9900K or wait for KS?

Discussion in 'Intel Processors' started by shadow2761, Sep 9, 2019.

  1. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    Was just days away from pulling the trigger in a new 9900K.

    Then saw the news about KS coming out in a few weeks time.

    I can't find any gaming benchmarks of the KS over the K. Should I wait for the KS or just get the 9900K this week?

    I will mainly be interested in high refresh gaming @ 1440p/4K.
     
  2. DogsofJune

    DogsofJune 2[H]4U

    Messages:
    2,691
    Joined:
    Nov 7, 2008
    If you are ready, then get what you need. There is no promises that the KS offers anything special.
     
    trick_m0nkey likes this.
  3. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    From what I read, the KS offers 5GHz on all cores from factory.

    But does the K not offer the same thing, or is that only once you OC it?
     
  4. mda

    mda [H]ard|Gawd

    Messages:
    1,612
    Joined:
    Mar 23, 2011
    KS is a binned K, that does 5GHZ all core boost.

    If you can overclock a K to 5GHZ all core, you have the stock settings of a KS. Not sure what your chances are on the standard 9900K, and does not say anything on how high an average 9900KS can overclock.
     
  5. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    Silicon lottery has a strong history of accurately calling likelihood of hitting x speed with each enthusiast chip.

    They say your average 9900k has a 30% change of successfully hitting 5Ghz on all cores from their testing.

    https://siliconlottery.com/collections/coffeelake-r/products/9900k50g?variant=15392435896406

    And only a 5% chance of 5.1Ghz on all cores

    https://siliconlottery.com/collections/coffeelake-r/products/9900k51g?variant=15392436093014

    However 91% will hit 4.9Ghz on all cores

    https://siliconlottery.com/collections/coffeelake-r/products/9900k49g?variant=15392435699798
     
    blurp likes this.
  6. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    I'm three for three on 5.0GHz all core overclocking with the Core i9 9900K. I've also got a 9600K and 9700K to do the same.
     
    Mav451 and sharknice like this.
  7. sharknice

    sharknice [H]ard|Gawd

    Messages:
    1,729
    Joined:
    Nov 12, 2012
    Mine is 5.1 all cores. Is it really that rare to hit 5?
     
  8. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    Not in my experience.
     
  9. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    Ahh ok thanks guys, I get it now.

    KS is just a 100% chance of 5GHz on all cores from factory and possibly higher with OC.

    K is by chance on lottery.
     
  10. Keljian

    Keljian Gawd

    Messages:
    646
    Joined:
    Nov 7, 2006
    200mhz is 5% performance. Tops.. just go the 9900k.
     
    Snowdog likes this.
  11. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    Will do, cbf waiting till mid october anyways lol
     
  12. ChronoDetector

    ChronoDetector 2[H]4U

    Messages:
    2,577
    Joined:
    Apr 1, 2008
    Just buy the standard 9900k and oveclock it and be happy, no need to wait for this so called special 9900KS CPU which most likely will be more expensive while you could have saved your money buying the standard version and do it yourself.

    Intel as of late has been rather disappointing and showing lack of innovation on their products. Still stuck on 14nm and have no answer to AMD's Ryzen third generation and this is the best they can come up with is rather laughable.
     
  13. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    There's nothing really laughable about 5Ghz 8 core/16 thread performance. Outside of server workstation loads that's a chip that will do everything and do it well.

    And just like the 9900K - AMD can't actually beat the 9900KS with third gen Ryzen IPC -- and since most 'things' rely on clock speedmore than core count - especially when you get to 8 core/16 thread --- Intel retains performance crown.
    However, the price might be laughable.

    Based on Silicon lottery's numbers he can pretty much rest assured of 4.9Ghz across all cores on a 9900K.

    That'll be imperceivably different than 5.0Ghz -- but there is something 'amazingly elusive' about that round 5Ghz number for us long time hardware enthusiasts. For most of us - it's always been just out of reach each generation -- going back to Sandy Bridge. And it's been the - Holy Grail for long over a decade.
     
  14. sharknice

    sharknice [H]ard|Gawd

    Messages:
    1,729
    Joined:
    Nov 12, 2012
    Doesn't the ks also remove the gpu too? So you don't have the option to be able to run a couple extra monitors or use intel's hardware encoder, etc. You have to have a discrete gpu so if you want to use it as a media player or server or something in the future you have to keep a gpu in there.
     
  15. vegeta535

    vegeta535 2[H]4U

    Messages:
    3,153
    Joined:
    Jul 19, 2013
    People that are buying a 9900k are more then likely going to be using a discrete gpu. It is only useful for back up really.
    From what I gather AMD has Intel matched or beaten on IPC with 3rd gen Ryzen. They just can't match the clock speed.
     
  16. sharknice

    sharknice [H]ard|Gawd

    Messages:
    1,729
    Joined:
    Nov 12, 2012
    Yes, it's just a nice to have thing in case you need it.
    I have a 9900k but I use more monitors than a single discrete gpu can handle. And when I build a new PC I use my old one as a server.
     
  17. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    From what I’ve seen Intel is still a hair faster IPC at identical clocks when compared across many tests. Toss on another ~500mhz with Intel and it’s still Intels performance crown to bear

    https://amp.hothardware.com/reviews/amd-ryzen-9-3900x-vs-core-i9-9900k-ipc-shootout?page=3
     
  18. Emission

    Emission [H]ardness Supreme

    Messages:
    4,179
    Joined:
    Dec 6, 2005
    Jay says 9900KS is just Intel's version of an FX-9590, aka not worth it.

     
  19. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,470
    Joined:
    Aug 16, 2010
    AVX too?
     
  20. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    Yes and no. All three of the 9900K's I've tested can do 5.0GHz under AVX, but they are all very close to throttling. Two of them were tested on open air test benches with a fan on the motherboard VRM's. Those are cooled with a 360mm radiator and a higher end water block. In an actual chassis, I have no doubt that they would be unable to handle AVX work loads at 5.0GHz. The third one is the one in my personal machine and it can handle AVX workloads at 5.0GHZ. Again, its borderline as well but that one is in an actual chassis. I'm also cooling it with a 480mm radiator. My ambient temperatures in my office tend to sit at 72F or less. Most people as I understand it do not keep their homes that cold.

    If I'm testing a bunch of hardware in my office, the ambient temperatures go up considerably, and at that point my 9900K probably wouldn't be able to do 5.0GHz with AVX at that point. The 9600K and 9700K can both do AVX at 5.0GHz too. They do it with a little more head room, but not by much. Really, I think if you want AVX at 5.0GHz consistently, you need to delid these things and either use a copper heat spreader, liquid metal TIM, or direct die cooling.
     
    Meeho likes this.
  21. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    Do you recall what voltage you used?
     
  22. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    The one in my personal system is 1.30v. The others were around 1.32v to 1.35v if I recall correctly. One of those chips went to its new owner, as it was one I was lent to start with and the other went back to Kyle. I only had it briefly.
     
    Archaea likes this.
  23. Keljian

    Keljian Gawd

    Messages:
    646
    Joined:
    Nov 7, 2006
    I should mention here, the 9900kf (no GPU version) hits higher clocks on average than the 9900k, so if clock speed is your deciding factor, go for that


    Also, I run my 9900k per my sig. Transient loads get 5ghz, full continuous load gets 4.8ghz.
     
    Last edited: Sep 10, 2019
  24. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    Silicone lottery used 1.3 as their threshold for success at 5.0Ghz and that was the 30% number. (Also with an AVX offset of -2)

    Based on your experimentation it sounds like the cooling capacity is the big hurdle if you start cranking that voltage to successfully hit the 5.0Ghz.
     
  25. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    I'd say so since all the ones I've tested border on throttling with AVX enabled at 5.0GHz.
     
  26. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    So should I just get the KF version since I will always be using a dedicated GPU. Currently have RTX 2080 Ti.
     
  27. Keljian

    Keljian Gawd

    Messages:
    646
    Joined:
    Nov 7, 2006
    Unless you have a use case for quick sync, yes
     
    Dan_D likes this.
  28. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    I find contradicitng info online about the KF vs K.

    Some people rekon the KF are just failed K chips and hence have the iGPU disabled and also don't clock as high.

    Some rekon the KF will clock higher mote often than not.
     
  29. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    We've seen models with the iGPU disabled clock higher in the past. The Core i7 7740X was a stupid CPU, but the net effect of having the iGPU disabled was that it did indeed clock higher. So there is some potential for that.
     
  30. N4CR

    N4CR 2[H]4U

    Messages:
    3,834
    Joined:
    Oct 17, 2011
    Uhh, AMD has superior IPC hands down, that's why they're within 5% perf at far lower clock and there is no difference at the OPs 4k and pratically none at 1440p resolution.
     
  31. freeagentt

    freeagentt [H]Lite

    Messages:
    67
    Joined:
    Dec 5, 2018
    Indeed. https://valid.x86.fr/show_oc.php?id=598279

    Even with our cold winters I still couldn't do it. Neither will my x5690 e.s. nor my 3770k.. at least on air. As usual, I will probably try again this winter :D
     
  32. Nightfire

    Nightfire [H]ard|Gawd

    Messages:
    1,698
    Joined:
    Sep 7, 2017
    Silicon Lottery and Intel standards are going to be much higher. Even the 4.8 ghz duds that Silicon Lottery have might be able to hit 5.0 ghz with enough voltage and cooling.

    On average, it is looking be be the difference between an 8086k and an 8700k. For those with limited cooling, it may be worth it.
     
  33. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    There are no benchmarks comparing the Intel Core i9 9900KS to the Core i9 9900K. They are both CPU's with the same 8c/16t count and 5.0GHz boost clock. The only difference is that the KS model will boost on all cores to 5.0GHz in multi-threaded applications. In games, the two should be identical. More importantly, Intel isn't sampling the Core i9 9900KS at present. I tried to get one and it isn't out yet and isn't being sampled to reviewers.

    It may be better binned than the 9900K, but as you said, there is no guarantee of that.

    You are mostly correct. The difference is that the 9900K can boost to 5.0GHz on any specific core, but doesn't boost to 5.0GHz on all cores at the same time. You have to overclock it to achieve this. Doing so is relatively easy in my experience. Doing so with AVX is something I wouldn't count on. I've tried the 5.0GHz on four ASUS motherboards and it worked with all the Intel CPU's I currently have or have access to. The only thing is, they won't all do it with AVX. The profiles default to an AVX offset of -2.

    Correct. As for the chances, Silicon Lottery has their numbers, but my odds since that site came into existence have personally been better. I'll get into the "why" of that shortly.

    You are mostly correct. However, AMD generally offers superior IPC over Intel right now. The Intel CPU's are only competitive assuming equal core counts because they can clock considerably higher.

    Unknown. The Core i9 9900KF has a disabled GPU. The rumor is they are CPU's where the iGPU failed to pass QC and was disabled. Aside from using QuickSync, I don't see any reason to keep the GPU. Many of today's GPU's can run at least three monitors, so I don't see the point of having the iGPU outside the aforementioned encoding scenario.

    I don't know about being used for backup, although I suppose you could. I'd still slap in an old GPU if I needed to. I've got some old GPU's but they all beat the iGPU hands down. As for AMD's IPC, your on the money. There may be a few cases where Intel is still faster in terms of IPC, this was something I saw in games but for the most part I think its safe to say that Zen 2 has higher IPC than the many Skylake refreshes.

    No, it isn't. It isn't even in the data in your link. Intel loses many of those benchmarks as it did when I tested it.

    I wouldn't say: "hands down." I'd say: "most of the time" for sure. In gaming, people often repeat a couple of myths. I'm going to straighten this out right here. The aggregate number thrown around in the 36 game test is misleading. Aggregate numbers are a generalization which is useful at times, but it isn't the end all be all of the story. There are games with as much as a 15-20% performance difference between Intel and AMD. There are cases where AMD's minimum frame rates are half what they are for Intel. My testing in Destiny 2 was an example with this. What was not published as of yet was Destiny 2 data concerning the 3700X, which showed a minimum of 18FPS at 4K vs. 54FPS for the Core i7 9900K at stock settings. Their averages are the same and AMD manages higher maximum FPS, but the minimums are what you really should be concerned with. With gaming, you need to be careful not to get caught up in the "Intel is only 5% faster so it doesn't matter" myth.

    Secondly, people also assume that because you are more GPU bound at 4K than CPU bound, AMD is just as good at gaming. It isn't. At 4K, those minimum frame rates are much lower for both CPU's. Intel having a decisive lead in some games can be the difference between a smooth experience or not at that resolution. I'm not saying AMD can't do 4K gaming, it can, but when selecting a CPU for gaming, you need to really look at the data. Almost all of the CPU reviews out there avoid showing you anything but average frame rates. I didn't show those in anything other than Destiny 2, which wasn't part of the initial review due to the issue Ryzen 3000 series CPU's had of not being able to run it at all.

    I think it was Hardware Unboxed that actually did a video showing minimums, maximums, averages and 99th percentile etc. In most of the games, even a Core i7 9700K is faster than AMD's CPU's. Sometimes faster than the Core i9 9900K. Essentially, Intel is pretty much always faster at gaming and while the aggregate number may be 6% across 36 games or whatever it was, there are plenty of cases where that doesn't tell you the whole truth.

    I don't think its actually as unlikely as Silicon Lottery claims to hit 5.0GHz on a Core i9 9900K. Silicon Lottery guarantees a certain overclock on their pretested processors and I think their standards are tighter than they need to be to ensure that anyone can achieve the result they are paying for on virtually any motherboard and system configuration. That means that the CPU's they guarantee have to be able to do it at less voltage, to account for use on an AIO, air cooler, or less than ideal ambient temperatures. They also need to be able to do it on motherboards with less than ideal VRM's.

    I've never used an AIO on a 9900K and I'm 3 for 3. I've also never used a motherboard with really shitty VRM's or tried these in a room with horrendous ambient temperatures. So while I don't think 5.0GHz is all that hard to achieve, I can see why Silicon Lottery only says its about 30%. Two of the ones I hit 5.0GHz on required upwards of 1.35v. The other only needed 1.30v. While I can run them all at 5.0GHz in my office, I would never guarantee that two of those three CPU's could be run like that anywhere else. Especially not with AVX.
     
    sharknice, Nightfire and Dayaks like this.
  34. Archaea

    Archaea [H]ardForum Junkie

    Messages:
    9,360
    Joined:
    Oct 19, 2004
    Well - I'll not counter your results - but the link I provided as reference did in fact say Intel 9900K was a hair faster than AMD when both were at 4.0Ghz and IPC was tested on a level playing field:

    From their conclusion:

    Performance Summary: These Intel and AMD multi-core processors traded blows across all of our CPU tests, but Intel won more often than not. The blue team notched IPC wins in SANDRA's Dhrystone integer tests, Geekbench, POV-Ray, LAME MT, 3DMark, and Middle Earth: Shadow of War. AMD stole single-threaded victories in SANDRA's Whetstone FPU tests, Cinebench, and Y-Cruncher. While not an outright thread-for-thread win for AMD, the company has obviously worked hard to improve 3rd Gen Ryzen performance in lightly-threaded loads, while its multi-core scaling is downright impressive.
     
  35. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    54,492
    Joined:
    Feb 9, 2002
    I linked my own tests for comparison. Keep in mind I didn't run all the same tests that site did. I also conducted my tests at 4.3GHz, not 4.0. It's also worth pointing out as I did, that a 3900X with four cores disabled isn't quite the same thing as an 8c/16t Ryzen 3000 CPU. It was simply the best I could do at the time.
     
  36. sharknice

    sharknice [H]ard|Gawd

    Messages:
    1,729
    Joined:
    Nov 12, 2012
    Well rumor is the 9900k will become cheaper when the KS comes out. So maybe wait even if you decide to get the K.
     
  37. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    snap, I pulled the trigger on K last night.

    Got it here ready to install over the weekend.
     
  38. e-Killer

    e-Killer n00b

    Messages:
    40
    Joined:
    Dec 3, 2005
    Look for a KF on Newegg, usually they put on sale at 419.99 + FreeShipping.
     
  39. Dayaks

    Dayaks [H]ardness Supreme

    Messages:
    7,119
    Joined:
    Feb 22, 2012
    About two weeks ago I got a 9900KF. I didn’t see anything worth waiting for. I set it to 1.35V and 5Ghz all cores. I haven’t had time to screw with it further but seems completely stable.

    For Timespy CPU tests I am at 70C with a custom loop set to 35C liquid temperature so I am happy with it. I’ve only gamed on it / encoded with Adobe Premiere Elements / Timespy.

    Very happy with it. VR is much better on it than my 2700x.
     
  40. shadow2761

    shadow2761 [H]Lite

    Messages:
    106
    Joined:
    Mar 11, 2016
    Nice. I will be installing it over the weekend on the Asus Strix Z390-E + G Skill 3600MHz C14 RAM.

    I hope to hit 5GHz with temps around 80°C max with a H115i Platinum. I don't like the idea of being around 90°C.