Page 143 of 158 FirstFirst ... 4393133141142143144145153 ... LastLast
Results 1,421 to 1,430 of 1571

Thread: Kepler performance

  1. #1421
    It is listed as a gtx680, although it does come up when I search for gtx670. Quite mysterious indeed, and it could be a misnamed gtx670.

    If it is, it doesn't really undercut the HD7970, although 1 SKU isn't really a good enough sample to check prices.

  2. #1422
    640k who needs more?
    Join Date
    May 2010
    Posts
    863
    Quote Originally Posted by apu4life View Post
    @APU_Fusion

    When I say caught with their pants down,
    I was mostly thinking to this current situation where AMD has to sweat to even brake even on performance with their higher manufacturing cost GPUs, against Nvidia's low cost/small die GPU.
    Cost is determined not only by die size, but also by how many working dies you get from each wafer (yields). And with the yields currently being reported, my guess is that Kepler at this time, probably is quite a bit more expensive than Tahiti.

    Quote Originally Posted by apu4life View Post
    With AMD's traditional binning/clocks vs Nvidia dynamic boosting, AMD is on a huge disadvantage here.
    Even their projected 1GHz edition should bring kinda lowish 75MHz.

    So when 28nm supply issues inevitably do get resolved,
    (mind you USA is the only country where 680 is supply constrained), Nvidia will just take another piece of AMD discrete GPU marlket.
    Definitely not what you see here (Sweden). And being regarded as an "early adopter market", Sweden usually get a proportionaly big allocation from the start. The 7970 intro being a very good example of that.

    This not being based on stock levels at the biggest e-tailers, but on the info they have posted about coming deliveries and stock at their suppliers.

  3. #1423
    640k who needs more?
    Join Date
    May 2010
    Posts
    863
    Quote Originally Posted by apu4life View Post
    casual look give 12 SKUs being IN STOCK at Swedish retailers.

    http://www.prisjakt.nu/search.php?query=gtx+680
    A closer look shows it's very few cards, from small vendors. A handful of cards here and there is a far cry from readily available. When I check availability I go to the major e-tailers and look.

  4. #1424
    Senior Member
    Join Date
    Jan 2010
    Posts
    1,889
    Quote Originally Posted by PabloBroon View Post
    Nothing wrong with that unless every retail 670 struggles to go past 980mhz.

    Far too early to jump on that bandwagon, especially since i don't remember reading about 680's not going past 1058.

    The 670's are just cut down 680's, it's expected they could clock to similar speeds, but does make the 680 and 690's look poor in my opinion, but there will always be a market for the epeen people.
    No, they are not just cut down 680's. They are chips that can't be 680's. Lower binned, defective, whatever label you prefer. If they had 680 chips, they wouldn't cut them down to sell for $100 less. They'd make 680's with them. I'm not trying to say you don't know this. It's just that saying they are "just cut down 680's" isn't correct.

    Quote Originally Posted by apu4life View Post
    Nvidia does not say Boost Clock is max. clock that the card will not overshoot.


    Actually quite the contrary:
    I think the problem arises that if I buy a Kepler card I'm not guaranteed to get one that will perform like the review samples did. Another issue here is they can send ringers to the reviewers and then point to the guaranteed specs and say, "We never guaranteed you'd get performance like the review samples. Some people will get performance better than the reviews, some worse. It's the luck of the draw".

    This is the equivalent of AMD's AWSUM switch or the unlocked bios on 6950's being the only way cards were tested, even though you weren't guaranteed to achieve those performance figures. At least in those instances the reviewers were capable of testing samples in a state the consumer was guaranteed to get.

    The reviewers need to put in bold print:
    According to nVidia's published specifications for this product, our sample is non-typical. It performs better than what nVidia guarantees the cards they sell will. While we endeavor to bring you reports that accurately show performance that you can expect the retail product to achieve, this might not be the case with this product as nVidia gives us no way to test at the guaranteed specifications.

    Actually, I'm wrong. nVidia doesn't guarantee that your card will boost at all. The boost specs are only "typical" not guaranteed at all. This feature should be able to be shut off for reviews.

  5. #1425
    I wish someone would do a benchmark that graphed the GPU clock compared to FPS.

  6. #1426
    Good point Relayer, lower binned would of been more accurate, but since i was trying to point out how similar it was to the 680, i felt cut-down worked better.

    As for the boost thing, i thought that was guaranteed, either way i've not saw anyone who has a 680 not get at least the advertised boost speed yet though.

  7. #1427
    Senior Member
    Join Date
    Jan 2010
    Posts
    1,889
    Quote Originally Posted by PabloBroon View Post
    Good point Relayer, lower binned would of been more accurate, but since i was trying to point out how similar it was to the 680, i felt cut-down worked better.

    As for the boost thing, i thought that was guaranteed, either way i've not saw anyone who has a 680 not get at least the advertised boost speed yet though.
    I have little doubt that these chips will achieve the claimed average boost rate. I'm not so sure that they will be able to achieve the clocks that review samples do.

    [H]OCP
    *snip*
    (Kyle saw a GTX 680 sample card reach over 1300MHz running live demos but it could not sustain this clock.) The actual limit of the GPU clock is unknown. As each video card is different,
    So, when I read [H]'s review should I feel confident that his performance will accurately portray what I can expect of a 680 that I buy? If not, and I don't believe I should, then what's the benefit of the review? None to me as a consumer. It's very beneficial to nVidia though to have a card that performs this good in the hands of a reviewer who's readership is as large as [H]'s. The cards need to be reviewed at a performance level that I'm guaranteed to have from the card I buy. GPU boost needs to be able to be turned off if nVidia isn't going to guarantee I'll get the same clocks as the reviewer's samples get.

  8. #1428
    The only way to test the guaranteed performance of a 680 is to get one (3? 5?) at retail and bench it with a steady stream of pre-heated air.

    Using a heat-gun to warm the air around the card, boost should be disabled and real retail minimum performance can be established

    Why haven't any reviewers tried this???



    *idea isn't mine, but found on these forums

  9. #1429
    Quote Originally Posted by integrated View Post
    The only way to test the guaranteed performance of a 680 is to get one (3? 5?) at retail and bench it with a steady stream of pre-heated air.

    Using a heat-gun to warm the air around the card, boost should be disabled and real retail minimum performance can be established

    Why haven't any reviewers tried this???



    *idea isn't mine, but found on these forums
    Yeah, they should get a sizable retail sample (at least 4), and ignore the lowest and highest scores (if they are too apart from the others). Also, using a standard, relatively high, room temperature should be a must.

  10. #1430
    You guys may want to take a look at this thread here:

    http://www.overclock.net/t/1253432/g...sion-hands-on/

    This guy got his 670, and it's boosting to around 1.3GHz I think, that's OCed. It's beating every single 680 shown, at least in 3DMark11 and Heaven.

    Pretty impressive.

    I wonder how the 670 is boosting so much, though. The headroom there is significant.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux