How exactly did Intel make a 15W version of the 7920HQ?

Discussion in 'Hardware Components and Aftermarket Upgrades' started by Peon, Oct 18, 2017.

  1. Peon

    Peon Notebook Deity

    Reputations:
    406
    Messages:
    1,980
    Likes Received:
    125
    Trophy Points:
    81
    Usually, lowering TDP by 66% would require a die shrink or at least a new architecture, but the 8650U is neither...
     
  2. yrekabakery

    yrekabakery Notebook Guru

    Reputations:
    22
    Messages:
    69
    Likes Received:
    66
    Trophy Points:
    26
    By throttling it.
     
  3. Starlight5

    Starlight5 I'm a cat. What else is there to say, really?

    Reputations:
    215
    Messages:
    2,195
    Likes Received:
    764
    Trophy Points:
    131
    Lower voltage & lower power limits.
     
  4. tijo

    tijo Sacred Blame Super Moderator

    Reputations:
    7,011
    Messages:
    10,022
    Likes Received:
    932
    Trophy Points:
    531
    There's always some variations in between CPUs so they are likelier to take the better ones for those rather than for the HQ parts.
     
    Falkentyne likes this.
  5. bennyg

    bennyg Notebook Deity

    Reputations:
    308
    Messages:
    1,176
    Likes Received:
    482
    Trophy Points:
    101
    Max Spec Marketing chalks another victim methinks

    TDP and base clock indicate the behaviour of the CPU under sustained workload. While it *can* turbo to similar 4ghz+ speeds, the 15W CPU will quickly drop off in a direct comparison with both CPUs under sustained load to maintain its lower power limits. Turbo only adds frequency beyond the base clock when the power and thermal limits allow it.

    The user experience may be indistinguishable under very light use, single webpage, opening a text only document, but try and run an intensive game or encode a video for example they will be chalk and cheese.

    Intel has done nothing special, just creating different performance profiles for insertion into different products - the U CPUs usually go into tablets and small/light/thin laptops vs the high end HQs into bigger laptops and, if you buy the wrong brand, high end gaming laptops :p

    Any CPU has this characteristic, for example my 3940xm running a 32M Pi (multicore 100% workload) in TSBench locked at various speeds comparing power consumption, voltage and performance (time to complete a fixed workload)
    1.0GHz - 14.5W 0.93V - 32.6 sec
    1.5GHz - 18.5W 0.94V - 21.6 sec
    2.0GHz - 22W 0.95V - 14.8 sec
    2.5Ghz - 26W 0.96V - 12.1 sec
    3.0GHz - 32W 1.03V - 9.8 sec
    3.5GHz - 51W 1.21V - 8.0 sec
    4.0GHz - 61W 1.29V - 7.4 sec
    4.5GHz - 88W 1.4V - 6.6 sec

    Intel could take this same exact piece of silicon, call it a 15W 1GHz base clock part or a 4.5GHz base clock 88W part, or anywhere in between.
     
    Last edited: Oct 18, 2017
    Ashtrix, SkidrowSKT, jaug1337 and 7 others like this.
  6. Silvr6

    Silvr6 Notebook Evangelist

    Reputations:
    142
    Messages:
    360
    Likes Received:
    81
    Trophy Points:
    41
  7. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,633
    Messages:
    13,551
    Likes Received:
    15,694
    Trophy Points:
    931
    It is "briefly" way faster, then it times out on the Turbo and ceases "melt-down".

    Benchmarks that favor Turbo Timeout's - running only as long as the timeout's are operating - can make these new 15w CPU's look competitive with 45w / 91w / 95w CPU's, and for some applications they may be.

    But if you have long running things, the thermal headroom and power headroom will disappear before the job is over, and power throttling + thermal throttling will make them fall apart.

    It's an illusion, don't fall for it :)
     
  8. Silvr6

    Silvr6 Notebook Evangelist

    Reputations:
    142
    Messages:
    360
    Likes Received:
    81
    Trophy Points:
    41
    Did you happen to read the review, yes they throttle, but they get up to speed and temperature after 1 benchmark run. So what you are saying is that all the other reviews showing this with a massive performance increase over the dual cores is wrong? Or the fact it is as fast as an i5 7300HQ which is 45W TDP. Maybe you should read my thread with non scientific benchmarks comparing the 2 laptops in my Sig. For intents and purposes. The 15W i7 8550u beats out my i7 3840QM,

    http://forum.notebookreview.com/thr...ing-behavior-on-dell-inspiron-13-7000.809595/

    Its no illusion and I haven't fallen for anything i've just managed to replace my M4700 with an Inspiron 13 and didn't lose any performance like most people proclaim you do.
     
    tilleroftheearth, Peon and Vasudev like this.
  9. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,633
    Messages:
    13,551
    Likes Received:
    15,694
    Trophy Points:
    931
    You are answering what I already said. The short run benchmarks show great performance, but when the timer runs out and it needs to power throttle before it thermal throttles, it's game over - it turns from a beautiful princess into a dust bin.

    Hey, believe what you want to believe, if you don't look at it too hard, or test for long runs, you'll remain happy, I'd hate to take that away from you.

    For everyone else, watch yourselves it's just a 15w CPU, it can only pull the rabbit from out of the hat for a short time. :)

    5 Minutes on Tech: Intel 8th Gen CPU - Worth Upgrading?

    The new 8th gen are a nice upgrade within the 15w realm, but 45w/95w CPU's have nothing to worry about :)
     
    Last edited: Oct 20, 2017
    Peon and Vasudev like this.
  10. tilleroftheearth

    tilleroftheearth Wisdom listens quietly...

    Reputations:
    4,335
    Messages:
    11,860
    Likes Received:
    1,887
    Trophy Points:
    631
    Yeah, yeah...

    Keep believing what you want to be true, but believing isn't making it so. ;)

    That video is laughable. She's comparing an i7-7700HQ to an i5-8250u, uhm... okay.

    See:
    https://ark.intel.com/compare/97185,124969,124967,122589,124968


    See:
    https://www.cpubenchmark.net/compare.php?cmp[]=3042&cmp[]=3064&cmp[]=2906


    If we try a little harder to compare apples to apples, we see that at almost double the TDP (assuming we're using a chassis with TDP configured to 25W for the i7-8550U) the 7th gen HQ model is about 11% faster in multicore tasks. Not a good trade off for a mobile platform, imo.

    The biggest trade off for the still not comparable i7-8650u top 8th gen model right now is double the RAM to 64GB total and 4 extra PCI lanes for the HQ. While my workflows demand the most RAM I can stuff into my platforms; battery life is equally important for a mobile system too, for me.

    What is interesting is that the u model is equal to the hq model in single threaded 'responsiveness'. That is the icing on the cake for me.

    Will the 8th gen processor throttle in a 15W capable chassis? Uh, yeah. That's because they're able to work at up to 25W TDP in an appropriately designed system.

    With same 14nm tech size, this is a phenomenal increase in performance at the lower TDP. Ignore at your own risk. :)

     
    Peon and Starlight5 like this.
Loading...

Share This Page