The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.
Problems? See this thread at archive.org.
← Previous pageNext page →

    NVIDIA Geforce GTX 660M Release Information + CUDA Core Count

    Discussion in 'Gaming (Software and Graphics Cards)' started by yknyong1, Mar 22, 2012.

  1. maxheap

    maxheap caparison horus :)

    Reputations:
    1,244
    Messages:
    3,294
    Likes Received:
    191
    Trophy Points:
    131
    tell me about it! :)
     
    maxheap, Mar 27, 2012
    #101
  2. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    You reckon they are going to disable half the units on kepler to get it working in a notebook?
     
    Meaker@Sager, Mar 27, 2012
    #102
  3. maxheap

    maxheap caparison horus :)

    Reputations:
    1,244
    Messages:
    3,294
    Likes Received:
    191
    Trophy Points:
    131
    I think it will be GK104 based and they will cut 95w to bring it down to 100w, what I guess will be just disabling half the cores :) (but I have no idea definitely, as they managed to trim 70w from 560Ti at the expense of no CUDA cores, but memory interface / core clock / memory clock, definitely it is involving a ton of electrical engineering which I don't know squat :)) also half a gtx 680 is somewhat equal to 580, can you imagine gtx 580 performance in a lappy :eek:
     
    maxheap, Mar 27, 2012
    #103
  4. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    No, the 680 is not twice as fast as the 580.

    It will most likely have clock/voltage reductions since its TDP is not too bad already.
     
    Meaker@Sager, Mar 27, 2012
    #104
  5. nissangtr786

    nissangtr786 Notebook Deity

    Reputations:
    85
    Messages:
    865
    Likes Received:
    0
    Trophy Points:
    0
    the gt660m is a 40-45w 128bit card. I have seen nowhere where it says its 192bit.
     
    nissangtr786, Mar 27, 2012
    #105
  6. nissangtr786

    nissangtr786 Notebook Deity

    Reputations:
    85
    Messages:
    865
    Likes Received:
    0
    Trophy Points:
    0
    Meaker the gtx680m should be 2x performance per watt if its a kepler.
     
    nissangtr786, Mar 27, 2012
    #106
  7. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    I think the post was edited, but since kepler no longer has a shader clock and 2 shaders are needed to do the same work as before it does not really work that way.

    Also at no point was performance per watt mentioned nissangtr.
     
    Meaker@Sager, Mar 27, 2012
    #107
  8. nissangtr786

    nissangtr786 Notebook Deity

    Reputations:
    85
    Messages:
    865
    Likes Received:
    0
    Trophy Points:
    0
    nissangtr786, Mar 27, 2012
    #108
  9. ryzeki

    ryzeki Super Moderator Super Moderator

    Reputations:
    6,552
    Messages:
    6,410
    Likes Received:
    4,087
    Trophy Points:
    431
    It is valid to say you want a low power consumption machine that gives you great performance on the go. And that's a very good reason actually.

    But don't be delusional with it's performance. Half of those games are not run at their maximum details at all. Only Deus EX and Call of duty are confirmed max all. Batman has two more presets, Battlefield one more etc etc... And check the graph correctly. You are talking about a 30-40fps experience at those settings already at best. Only Dirt 3 managed higher performance. From the graph alone by nvidia, the 570m is around 20% faster on stock.

    I imagine this is because 384 Cuda cores running around half the shader speed than current fermi is basicaly near the 192 cores of the GTX460/560m which is why it has similar performance in the end. so the 680m will be at least running 768 cores at less power, and should have a nice speed bump.

    In the end, it's good to have a low power card that you can take on a smaller package. I am not against such progress. I am just pointing out performance isn't enough for a 60fps, 1080p, highsettings at all. I am talking only regarding performance, but in my case, I am interested in higher performing products, not just the same performance on lower power envelope (for my machine).

    thats why I am interested in the higher end versions of Kepler (680m) and AMD's equivalent in GNC. I want to see what 70w-100w tdp can get me now.
     
    ryzeki, Mar 27, 2012
    #109
  10. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Oh Cloudfire, sorry to burst your bubble, but

    NVIDIA has been known to use interleaved memory configuration on memory bus.
    Something like 'flex' interleaving.
    So it doesn't simply mean one can judge the bus width from the memory size nowadays.

    Hope the source below is informative enough :)
    AnandTech - NVIDIA's GeForce GTX 550 Ti: Coming Up Short At $150

    And the chart quoted is for the GTX 560M, not 660M.
     
    yknyong1, Mar 27, 2012
    #110
  11. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    He was talking about the desktop chip, as for the performance compared to the 580M yes there will be a huge jump.
     
    Meaker@Sager, Mar 28, 2012
    #111
  12. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    1.5GB 128bit have never happend on mobile GPU. Atleast not with Fermi, 400 and 500 series. Go look up various gpu-z of 445M, 460M, 560M and 555M. You will not find any 1GB/2GB 192bit or 1.5GB/3GB 128bit. Period.

    And I posted the chart to give people an idea what to expect going from 128bit to 192bit, which I thought was pretty clear. Since Kepler have turbo boost, perhaps it will greatly benefit by pushing those precious turbo Gigabytes through a wide memory bus. Or perhaps the 1.5GB was a typo and that 128bit is really enough for GK107. I don`t know :)
     
    Cloudfire, Mar 28, 2012
    #112
  13. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    But I sure hope GK107 will have 192bit. It seems impossible as the desktop GK107 is also 128bit though :(
     
    yknyong1, Mar 28, 2012
    #113
  14. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    Crap, good thinking. Perhaps it is not needed for GK107 after all :(
     
    Cloudfire, Mar 28, 2012
    #114
  15. Prema

    Prema Your Freedom, Your Choice

    Reputations:
    9,368
    Messages:
    6,297
    Likes Received:
    16,485
    Trophy Points:
    681
    Prema, Mar 28, 2012
    #115
  16. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    Cloudfire, Mar 28, 2012
    #116
  17. plancy

    plancy Notebook Evangelist

    Reputations:
    56
    Messages:
    550
    Likes Received:
    0
    Trophy Points:
    30
    :). I would like to see if it stacks up to 560m in real world performance, I'm probably going to get a 660m and not a $800 GPU and game at 720p, put the extra in a desktop maybe.
     
    plancy, Mar 28, 2012
    #117
  18. TheBluePill

    TheBluePill Notebook Nobel Laureate

    Reputations:
    636
    Messages:
    889
    Likes Received:
    0
    Trophy Points:
    30
    Wicked. And here i thought progress peaked with frozen pizza. :D
     
    TheBluePill, Mar 28, 2012
    #118
  19. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    hahaha :p
    10 char
     
    Cloudfire, Mar 28, 2012
    #119
  20. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    LOL :eek:
     
    yknyong1, Mar 28, 2012
    #120
  21. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    I am pretty shure we will see benches with 660M very soon. Pictures with internals of Lenovo Y580 have been posted in various forums recently
     
    Cloudfire, Mar 28, 2012
    #121
  22. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    I wonder what the real frequencies were.

    It will have to clock very well to match the 570.
     
    Meaker@Sager, Mar 28, 2012
    #122
  23. plancy

    plancy Notebook Evangelist

    Reputations:
    56
    Messages:
    550
    Likes Received:
    0
    Trophy Points:
    30
    I wish there was a 192-bit variant, that would be insane performance...
     
    plancy, Mar 28, 2012
    #123
  24. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    yknyong1, Mar 28, 2012
    #124
  25. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    I don't get it, what are you saying about that link?
     
    Meaker@Sager, Mar 28, 2012
    #125
  26. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    "If you ask us, the GeForce GTX 680 is our best graphics card ever"

    I felt a bit sour that NVIDIA did not release the best chip for desktop ie GK110.
     
    yknyong1, Mar 28, 2012
    #126
  27. TheBluePill

    TheBluePill Notebook Nobel Laureate

    Reputations:
    636
    Messages:
    889
    Likes Received:
    0
    Trophy Points:
    30
    TheBluePill, Mar 28, 2012
    #127
  28. long2905

    long2905 Notebook Virtuoso

    Reputations:
    2,443
    Messages:
    2,314
    Likes Received:
    114
    Trophy Points:
    81
    Why should they when their lower cards are already winner of this round?
     
    long2905, Mar 28, 2012
    #128
  29. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    There are a couple of realistic reasons why the top end chip was not launching IMO:

    1) The chip is too defective and yields are not where they need to be.
    2) The chip is not clocking where it should.
    3) The GK104 was so close in performance it made no sense to release it.

    I think its 1 or 2.

    Also Long, because they are not really that far ahead, maybe a few % but if you had a higher end chip you could slap it on a £600 card and claim victory and rake in the cash.
     
    Meaker@Sager, Mar 28, 2012
    #129
  30. lordbaldric

    lordbaldric Notebook Consultant

    Reputations:
    2
    Messages:
    203
    Likes Received:
    6
    Trophy Points:
    31
    And this isn't even taking into account the crappy compute/OpenGL performance of the 680, which are no where near "good enough".
     
    lordbaldric, Mar 28, 2012
    #130
  31. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    GK110 was not done because
    A) A 256bit mid range GPU could beat a 384bit high end AMD GPU and therefor
    B) Didn`t have to rush and build a monstreous GPU and
    C) They wanted to show what a small Kepler could do in terms of efficiency and therefor didn`t go overboard with heat, consumption etc. Meanwhile
    D) AMD is working on the 8000 series and the Dual GPU 7990 which is what the GK110 should be aiming for which is also the reason why Nvidia wait.

    But yeah, your points could also be valid Meaker. After all Nvidia seems desperate now that the CEO have reached out to Intel and asking if they could use Intel`s factories due to low yield.

    And I agree with lordbaldric. The 680 is crappy in other things than gaming. 7970 is miles ahead
     
    Cloudfire, Mar 28, 2012
    #131
  32. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    yknyong1, Mar 29, 2012
    #132
  33. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    Yeah I don`t see that hot thing going inside a laptop :p
     
    Cloudfire, Mar 29, 2012
    #133
  34. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Watercooled laptop, anyone? :p
     
    yknyong1, Mar 29, 2012
    #134
  35. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    If Nvidia could release the GK100 and make a profit they would. But yeah, thanks to gk104 being better than expected (Part. the mem controller reaching 6GHz) they dont have to launch it, even if it looses money.
     
    Meaker@Sager, Mar 29, 2012
    #135
  36. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Request to show them :p
     
    yknyong1, Mar 29, 2012
    #136
  37. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    Why? Boring inside pics with copper and boring plastic :p
     
    Cloudfire, Mar 29, 2012
    #137
  38. TheBluePill

    TheBluePill Notebook Nobel Laureate

    Reputations:
    636
    Messages:
    889
    Likes Received:
    0
    Trophy Points:
    30
    Really..

    The first photo of the inside the Y580 and the new processor was cool;


    [​IMG]
     
    TheBluePill, Mar 29, 2012
    #138
  39. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    hahaha that made my day. :laugh:

    Too bad I can`t give you rep (will later)
    "You must spread some Reputation around before giving it to TheBluePill again."
     
    Cloudfire, Mar 29, 2012
    #139
  40. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    If you like that, then the eve cluster runs 10,000 of those in parallel to generate the computational power needed.
     
    Meaker@Sager, Mar 29, 2012
    #140
  41. TheBluePill

    TheBluePill Notebook Nobel Laureate

    Reputations:
    636
    Messages:
    889
    Likes Received:
    0
    Trophy Points:
    30
    EVE huh?

    [​IMG]
     
    TheBluePill, Mar 30, 2012
    #141
  42. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Cute thing dude :)
     
    yknyong1, Mar 30, 2012
    #142
  43. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    I'm wondering if given great cooling, would a GT 650M outperform a GTX 660M?
     
    yknyong1, Mar 30, 2012
    #143
  44. ryzeki

    ryzeki Super Moderator Super Moderator

    Reputations:
    6,552
    Messages:
    6,410
    Likes Received:
    4,087
    Trophy Points:
    431
    Sure, well they are all the same chip. Aside from memory interface. even a 640, overvoltage with great cooling could decimate the other two in general, except in high res.
     
    ryzeki, Mar 30, 2012
    #144
  45. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Now I wonder if overvolting is possible on Kepler ;)

    Would overclocking still have the boost????
     
    yknyong1, Mar 31, 2012
    #145
  46. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,436
    Messages:
    58,194
    Likes Received:
    17,902
    Trophy Points:
    931
    You can disable boost, but if you dont then yes it will still try and boost itself. Since the 650M (7xxmhz) and 660M (8xxmhz) will have different core voltage there may be some wiggle room for 650M users at least.
     
    Meaker@Sager, Mar 31, 2012
    #146
  47. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Yeah, repeat of GF108, GT 525M to GT 550M all same chip... One could easily OC 525M in the Dell XPS to 540M clocks.
     
    yknyong1, Mar 31, 2012
    #147
  48. yknyong1

    yknyong1 Radiance with Radeon

    Reputations:
    1,191
    Messages:
    2,095
    Likes Received:
    8
    Trophy Points:
    56
    Any chinese overclocking result for 650M/660M since Meaker has been sharing sources with us? :)
     
    yknyong1, Apr 1, 2012
    #148
  49. Cloudfire

    Cloudfire (Really odd person)

    Reputations:
    7,279
    Messages:
    10,304
    Likes Received:
    2,878
    Trophy Points:
    581
    Here is another 660M with what looks like a 192bit configuration. Weird

     
    Cloudfire, Apr 2, 2012
    #149
  50. Kevin

    Kevin Egregious

    Reputations:
    3,289
    Messages:
    10,780
    Likes Received:
    1,782
    Trophy Points:
    581
    You can tag 3GB onto 128-bit bus though...

    If it is 192-bit, that card will be sitting between the 670M and 675M, which then causes me to question why the 670M would be in existence at all. So I can only imagine that Nvidia wouldn't do something THAT confusing to the market.
     
    Kevin, Apr 2, 2012
    #150
← Previous pageNext page →