intel i7-9700k/i9-9900k reviews - disappointed

Discussion in 'Intel x86 CPUs and chipsets' started by Jazper, Apr 16, 2019.

  1. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    Judging by your past false arguments that are wholly biased towards AMD, you gave up when you read the intro page. The Ryzen gets a slightly beefier cooler, because the Ryzen needs a slightly beefier cooler to reach anywhere near the clock speeds of the Intel.

    You know how I determined this? I researched, reading several recommendations on various forums and reddit. This was the cheapest adequate air cooler people recommended for the Ryzen. How many people here are even air cooling their Ryzen 2700X using a cheap or stock cooler?
     
    Last edited: Apr 18, 2019
  2. flu!d

    flu!d KDE Neon 5.16

    Joined:
    Jun 27, 2001
    Messages:
    13,404
    However, that's just maximum 'reported' boost speeds, it's unlikely the CPU would be running all eight cores at full boost all at once. The most logical scenario is the CPU will be jumping a certain number of cores cores in relation to full boost as the thermals related to all eight cores running at 5Ghz all at once is fairly extreme.

    Turbo Boost 2.0 drops boost levels based around the number of active threads.
     
    Last edited: Apr 18, 2019
  3. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    I'll weigh in again

    I have an Asus z370 prime p II board and an i7-9700k. It's pretty much as basic as a z370 board gets, I'm not looking to get enormous overclocks out of the chip. I got the Z370 because I have 3000mhz memory. I don't actually intend to do much in the way of overclocking at all, but avx matters to me as I do a fair bit of compiling, some of which is single threaded.

    The board pops on multicore enhancement by default, which allows all cores to hit 4.9, and voltages to go north of 1.45V. This triggers the 210W event under avx loads, but doesn't thermal throttle as I have a NH-D15.

    I've since turned Multicore Enhancement off.

    Having turned that off, the chip acts normally. Except that the AVX reduction multiplier on this board (by default) is negative 6 - meaning AVX loads trigger a 40x multipler. Having reduced the multiplier to 0, I now get the correct behavior of the 43x multiplier.

    If I set the core multipliers manually, then everything works as intended (-3 AVX offset works). But auto voltage means voltages of up to 1.4V by default. If I change the top voltage in "adaptive" voltage to 1.35, it seems the board just ignores it.

    The board isn't bad. If this behavior is common though, there are traps for new players..

    P.s. I have the cooler as I've had it since the 3770 days, it works, it's stable, and I'm happy with it. My life has changed a fair bit since then..
     
    Last edited: Apr 18, 2019
  4. demiurge3141

    demiurge3141 Member

    Joined:
    Aug 19, 2005
    Messages:
    1,525
    Location:
    Melbourne 3073
    Nope, Ryzen 2 will never reach anywhere near intel clock speeds as they are different architectures, nor does it need to. And manual overclock of a 2700X does not produce any meaningful increase in performance as XFR is already pushing that chip to the limit. It is completely fine on stock cooler and will boost over 4Ghz under 100% load. You are just adding a random $100 to the cost of Ryzen in order to make the false point that the Ryzen setup is similarly priced to the 9700k.
     
    Last edited: Apr 18, 2019
    mAJORD likes this.
  5. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,791
    Location:
    Griffin , Brisbane
    That's not what ths Screenshot shows - All the max column shows is that each core, at some point in time reached 100% , and 4.9Ghz but not simultaneously. you can make this screenshot with minimal total CPU load, because workloads will bounce around across cores and eventually they'll all peak at 100%


    Just because it's a synthetic benchmark doesn't mean it's not representatice of a real workload.. Cmon I think you're more than aware of this.. do the same with x265, or 7zip, and so on. Given that prime version uses no AVX, it's not even representative of load that could be encounted in modern software

    [/QUOTE]

    No, scenario 1 is an arbitrary mix of mostly idle, and an unknown amount of CPU load - the averages tell this story.
     
  6. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    Fair points, I can't prove conclusively that all 8 cores genuinely boosted at precisely the same time to 4.9GHz outside of what the software monitors say. I have an iStat CPU Gadget on my desktop and I can also run CoreTemp, and both of these show in real-time the CPU load on each of the 8 cores and their clockspeeds, but without using a stress tester, I can't show them hitting 4.9GHz across all cores during normal workloads.

    But then that follows the logic that during normal workloads, the CPU may not use 4.9GHz across all cores and maintains a lower 95-115W TDP, but during artificial workloads it does boost to 4.9GHz across all cores, however briefly, and thus uses the much higher 139W TDP. Again, let's remember what Intel advertises: that boost only reaches 4.9GHz on one core at full usage. Let's also remember that whatever the CPU is doing, the benchmarks show it beating the competition in most scenarios.

    Unfortunately, I can't provide any more proof than I already have in a lengthy article and plenty of screenshots. I think sensible people can see for themselves however that any claims of these Intel CPUs being excessively power hungry or not working as advertised are either exaggerated or outright false.

    Sorry no, I'm actually the one surprised that you seem to be unaware of the difference between real-world usage and synthetic benchmarks/stress testers :)

    Prime95 and most synthetic bench/stress testers are not, and have never been, representative of real-world workloads. In real-world software usage, rarely does every single core get 100% continuous usage for an extended period. The benchmarks most relevant to real-world usage contain a mixture of tests, don't necessarily use 100% resources, and are typically based on game engines as these use the most realistic combination of system resources.
     
  7. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    You can show all cores boosting to 4.9, in hwinfo while doing something like compressing a video in handbrake, or using 7zip - these are real world uses..
     
    Last edited: Apr 18, 2019
  8. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    You can't conclusively prove that the software is not mis-reporting this, as fluid says. The software may not be polling the cores quickly enough to catch them briefly cycling down to 4.4 or 4.6GHz for example. Don't worry about it, I think you can see there's a bit of biased nit-picking going on here by some members ;)

    As for your concerns, sorry I didn't answer your previous post. Obviously, take a look at my article for full details, but briefly: yes, Multicore Enhancement is very aggressive, and it's the first thing I was warned about when I got this chip. It's not based on Intel recommendations. Indeed the latest ASUS BIOS updates have changed the wording for setting to make it clearer, but still not clear enough as to how undesirable it may be:

    190418142813.jpg

    This is not Intel's fault, it's ASUS being silly by including a potentially destabilizing setting without making it clear. I have it disabled of course, and have had for a while now.
     
  9. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    https://imgur.com/a/srLtNSt

    Don't know what you're on about.. BTW on my rig, at 4.3ghz/handbrake it is hitting about 110W peak. Which is beyond the 95W TDP peak which intel quote. That being the case, there is no way you're running all cores at 4.9 putting out only 95W.

    You also clearly have your AVX offset set to the default of -3 (based on your bios shot) which means 4.6ghz under load.
     
    Last edited: Apr 18, 2019
  10. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

    Here Anandtech clearly shows intel exceeding 95w TDP on the 9900k and the 9700k.. in a real world example, at default settings

    https://www.guru3d.com/articles-pages/intel-core-i7-9700k-processor-review,5.html

    this aligns with guru3d's analysis.

    Naturally Turbo will kick in so long as temperatures are not exceeded, which explains this.

    If you use a "95w" cooler, then presumably the heat will kick the processor down to 3.6ghz which probably has the effect of keeping the the processor at 95w.
     
    Last edited: Apr 18, 2019
  11. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    OK, let's see if we can make sense of this train wreck of a thread.


    Welcome to my world. I've had the same feeling since this thread started.

    You initially claimed this:

    So let's see. If you disable Intel's built-in limits, you put your system at risk. The sky is blue, more in the full News at 6pm.

    You enabled ASUS' proprietary Multicore Enhancements, which do not follow Intel recommendations, and disabled the TDP limiter. Why were you expecting to remain at 95W?

    Furthermore, you claim the 9700K reaches 210W+ straight off the bat! WTF?

    Here's what I said:

    I then went on to prove this in screenshots, as well as a lengthy article that explains the context of those shots and my settings. But to summarize for the easily confused:

    - The 9700K will boost as advertised to 4.9GHz within 95W. Intel only advertises boost of 4.9GHz on one core as I noted earlier.

    - The problem is boosting to 4.9GHz across all cores at 95W, and while it can do that briefly, it starts to exceed TDP if you remove TDP limits. In my case, I posted screenshots that show my system reaching 139W maximum power draw while running 10 minutes of Prime95 where all cores are pegged at 100% (both max and average) and 4.9GHz across all cores (both max and average) without TDP limits and using a basic 150W rated air cooler. Similar results to the Anandtech article. Here it is again, this time in full size for people who don't like to click thumbnails:

    Hardcon19_27.jpg

    - And even though Prime95 is not reflective of real-world usage and is far more stressful, this is nowhere near "210W+ straight off the bat".


    So basically, we reach the point where you've, unresearched, found a way to disable all limits on your 9700K, push it to beyond what even a tester like Prime95 can achieve, and you're complaining about what exactly? That Intel should anticipate people defeating its protections and yet keep the CPU magically at 95W TDP? Does any other CPU in existence do this?

    Lastly, regarding this AVX red herring that's suddenly cropped up, AVX is not commonly used in desktop applications, so it is not reflective of real-world loads. From this Wikipedia article:

    Software that uses AVX

    How many of these are you using every day?


    /EDIT:

    And to anticipate the inevitable ":but I use x264 and x265 encoding/decoding all day, every day!", the answer is you want a 9700K, because even with the TDP limiter in place and no HT/SMT it trounces AMD's equivalent (orange Intel 9700K, blue=AMD 2700X):

    untitled.jpg
     
    Last edited: Apr 18, 2019
  12. asho444

    asho444 Member

    Joined:
    Feb 7, 2012
    Messages:
    1,037
    Location:
    Sunny Coast
    Question, is Intel's "protection" active by default it is Asus set as default?
     
  13. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    The TDP limiter is on by default at all times unless you manually override it:

    Hardcon19_18.jpg
     
  14. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    Man.. adversarial..

    ME was enabled by default on my board.

    I don't use handbrake day in day out. Also, x265 and x264 will load cores differently (obviously)

    I do 3d render, and I compile, daily. That's my load, that's avx based. More compiles and renders in a given time means less time wasted waiting for the pc - it needs to be rock stable so overclocking is not a good idea, and AMD is not a good idea either. (I'm actually reconsidering getting an i9-9900k, though I think that's want rather than need)

    The start of this thread was about the fact that a lot of numpties I've been watching on youtube have basically said "disable all the power limits and safety mechanisms" - I'm saying, that's stupid, don't listen to them.

    edit:

    People who need processing power, like video editors, programmers, 3d modelers all use AVX based software. You're saying that this is not common. Ok in my world it is, maybe I live in a different world from you? Lots of people do video editing, even for home videos. Most game engines use AVX..
     
    Last edited: Apr 18, 2019
  15. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    In my circumstance , I need stability, a crash or software patch that stops me from working is lost time, I don’t have time to lose on this stuff. I’ll happily wait a few moments more for compiling if it means there is no risk of crashing.

    Performance matters, stability matters more. If I need to reinstall everything I’m down for 4-6 hours due to the toolchains I use for work and that risk is too great with AMD as it stands. I lost a full day and a half when my x370 motherboard died.

    I think in terms of increase of performance, for a 4600mhz processor, 460mhz is 10%, 10% on 10minutes is 1 minute. My compiles are in the range of 15-20 minutes tops , usually much less - like 30 seconds. 2 minutes isn’t a big deal if it means stability.

    Hyperthreading will give me up to 20% more performance (looking at Anandtech) but cost is a consideration too and the 9900k is obscenely priced.
     
    Last edited: Apr 19, 2019
  16. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,791
    Location:
    Griffin , Brisbane
    If by protection, you mean the TDP limited to the level on the box - seemingly not, which is at least one of the key points of the thread.

    https://www.techspot.com/news/77313-do-need-re-review-core-i9-9900k.html




    What are the AUTO values?


    I'm well aware of the difference.. I'm also well aware whilst they can represent unrealistic load (a power virus) they don't all do this - To repeat again , that old Prime version doesn't support AVX, so regardless of whether you use anything that has AVX instructions or not , it's not a worse case scenario is it.. Further to this, Prime95's consumption can vary a lot depending on how it's configured with FFT size.


    Completely false. I and others have already given you examples.. Plenty of Encoders/transcoders, Filters (PS) , Rendering (obviously) utilise 100% CPU for extended periods dependent on the media they're working on - I can't believe I even have to spell this out to be honest. This is computing 101. As a REAL WORLD example, The video's I transcode from my race meetings can take up to 10mins to do and will use 100% CPU for large chunks of that time (The encoding part) . When I use POV ray to 3d render PCB's, can still take longer than a cinebench run.
     
    darkbastard, demiurge3141 and Jazper like this.
  17. flu!d

    flu!d KDE Neon 5.16

    Joined:
    Jun 27, 2001
    Messages:
    13,404
    Considering Windows and it's shocking NUMA scheduler, one core or eight cores at ludicrous speed or not, you're most likely still not seeing the true performance benefit anyway.
     
  18. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,791
    Location:
    Griffin , Brisbane
    These Benchmarks you've linked aren't with TDP limited in place..

    Go look at the "9900k@95w" Article again - a 9900K @ 95w is slower than this
     
    Jazper likes this.
  19. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    Sigh. Where to begin. The nitpicking is getting ridiculous, but I suppose the anti-Intel crowd must have their pound of flesh. Let me see if I can entertain you further :)


    No, by protection I mean what the English word protection actual refers to: prevent harm, i.e., a TDP which is safe, and which does not "Burn up the VRM" as the OP has try to state. Again, the TDP is actually irrelevant to the average user. Voltages, temps, and performance are the main consideration, and they're fine. Using results from a 9900K also don't apply to this specific discussion about the 9700K the OP and I have.


    It says at the bottom of that screenshot what the default is, but to confirm, when I run AUTO, it limits to 95W.


    Irrelevant. It is still far more stressful than any regular software, and I linked to at least one professional site which says as much, comparing it to Furmark. Many overclockers will not use Prime95 at all, AVX or otherwise, because they say it's not representative of a realistic load.

    I think the best example of the difference between a synthetic test and real-world usage comes from the news yesterday that the new Samsung Fold phone screens are breaking after a few days' usage, when Samsung put them through extensive stress tests that supposedly showed them lasting over 200,000 folds. That is: a stress test is not the same as real-world usage, or vice versa.

    Rubbish, I posted a detailed list of the types of software that use AVX, and it's not that common in real-world usage yet. Playing up this angle is not doing you any good, because I even posted a benchmark result which shows the 9700K soundly beating the top AMD chip when AVX loads are used anyway.

    I can't vouch for other reviews. The benchmarks I posted are from this review by AnandTech, and there is no indication whatsoever that they disabled the TDP limits. The full results are:

    Untitled.jpg


    /EDIT: I can see now that the confusion is arising because the TDP Limit is being seen as a hard limit by some, when it is apparently within Intel specs to exceed it. My screenshots already made that clear. But as I said, the TDP is largely irrelevant to users if they can run the CPU at reasonable voltages and temps - and my system is a prime example of that because I'm running closed-case and using a slightly-better-than-stock air cooler.

    But the OP is citing the 210W power draw of a 9900K at full 8/16 core load. This is not what the 9700K does. So if you want to direct your anger at the 9900K go right ahead, and I even noted in my article that i didn't choose the 9900K precisely because I considered it to be too hot to cool silently (bottom of page 3: "My choice came down to the Ryzen 7 2700X or the Core i7 9700K, because the Core i9 9900K is not particularly good value in my opinion, nor easy to cool silently.")



    /EDIT 2: To address Jazper's specific concerns:

    I think you mean that Multicore Enhancement was set to AUTO by default on your board. But even if it was set to Enabled, I'm fairly sure the TDP limiter still works to maintain a safe level on your chip - unless you manually override the TDP limit by entering a large value - which you did, correct?

    This makes no sense to me. Why would you consider getting an even more unstable and hotter CPU like the 9900K? The 9700K is fine, even if you use AVX loads - the benchmarks above prove it. Forget about TDP, your chip runs extremely well. What's happened is that you've listened to the wrong people and you're using incorrect BIOS settings.

    It took me over a month to research and write my article precisely because I had a lot of issues I had to research and resolve before I dialled in the perfect settings. And even then, I wound up with settings that run counter to the common advice that many supposed experts give: I set all of my C-States to Enabled, I enabled all of the limiters, I set my power profile in Windows to High Performance, etc. etc. All chronicled in my article, one man's bumbling journey towards learning about how his new system works. But I got there in the end.

    Modern CPUs and motherboards are extremely complex, and sadly, the vast majority of people are given options that they neither understand nor even need, and can do more harm than good when they try to overclock.

    Your world of 3D modeling is just that: your world. Most users do not do complex 3D modeling. Cinebench is not representative of the performance of a 9700K in most real-world applications; again, it is strongly promoted by AMD fans because it is one of the few benchmarks where the AMD chips can gain the upper hand with multicore performance.

    If we want to get really technical the issue is AVX 512, not just "AVX". Game engines using DX12 can use AVX, but it's not an issue as the 9700K is tested to perform at least as well if not better than the top AMD chip in games. It's AVX 512 that's the real killer in Prime95, and that's not being used in much software.
     
    Last edited: Apr 19, 2019
  20. OP
    OP
    Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,600
    Location:
    Melbourne Australia
    This patronising is getting ridiculous ... you did make it personal...

    First Multicore enhancement was enabled on my motherboard by default. No I didn't manually override anything, I left everything as it was, out of the box to start with.

    The 9900k is not "more unstable" - this statement also implies the 9700k is unstable. This is plain wrong. The intel platform is a much more mature platform than the AMD platform, this is both a good and a bad thing, bad in a sense it hasn't really evolved much since Haswell, good in a sense that it has matured and a lot of the bugs have been ironed out.

    Detailed list? you missed a few things..Like web browsers, Microsoft Excel (probably powerpoint too), the entire adobe suite, windows, linux, autocad (all variations of it), Visual studio, gcc, nearly all video compression and editing software, just about everything that uses a graphics card, including drivers..

    Did you miss the point where x87 floating point instructions were basically deprecated in favor of SSE/2/4 and AVX 1/2 when things went 64 bit?

    Turn your AVX negative offset up right up and see exactly what happens to your general desktop performance.

    I also resent the fact you suggest I'm using the wrong bios settings, and that I don't understand this stuff. I have 30+ years of experience tweaking these things, and I know my way around electronics better than most (I read data sheets, do you?, My day job involves designing electronic devices, what about yours?).. do you truly think I don't do my homework?


    AVX-512 hardware isn't even available on the desktop platform, including the 9x00 series, so it can't possibly be a problem for these chips, you knew that right? You do know that people use their computers for more than games. Games make up less than 5% of my total computer time these days.


    Stop spouting your uneducated opinion.
     
    tracer0 and mAJORD like this.

Share This Page

Advertisement: