Single-core performance of Intel's Sunny Cove chips Surface - Shows Big IPC Increase

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jun 17, 2019.

  1. user1

    user1 Maha Guru

    Messages:
    1,422
    Likes Received:
    465
    GPU:
    hd 6870
    Clock speed is kinda overrated when ln2 5ghz zen2 chip beats an ln2 cooled 6ghz skylake x

    gotta stop thinking in mhz, the cores are wide and the clocks are "low" with these. and thats ok in my book.
     
  2. sverek

    sverek Ancient Guru

    Messages:
    5,367
    Likes Received:
    2,265
    GPU:
    NOVIDIA -0.5GB
    All I care is gaming frame count and benchmark numbers.

    Clock speed is just a number, for some reason majority losing their breath over it. Makes a great bait though.
     
    user1 likes this.
  3. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    Why? Hardly anything since Zen1 could hold back your 970 back.
    At time you are about to upgrade, you will look at benchmarks for given resolution and GPU will play major role anyway.
    Unless you intend to play on low.
     
  4. user1

    user1 Maha Guru

    Messages:
    1,422
    Likes Received:
    465
    GPU:
    hd 6870
    True
     

  5. sverek

    sverek Ancient Guru

    Messages:
    5,367
    Likes Received:
    2,265
    GPU:
    NOVIDIA -0.5GB
    I don't change CPU every few years, so I want new CPU/mobo/ram to last as long as possible.

    1) We don't know how far will GPU develop in a few years (= we don't know how fast CPU cap will be reached)
    2) I prefer smooth gameplay over quality. So I tend to lower game settings until my framerate is stable.

    As weird as it might sound, BF4 caps my CPU (2500k@4.5) on 96FPS before it cap GTX970 on medium settings on 1440p@96hz panel.
    Unless I go full retard and play on highest settings, CPU is the limiting factor right now.

    So yeah, CPU cap does exists and it can easily be reached by lowering quality a bit. We know that by 720p/1080p benchmarks.

    Say if I want to get 144hz panel and push framerate as close as possible to 144fps, I know that CPU might be limiting factor in some games.
    And no matter how much I am willing to sacrifice video settings, CPU load tends to be static, therefore, once cap is reached, there nothing to be done, beside overclocking.

    I am talking from my experience and benchmarks that I see. I am not eye-candy whore that instantly locks all video settings on ULTRA.
     
    las likes this.
  6. Exodite

    Exodite Ancient Guru

    Messages:
    1,913
    Likes Received:
    171
    GPU:
    Sapphire Vega 56
    That's similar to my experiences with the 2600K.

    In the Division 2 in particular, though I'm sure it applies to some other modern titles as well, the CPU is definitely the limiting factor if I try to optimize the settings. All Ultra, or maxed out, it's not quite so bad but when I try to squeeze out more by optimizing the game settings I'm clearly limited by the CPU. So the trick ended up being bumping the settings juuust enough to get as much eye candy at the frames my CPU allows. :)

    Though that's hardly the norm I think, we have to recognize that Sandy Bridge is pretty old now and while I've yet to see any issues hitting 60 FPS in any modern titles it can be problematic if you're targeting 90+.
     
  7. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    We know one thing for sure. CPU use of games is not going to progress as much vertically as it is going to progress horizontally.
    Hell, All done in last few years was to decrease verticality and increase horizontality.
    - - - -
    Your issue is hardly IPC and clock (verticality). It is lack of cores. (horizontality)
     
  8. sverek

    sverek Ancient Guru

    Messages:
    5,367
    Likes Received:
    2,265
    GPU:
    NOVIDIA -0.5GB
    Oh yeah, you right.

    That's why I am looking up to Zen2, at least 8 (16) cores, threads. Maybe 12/24.
    Single core speed wouldn't mean much, if there no power behind it.

    Looks at short-lived 6600k.
     
    Fox2232 likes this.
  9. Astyanax

    Astyanax Ancient Guru

    Messages:
    3,656
    Likes Received:
    980
    GPU:
    GTX 1080ti
    MSI didn't do crap.
     
    Aura89 likes this.
  10. Aura89

    Aura89 Ancient Guru

    Messages:
    7,673
    Likes Received:
    921
    GPU:
    -
    He's not going to because he has no information, just as he never replied to my last request for him to answer essentially the same question.
     
    Fox2232 likes this.

  11. coth

    coth Master Guru

    Messages:
    407
    Likes Received:
    35
    GPU:
    MSI GTX 970 Gaming 4G
    If you look around stock results in their database, you'll an average of non-K 3770 with stock multiplier of around 370-375, which puts it well in its position. Probably just a mistake in the table. Could have been 363, not 263. This is 2017 version of the test, so they probably not paying much attention to older CPUs.
     
  12. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    But you are still unable to repeat it here.
     
  13. Aura89

    Aura89 Ancient Guru

    Messages:
    7,673
    Likes Received:
    921
    GPU:
    -
    Lol yes, MSI totally did all the tests and got an average of 13% or just did one test and got 13% ipc increase and mysteriously posted it somewhere you won't tell anyone about.

    Even though that is impossible. There is not a single reviewer or company that could test ALL situations. As long as there is one application that shows 15% or higher, there is no lie, but you don't get that because YOU don't want to. All the while not showing your sources either.

    It's like talking to a wall except i'd probably have more luck with the wall.

    Guys, we've found another intel worker. Hello intel
     
    Last edited: Jun 21, 2019
  14. oxidized

    oxidized Member Guru

    Messages:
    198
    Likes Received:
    29
    GPU:
    GTX 1060 6G
    Or maybe it's you being the AMD worker here? Ever thought about that?
     
  15. vbetts

    vbetts Don Vincenzo Staff Member

    Messages:
    14,714
    Likes Received:
    1,296
    GPU:
    RTX 2070FE
    @Aura89 @oxidized

    Let's not argue like kids and go back on topic, or argue like adults please.
     
    SplashDown likes this.

  16. Aura89

    Aura89 Ancient Guru

    Messages:
    7,673
    Likes Received:
    921
    GPU:
    -
    Oh are you prepared to show us the mythical results from MSI that has tested all tests then?

    We're waiting.
     
  17. SplashDown

    SplashDown Master Guru

    Messages:
    608
    Likes Received:
    67
    GPU:
    EVGA 980ti Classy
    LOL .... all of these threads about chips all end the same way, badly. Alot of you guys know much more than I do I feel stupid sometimes I just don't know as much in depth but all this battleing about it ... man waste of energy and stupid. Peace out ...
     
  18. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    1st) MSI used AMD's slide. Those 13% were there written by AMD.
    2nd) AMD later decided to use different test to show therefore used different value
    3rd) IPC varies from app to app
    [​IMG]
    Or is it 18%? Based on which tests? Add any new test suite and average value goes UP or DOWN.
    That dude has some trauma which is returning to him.
     
    Aura89 likes this.
  19. Arbie

    Arbie Member Guru

    Messages:
    169
    Likes Received:
    58
    GPU:
    GTX 1060 6GB
    Maybe because the Q9650 had 12MB of L2 cache. That enabled it to keep up with newer chips for a long time in specific (primarily integer) benchmarks, while falling way behind in general.
     
  20. Carfax

    Carfax Ancient Guru

    Messages:
    2,560
    Likes Received:
    274
    GPU:
    NVidia Titan Xp
    It's quite simple to resolve this debate about the IPC increase. AMD's Robert Hallock has already addressed it on twitter. The 15% figure was taken from SpecINT 2006, and the 13% figure was from CB 1T. Between the two, the SpecINT figure is more reliable as a means of estimating performance increase for average users.

    In heavy AVX/AVX2 applications, the IPC increase will be 30 to 60+% in real world, and up to 100% in synthetic benchmarks.
     

Share This Page