limited somewhere below max turbo, 8C8T is probably more efficient on high-IPC workloads and 6C12T is probably more efficient on low-IPC workloads. In the high-IPC case, more cores clocked lower will always be better, and in the low-IPC case, being able to handle more threads with less silicon active / leaking is probably better by an even bigger margin.

        If it’s thermally-limited somewhere below max turbo, then 8C8T gets a bit of extra advantage over where it stands in the power-limited case, but it’s still possible for 6C12T to win that for the same reasons it can win when power-limited.

    • chuckula
    • 1 year ago

    Look rumor people, claiming that Intel will finally get 10nm chips out next year is stretching things.

    But Intel heatspreaders of [url=https://www.youtube.com/watch?v=bYHg8vJSz-Q<]unusual thermal conductivity?[/url<] I don't think they exist.

      • Redocbew
      • 1 year ago

      Inconceivable!

      • freebird
      • 1 year ago

      Well, this doesn’t seem to be a rumor any more…Ice Lake-sp (1st 10nm Server part) isn’t slated until mid-2020 probably, going by the slide and asterisks.

      [url<]https://www.anandtech.com/show/13119/intels-xeon-scalable-roadmap-leaks-cooper-lakesp-ice-lakesp-due-in-2020[/url<] Intel may be skating on "thin ice" with only a 28-core Cascade-SP server chip to go up against 7nm Rome/Epyc. Not sure what Cascade-AP may bring.

    • Laykun
    • 1 year ago

    For me the 9700K is probably the most interesting product in this line up. Although I do a lot of compilation at home and do benefit from hyper threading, 8 real cores at high clock speed is still better than 4 with 8 threads on my 6700k.

      • Srsly_Bro
      • 1 year ago

      Can you think of a time where 8 cores were not better than 4 cores of essentially the same uArch at the same clock speeds?

        • highlandr
        • 1 year ago

        Except up until recently, you couldn’t get 8 cores at the same speed as 4 from Intel. It was either a speedy quad, or a slower 8 core HEDT (that cost considerably more). Stratospheric clocks on an 8 core chip is new from Intel, and certainly welcome!

      • rudimentary_lathe
      • 1 year ago

      Agreed. It wasn’t long ago that an i7 gave you just 4 physical cores and 8 threads. If I can get 8 physical cores for a similar price, I’m not going to miss hyperthreading, at least for my workloads.

      My next upgrade looks to be either the 9700K or Zen 2.

      • LauRoman
      • 1 year ago

      But people are pissed because they have a 6c/12t in the 8700k. Adding 2 cores and removing 4 threads is a step back in some people’s tgoughts.

      I honestly think it is a move to not overlap with their x or server parts…

      Also, are some people gonna be pissed because they have solder and thus harder to delid?

        • K-L-Waster
        • 1 year ago

        The extra threads from HT aren’t as effective as physical cores — an HT thread is maybe 30% of a physical core in real world terms. In the majority of use cases 8/8 will probably match or beat 6/12.

        If the chip has a soldered IHS there is no benefit from delidding.

          • LauRoman
          • 1 year ago

          I’m not sure i disagreed, but when buying into more cores, you probably are doing it for a reason, so edge cases tend to matter, unless you are doing it for some pimping and braggung rights, but there’s stupid rgb for that.
          If this turns out to be true i’m gonna be pissed because ht could have been enabled on all chips and simply been a requirement for the 400 (or 390 i think it is) series that it be disabled by default in bios if it is a vulnerability mitigation, but then why would the i9 have it?
          But forget that last argument, what really would rub my cheese is that core/thread config would stay the same on mobile as the 8th gen, but we shall see.
          As for the solder as tim, i said “some people”.

          • tygrus
          • 1 year ago

          HT was closer to a 25% maximum benefit. In the early days it was 20% but Intel has made a few changes since then. Maybe a 6/12 HT = 7.5 real cores with some workloads. AMD was getting closer to a 30% boost from their version of HT.

          Diminishing returns as we add more cores & threads to software. Microsoft just loves more CPU/RAM/network/SSD, the amount of junk running in the background of Windows 10 is ridiculous.

          Soldered Intel IHS is less likely but I would never say never.

            • Waco
            • 1 year ago

            HT/SMT can give you up to 100% increased throughput. Putting a number on it makes no sense, since it absolutely depends on the workload you present to the chip.

            For example, the workloads that I typically build systems for benefit almost 95% from SMT/HT because they’re very bubbly. If you’re running something that can already max out the execution units and can keep them busy, of course SMT/HT won’t gain you anything at all. The number of workloads that can do that that *aren’t* power viruses are very low.

    • Krogoth
    • 1 year ago

    Doesn’t make sense to pull this off. I’ll take it a with a massive grain of salt.

      • DeadOfKnight
      • 1 year ago

      Why doesn’t it make sense? It makes a lot of sense now that they have some real competition. I mean, it’s no doubt surprising if they do this, but that’s mostly because it’s been so long since we saw them noticeably try to push the limits of what they can do. The only thing there that looks to be a bit crazy is the TDPs, but maybe there is no GPU for these parts?

      • chuckula
      • 1 year ago

      [quote<]I'll take it a with a massive grain of salt.[/quote<] And a dollop of silicone paste!

      • Wirko
      • 1 year ago

      Barrrh, I poured salt in my coffee by mistake, it’s now a salt lake!

        • Krogoth
        • 1 year ago

        Utah is not impressed

        [url<]https://www.youtube.com/watch?v=pIJvh8GiyKc[/url<]

    • jihadjoe
    • 1 year ago

    i7 loses hyperthreading lol no thx

      • xeridea
      • 1 year ago

      I don’t understand what big deal with this is. Many workloads can’t use more than 8 threads anyway. Games can use maybe 6, video transcoding gets diminishing returns after about 8 or 10. I know there would still be many tasks that could, such as gaming while streaming, or many work tasks, and for those, go ahead and get the 9900K if extra money is worth it for you.

      Also, gains on Intel CPUs from HT are pretty small. SMT on Ryzen gives much better improvements.

      I would be more concerned if they fixed Meltdown yet.

      And….. what would Intel product lineup be without segmentation?

        • dragontamer5788
        • 1 year ago

        Hyperthreading is a feature we [b<]know[/b<] is cheap to implement with easy gains in certain tasks. But Intel wants to charge extra for it. Its not like Hyperthreading costs much silicon. In fact, its probably a fuse that Intel is blowing up. Artificially restricting i7 chips is not really cool, especially because we know that the i7 is going to be $400, or maybe even more.

          • aspect
          • 1 year ago

          I thought it isn’t even hardware and that traditionally the difference between an i5 and i7 is the i7 having microcode that enables hyperthreading.

            • Bauxite
            • 1 year ago

            Its an e-fuse or whatever they call it now, otherwise countless linux users would patch it in for FREE. They claim turning off HT is due to binning but dollars-to-donuts its mostly segmentation and not actual failures. Testing stability of “bonus features” in linux would be easy anyways.

            Sometimes they mess up burning those traces at the factory, early batches of E3 1225v3 Xeons have 4C8T instead of 4C4T as some happy Dell T20 owners know. Various internet posters of typically dubious reputation will occasionally claim to have gotten a lucky “magic” retail extra unlocked chip, but so far it always turns out to be ES. Several lucky AMD owners have gotten some free Ryzen upgrades as well, those at least are hard confirmed as real.

            However there [i<]are[/i<] some things disabled by microcode in current versions of windows that work totally fine in linux. Running unlocked pentium G3528 on non-Z series chipset motherboards is one of those features. Intel actually (ab)used Microsoft's updates to cheat their customers out of this well after it was released.

            • Bauxite
            • 1 year ago

            tl;dr:

            Intel may not QC their sloppy IHS assembly beyond making sure it runs at stock, but they make [i<]damned[/i<] sure you don't get one penny's worth more of features than you paid for.

          • Bauxite
          • 1 year ago

          Darth Intel: Pray we do not segment the products further.

            • Bauxite
            • 1 year ago

            AMD-won-kenobi: Use the cores, Luke! Trust in your enabled features!

            • JustAnEngineer
            • 1 year ago

            [url<]https://www.youtube.com/watch?v=WpE_xMRiCLE[/url<] "This deal is very fair, and I'm happy to be a part of it."

      • MastaVR6
      • 1 year ago

      (Disclaimer, not a chip developer) I’m going to go with the meltdown and spectre vulns being something too difficult to resolve in hardware or a lack of confidence in continued use of hyper threading being secure. Intel may just have decided to forgo using it in enterprise environments as a business choice to avoid something even worse or a marketing failure from its continued use. Time will tell if I’m even partly correct.

        • Puiucs
        • 1 year ago

        HT doesn’t affect the vulnerabilities you mentioned and besides, Intel said that the newer chips will have updated microcodes/architecture changes to mitigate them.

        When people buy an i7 they expect HT to be enabled. It’s the reason why they pay so much for a high end CPU, you want all of the bells and whistles. There are many applications that can take advantage of the extra threads.

        If it’s true then It’s just disingenuous from Intel’s part, but i expect this leak to be fake. Something just doesn’t add up.

        • Spunjji
        • 1 year ago

        Then why is HT still enabled on the top-end i9? I’m seeing this theory bandied about a lot but nobody’s explained that rather large hole in it.

        The simplest explanation is market segmentation because that’s always been Intel’s reason for disabling HT on their mid-range chips.

          • techguy
          • 1 year ago

          The reason people are floating this argument is because they fail to account for the very fact you cite.

      • Chrispy_
      • 1 year ago

      and cache.

      At least we get soldered IHS again; Intel have finally realised that with decent competition they can’t just keep forcing dumb decisions and laughing at the poor consumers with no other choice.

      THANKS SCOTT!

      • Goty
      • 1 year ago

      Just mentally shuffle all of the product designations down by 2. Problem solved!

    • techguy
    • 1 year ago

    I personally guarantee that 9900k will not use a soldered IHS.

    If Intel sold me a $1000 7900x last year that used TIM, they better not sell a mainstream part for half the price (or less) that uses solder the next year.

      • LASR
      • 1 year ago

      That’s not a logical reason. You want Intel to intentionally underdeliver on a new mass-market product, just so you, in your narrow customer segment of on an older, high-end product, don’t feel chumped?

      What Intel is worried about is Ryzen/ThreadRipper2. They’ve lost the battle on core count. Clock speed is where their advantage is now. So it makes sense they do everything to pump that up sky-high to make up for it.

        • freebird
        • 1 year ago

        You mean Coffee(caffeine)-high… Sky-high was 6th edition. Apparently, they had to jettison Hyper-threading to lighten the load to get Coffee that (hot)high on some models. They may have to give away free water coolers with their products if Ryzen 7nm turns out as expected next year and Intel can’t get their 10nm fixed.

          • faramir
          • 1 year ago

          This is genuinely funny and totally undeserving of negative votes.

            • freebird
            • 1 year ago

            So cay, just trying to have a little fun… I’m used to the hater on here by now…

            I was only serious about the water-cooler comment, but hey maybe they can sell the water-cooling system as a “coffee” maker, but that would mean they have to stick with the coffee iteration. Nobody wants a Cannon sitting on there CPU/PC.

      • Phr3dly
      • 1 year ago

      That’s like saying Samsung better not release a better performing phone this year than last year, and ISPs better not increase bandwidth this year over last year.

      Your 7900x isn’t any slower or worse than it was when you bought it just because faster and better products come out.

      • chuckula
      • 1 year ago

      [quote<]If Intel sold me a $1000 7900x last year that used TIM[/quote<] If it didn't use TIM then you'd never be able to cool it you know.

      • albundy
      • 1 year ago

      thanks for the guarantee! your refund checks wont bounce, right?

      • Srsly_Bro
      • 1 year ago

      [quote<]they better not[/quote<] or what?

        • chuckula
        • 1 year ago

        Or he’ll write them a letter telling them just how angry he is!

          • Srsly_Bro
          • 1 year ago

          Yeah, so I guess they better.

      • Kretschmer
      • 1 year ago

      Intel sold me a four-core i7-7700K a few months before they released a 6-core i7-8700k. It took teams of surgeons 48 hours, but I survived.

        • Spunjji
        • 1 year ago

        “We need a gallon tub of preparation H in here, STAT!” 😀

        • chuckula
        • 1 year ago

        The only thing that gets people more angry than Intel not doubling the core count of its chips is when Intel doubles the core counts of its chips.

          • Kretschmer
          • 1 year ago

          I mean, the 7700K is a great chip, and I’m somewhat glad that I built when I did for personal reasons. New tech doesn’t make your old tech stop functioning.

        • Wirko
        • 1 year ago

        Now you have 4 legs and 2 arms, or is it the other way around?

      • Spunjji
      • 1 year ago

      I can’t help but read this as you being angry with yourself for buying an overpriced knee-jerk product. Intel are not your friends.

      • K-L-Waster
      • 1 year ago

      Ah yes, the “I bought top of the line so it has to *stay* top of the line damnit!” argument.

        • techguy
        • 1 year ago

        Not even remotely close to the argument I am making.

        Intel will not solder the 9900k – for business reasons.

        Soldering costs money. Why do you think they stopped doing it on all the mainstream parts starting with Ivy Bridge? If you buy the “durr, the die is too small for soldering” argument that is naive.

        The “i9 9900k”, despite having an “i9” in the moniker, is still a MAINSTREAM part. The argument I was actually making by mentioning my 7900x as a point of comparison is that if Intel can’t be bothered to solder a chip they can sell for $1000, why in the world would they cut into their margins to do it on a $350-500 part?

        But go ahead and throw out your downvotes folks, we’ll see who’s right in the end. If I’m wrong, I’ll come out and say it. You know why? Because if I’m wrong, Intel will have finally offered me what I’ve been asking for since Ivy Bridge – a mainstream chip that doesn’t need to be delidded in order to run at high clock speeds. I’ll be happy to be wrong, in this case.

          • Spunjji
          • 1 year ago

          That’s a much better statement of the argument. Seems reasonable to me, but I still wouldn’t be sure enough to put a guarantee on it. 🙂

            • techguy
            • 1 year ago

            I often state my argument too simply, at first. I figure everyone else thinks the same way I do but I am constantly reminded that is not the case.

            • Spunjji
            • 1 year ago

            Amen to that! If it helps, in this case the phrase “they’d better not” was what led my thoughts away from what you were going for.

            • JustAnEngineer
            • 1 year ago

            I often state my argument too briefly, at first. I figure everyone else is a rational being and knows their job, but I am constantly reminded that is not the case.

          • K-L-Waster
          • 1 year ago

          Y’see, *that* argument actually makes sense.

          Your original one where the *only* supporting statements concerned your recent purchase of a 7900X, OTOH….

            • techguy
            • 1 year ago

            Fair enough. I’ll try to explain my thinking better in the future.

            • MastaVR6
            • 1 year ago

            The ass-u-me colloquialism applies even here.

    • USAFTW
    • 1 year ago

    That’s understandable, can’t run an 8 core 16 thread chip with what Intel considers to be thermal paste.
    If I’m not mistaken the rest will still have peanut butter ounder the IHS. Sigh…

      • auxy
      • 1 year ago

      Intel’s paste is good! Stop repeating this stupid meme! The problem is the assembly process, not the paste!

      The adhesive expands as it hardens and pushes the IHS up off the CPU die. It is a known issue and Intel does not consider it important enough to fix. (‘ω’) Or allows it to happen on purpose depending on who you ask.

        • psuedonymous
        • 1 year ago

        [url=https://images.anandtech.com/doci/8227/1b%20CPUIHS.png<]Chart to back this up[/url<]. We knew this back with [s<]Haswell[/s<] Ivy Bridge, but the meme refuses to die. As for "but why the gap?": mechanical stress. When pressure is placed on the IHS, that force needs to be transferred somewhere. Ideally, you want to spread it evenly across the perimeter so it can be transferred to the socket assembly, and no load placed on the die. But there's a problem: the IHS needs to be thin to perform well, and thin copper flexes. If there were no gap between the IHS underside and the top of the die, pressure on the IHS would be transferred directly to the die. Any off-centre pressure (e.g. mounting a heatsink) risks putting pressure onto a corner, and cracking the die. The gap allows for the IHS to flex without contacting the die. What about solder? [url=http://overclocking.guide/the-truth-about-cpu-soldering/<]READ THIS[/url<] about what needs to go into soldering an IHS to a die and the limitations on die size. But for mechanical performance specifically: because the solder bonds the die to the IHS mechanically as well as thermally, load is spread across the die far more effectively (TIM provides effectively no mechanical coupling so force is concentrated at the contact point).

        • Krogoth
        • 1 year ago

        The problem was never the paste. The issue was QC was a lot less consistent because the solder made fitting more uniform. Ivy Bridge and first batch of Haswell had inconsistent fittings. You either got a chip that was fine or it had a tiny gap(s) that created “hotspots”. That is why simply delidding and reattaching IHS yield significant improvements on these chips. Intel addressed this QC issue with second batch of Haswell a.k.a Devil’s Canyon and onward chips.

          • Bauxite
          • 1 year ago

          Intel addressed jack squat, even their latest fancy 8086k has terrible assembly. Every delid done right has serious temperature drops when paired with respectable coolers. Every single one.

          Their consumer QC is basically “does it boot and not melt enough to crash under full load at base clocks with OEM junk cooler” at this point. They don’t give two spits about the actual thermal performance.

            • Krogoth
            • 1 year ago

            QC has been mostly fixed thought. They are still a few duds here and there. It is a far cry from Ivy Bridge and early Haswell units.

            8086K isn’t really “fancy’ either. It is just a “8700K” with a different model number and higher turbo speeds (a.k.a factory overclocked). You can get the same speeds with nearly every 8700K out there if you are adventurous enough.

            • Spunjji
            • 1 year ago

            I think Bauxite’s main point there still stands – if the heat problem was purely QC related then there would be far more variability in the improvements people see after a delid. In reality the process seems to offer consistent and significant benefits (when done correctly).

            • Krogoth
            • 1 year ago

            Proper IHS fitting accounts for majority of those returns. It is surprising how tiny “mm” airgaps can affect thermal dissipation of the chip.

          • techguy
          • 1 year ago

          It’s both. From someone who’s tested multiple TIMs on many delidded processors, I can assure you that both the efficacy of the TIM and the gap between the die and IHS contribute to the problem of thermal runaway on modern Intel chips.

            • Krogoth
            • 1 year ago

            It is entirely a QC issue on manufacturing.

            Intel’s manufacturing side was so used to using solder (Been using since Pentium 4 until Ivy Bridge). They never pay any special attention to “minor imperfections” in mounting with mass production lines.

            Ivy Bridge and early Haswell brought the issues to forefront when it started effecting product running at stock speed/voltages. They managed to clean it up for the most part with Devil’s Canyon and onward.

            FYI, I happened to one of the lucky ones who has an Ivy Bridge (i5-3570K) that wasn’t impacted it (temp offset for cores at load is around ~2-4C [which is normal] not 5-15C difference that affected some Ivy Bridge users)

        • Goty
        • 1 year ago

        I don’t think it really matters whether it’s the paste itself or the implementation that’s at issue when the solution to both is to just solder the IHS.

          • Krogoth
          • 1 year ago

          The main reason why Intel moved away from solder on their packaging is because of long-term material fatigue/thermal cycling in what they use today where the solder could theoretically break the silicon.

          Bean counters were completely sold on the idea since it lowered production costs.

          The whole IHS/paste issue was mainly a problem with products made during the transition period. It is mostly a non-issue now, unless you want to shave off a few more degrees. Besides if you are going in that direction, the chances are that you were going to de-lid the chip anyway whatever it used paste or solder.

      • freebird
      • 1 year ago

      Peanut butter??? I’d rather have the i-9xxx SMORES Edition. Ryzen wouldn’t have anything to compete with that.

      I guess the Intel fanboys on here don’t like smores…

Pin It on Pinterest

Share This