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. BReal85

    BReal85 Master Guru

    Messages:
    487
    Likes Received:
    180
    GPU:
    Sapph RX 570 4G ITX
    Intel had years to change nodes, yet they cannot do it desktop-wise. Of course they will leap forward. As AMD will leap next time.

    "Especially outside of low to mid-end market."

    Check how many % of PC owners buy high-end GPUs (1080ti-2080-2080Ti) and how many buy cards below those. Fortunately AMD had competition to offer up to the mid-high segment after R300 (RX 470, 480, 570, 580, Vega 56) that had better price-performance ratio. Just people thought AMD cards are still for egg frying (after the R9 290-290X reference models' 90-95). However this was an absolute urban legend as the mentioned reference cards were about 10 degrees cooler than the R9 cards. 80-85 degrees for a reference card is absolutely good. And of course the real problem was there were low stock numbers at the beginning. Other problem was the mining craze that hit AMD card prices much more. And the fourth thing is AMD had worse drivers earlier (many years ago) which remained another urban legend up until the last years. However since the end of the Catalyst era many things changed: they started releasing Catalyst drivers at the end more frequently, and the drivers since Crimson were much better in quality, plus NV driver quality got worse since Crimson, so the drivers are equal in quality and quantity today.
     
    Last edited: Jun 17, 2019
  2. fredgml7

    fredgml7 Master Guru

    Messages:
    242
    Likes Received:
    87
    GPU:
    Sapphire RX 7600
    I think nobody doubts that Intel will fight back hard in the next years, and that's pretty good to all of us as consumers, but if until there AMD gains a considerable market share It would be even better.
     
  3. Hewyx

    Hewyx Guest

    Messages:
    1
    Likes Received:
    0
    GPU:
    1080TI
    Here is my take on the refresh,,, AMD gifted this to Nvidia I bet there is a mountain of each of these chip that didn't quite make the grade for example 2080 TI short a few CC now it is a 2080 Super and so on. My wish is that the GPU makers would put a cache drive port (M.2 PCE 4X3X-4) on the cards.
     
  4. CPC_RedDawn

    CPC_RedDawn Ancient Guru

    Messages:
    10,409
    Likes Received:
    3,077
    GPU:
    PNY RTX4090
    Just trying to build up hype and take away some of the limelight from AMD. It's to try and convince all those current Intel owners thinking of moving over to AMD to wait. Simple as that.

    Or it could be that Intel has finally figured out their issues with the 10nm process and that said process, including new architecture, including processor layout, memory improvements, etc are all better than TSMC's 7nm process.

    Also, will Sunny Cove be using a chiplet design like Ryzen 3000? With a 14nm I/O die and 7nm core dies? Or will it be a traditional Intel monolithic chip?

    If its the later and it is monolithic and it does come in 14core/28thread parts and these numbers are to be believed that is going to be one VERY hot running chip indeed! Sure 10nm will pull less power but Intel will need to pull a miracle out their asses to get a monolithic chip design with this many cores to cool well. Certainly they will have no chose but to go with solder rather than tooth paste this time but even then with so many cores and such a big clock speed. 10nm has to be near perfect, and I doubt we will see any amazing overclocks on a chip like this.

    Still exciting stuff if true though, I have wanted nothing but competition for years and now we are getting it in the CPU market! FINALLY!!!

    Though at this point I couldn't care less, I made the jump back to AMD for the first gen Ryzen chips, mainly because we were told AM4 socket will be getting support up until 2020 at the earliest. So we will be getting ZEN 2 (ryzen 3000) and a ZEN 2+ (ryzen 4000) upgrade after as well until AMD decide to release AM4+ or AM5 for Ryzen 5000. No need to upgrade anything other than CPU, keeping your exisiting motherboard and RAM. I am glad I splashed out on a decent X370 board with good a good VRM.
     

  5. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    then tell me why a q9650 scores the same single thread score as an i7 3770, or i5 4670, it is not representative of realworld at all. https://valid.x86.fr/bench/1 this chart is a mess

    also I dont use geekbench, scores are significantly effected by the operating system.
     
  6. d2v

    d2v Guest

    [​IMG]
    in cpuz chart, 9900k 5.0 ghz: 543 point
    in this photo is 600--> fake
     
  7. BlackZero

    BlackZero Guest

    Comparing different versions.

    Each version comes with its own relatively small data base, and the algorithm is obviously adjusted with each release. Also, as a general indicator of performance in most common instructions, it's pretty accurate from what I can see.

    Clearly, maintaining such a utility isn't an easy task, and all other similar utilities do exactly the same with different revisions.
    [​IMG]
     
    Dragam1337 likes this.
  8. fantaskarsef

    fantaskarsef Ancient Guru

    Messages:
    15,693
    Likes Received:
    9,572
    GPU:
    4090@H2O
    Yes, I too was surprised how those numbers pop up without a single validation, posted by some guy on some forum somewhere in Asia. Credible is something different to me. Although at least there's expressed doubt here, on other sites I did not read it like that.
     
  9. fantaskarsef

    fantaskarsef Ancient Guru

    Messages:
    15,693
    Likes Received:
    9,572
    GPU:
    4090@H2O
    You miss the point: Don't trust any numbers besides people doing real benchmarks. You're just trying to throw dirt when in reality, if you don't trust AMD's numbers, you can't trust Intel's, and with that it leads you to what we say: Proper benchmarks needed.

    I don't trust poorly written "pictures of text" with no credibility at all.
     
    airbud7 likes this.
  10. fantaskarsef

    fantaskarsef Ancient Guru

    Messages:
    15,693
    Likes Received:
    9,572
    GPU:
    4090@H2O
    Probably right, not so much. Which still is not "have never been cought" at all. Stop cheaping out and keep true to what you say yourself, avoid marketing numbers alltogether.
     
    airbud7 likes this.

  11. fantaskarsef

    fantaskarsef Ancient Guru

    Messages:
    15,693
    Likes Received:
    9,572
    GPU:
    4090@H2O
    And under what conditions do you believe in marketing material? Because usually isn't always.
    I personally find those PR slides interesting, but it's not much use until confirmed by somebody not paid, or giften free hardware, by companies. Haven't suffered due to being strict about not giving in to either of AMD's, Intel's, or Nvidia's stuff right away.

    And again, that picture, because it's not even a slide or proper table, lacks any credibility besides coincidentally matching a PR slide. And that's what this thread's about, sadly, a single picture without any way to confirm it's numbers. And I don't see much reason to defend it's credibility.
     
    airbud7 likes this.
  12. D3M1G0D

    D3M1G0D Guest

    Messages:
    2,068
    Likes Received:
    1,341
    GPU:
    2 x GeForce 1080 Ti
    I was referring to those who say that Intel provides more accurate pre-release numbers. The fact that Intel flat-out lied about their 28-core 5 GHz CPU should permanently dispel any myths of Intel being more honest in their numbers.
     
  13. Tsenng

    Tsenng Active Member

    Messages:
    52
    Likes Received:
    17
    GPU:
    Zotac RTX 3080 AMP
    Am i reading correctly when they say 4core cpu??? Who in their right MIND wants a 4 core anymore...
     
    MonstroMart likes this.
  14. Reddoguk

    Reddoguk Ancient Guru

    Messages:
    2,660
    Likes Received:
    593
    GPU:
    RTX3090 GB GamingOC
    Depends on the price. I agree 4 cores is weak in 2019 but if IPC is really that much better and it costs £100, 8 threads/cores is still plenty enough to be a good gaming chip.

    Just like the G4500 which beat most cpus in gaming @ the time with only 2 cores. Maybe because games @ the time were only using 1 or 2 cores max but you get the point right.
     
  15. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    What about 10% more performance with 25% lower power draw? Have you seen that? I did. Now my card happily ticks in family member's PC together with rest of components set up for power efficiency.

    I see that you like to repeat worst things you read somewhere. And you bring here 1060 as some winning argument. While it is argument showing utter defeat of sense.
    1060 and cards under it on nVidia's side were worse performers than AMD's counterparts at same price points. In some cases AMD provided even 50% higher performance per $.
    Sure, there was quite lower power draw on nVidia's side, but what do you expect from much slower HW.

    Every time nVidia managed outsell AMD 5:1 with worse product, customers ensured them that they can do whatever they want to. And so nVidia did.
    Message to AMD was different: "Whatever you do, we do not care." And so AMD did stop providing much better performance per $.

    But to the lying. True lies came from many review sites. Mostly unintentionally, because they did not have skill set to actually understand OC.
    I have seen crypto miner communities. And as much as I dislike mining, those people pushed boundaries of what can be achieved through BIOS modifications far more that any traditional tech site.
    Tragic, seeing random nobody to do better job at extracting performance from graphics card than some revered tech site.

    Some tech sites were completely stupid, like that one which claimed that Fury X's loop liquid reached boiling point. While GPU could not even reach such temperature due to thermal shutdown trigger being much lower.
    - - - -
    intel not lying? All their mobile SDP/TDP values are lies.
    nVidia not lying? False HW specifications. Half dozen of variants of one GPU sold under one name?
    AMD lies? Like one case of RX-570 which was marketed under RX-580 in China? Box clearly stated 2048SP version.

    Want to throw generic accusations? Be specific! Maybe we can learn from you for once.
     
    carnivore likes this.

  16. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    The scores can change wildly between releases, its not a benchmark if you keep changing the mark.
    If you make a test , it should remain fixed until you devise a new test. Cpuid has a problem with this , they like to "fix" their tests after the fact, why bother listing the benchmark version if it changes constantly.
    [​IMG]
    I ran all of these on a e5450 at 3.9ghz, same session of windows with all of them loaded first, ran each 3 times incase it randomly scored low or high, the scores did not vary more than 2 points between runs
    .
    note that all of these versions list the same benchmark version.

    I would not have a problem if they listed different benchmark versions, but they dont, they just "update" the test without telling anyone

    point is you cant use cpuz score as a reliable benchmark(specifically for comparison, between different systems, unless special care is taken to make sure the versions are exactly the same ). and no one should.
     
    Last edited: Jun 18, 2019
    Fox2232 likes this.
  17. BlackZero

    BlackZero Guest

    The CPU can't score the same results each time if there's other processes taking up CPU time.

    It's clearly a pretty sensitive test, which is a good thing.
     
  18. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    I loaded all the programs first, then ran them separately, background load is the same for all of the instances at run time. seeing variance is not a good thing in this case.
     
  19. BlackZero

    BlackZero Guest

    I don't see such variances unless there's OS processes doing tasks.

    For a 30 second test, sensitivity should be as they have it.
     
  20. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    for each instance, i only saw a 2 point variance between 3 runs , the differences you see are almost entirely due to differences between versions.
    If they were the same test they would not vary more than 2 points
     

Share This Page