Nvidia GeForce GTX 1080 Ti Arrives late March

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Feb 20, 2017.

  1. Denial

    Denial Ancient Guru

    Messages:
    12,382
    Likes Received:
    1,618
    GPU:
    EVGA 1080Ti
    I don't think they change their mind about their goals - they just change their marketing focus. The goal for the architecture is probably the same the entire way through.

    Pascal, GP100 for example packs pretty decent DP numbers over Maxwell (since they basically skipped DP in Maxwell) yet Nvidia's marketing focused more on FP16/INT8 performance because they are trying to get more traction in the compute/deep learning space. The design goal didn't change, the focus of marketing did - all the design goals were met.
     
  2. Netherwind

    Netherwind Ancient Guru

    Messages:
    6,858
    Likes Received:
    1,026
    GPU:
    RTX 2080Ti
    Come to papa!
     
  3. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,004
    Likes Received:
    137
    GPU:
    Sapphire 7970 Quadrobake
    Is 12nm even "out"? If they deliver parts of it for the auto industry by the end of the year, doesn't that mean that they have to have something taped out of that process on June at the latest? What I am basically trying to say is that if Pascal was indeed Volta, then it would be called as such and be done. Volta seems to be as radical a redesign as Fermi was, and for one reason or the other (for which we can all argue without actually knowing), it is definitely delayed.

    And to the subject itself: I'm not sure I would be comfortable to give ~$700-$850 for something on the Titan XP level of performance, with packing even less hardware.
     
  4. Denial

    Denial Ancient Guru

    Messages:
    12,382
    Likes Received:
    1,618
    GPU:
    EVGA 1080Ti
    12nm is just an improved 16nm, I guess TSMC is marketing it as 12 because they are miffed Samsung marketed theirs as 14nm lol. There isn't that much information about it - but AIDA64 is reporting that Volta (specifically GV100) is showing up in Nvidia's drivers - which in the past indicated that they have engineering samples already. No one knows if they are on the improved 16nm process though. That being said, TSMC specifically stated that it was designed for high performance/high frequency. So aside from Nvidia, I can't think of anyone else that would be using it - all the phone SoC's launching this year are already 10nm.

    No one knows what Volta is, it was speculated that it was a radical redesign but the evidence of that was from Pascal being added. That entire rumor spawned from when they moved Volta back to 2017 and added Pascal. People assumed it was too big for one release so they split the release into two parts. It's kind of similar to what AMD did with Polaris/Vega - they focused on the front end with Polaris and the back end with Vega. Pascal got a bunch of "front end" changes - Volta is rumored to have back end changes. But it's all speculation. The only thing we know 100% was that Nvidia stated that the biggest, number #1 feature for Volta was stacked DRAM. Which Nvidia obviously did with Pascal. So yeah, the "name" got delayed - but it's biggest feature didn't and obviously Pascal had a bunch of other features as well which all most likely fell under what they planned for Volta anyway.

    As for the Ti, I wouldn't be surprised if Nvidia put the MSRP at $650 and FE edition at $750. You can already find factory overclocked 1080's for $580 and every once in a while they go as low as $565 despite the MSRP being $600. I can definitely see Nvidia dropping the MSRP of the 1080 down to $500 and slotting the Ti at $650.. especially if Vega is competitive (which it looks like it will be).
     
    Last edited: Feb 21, 2017

  5. zipper

    zipper Maha Guru

    Messages:
    1,029
    Likes Received:
    79
    GPU:
    GTX 680M
    NVIDIA_DEV.1D81 = "NVIDIA Graphics Device [GV100]" showing in the latest Quadro 378.66 driver.
     
  6. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    You are probably more right than my statement. They just picked value which gained highest percentage boost.

    Side Joke:
    Let's make study where we will have 2 groups of 100 people. 1st 100 will eat carrot every day. 2nd group will not eat carrot.
    - measure change in eye sight.
    - measure blood pressure
    - measure cholesterol
    - measure aggressive tendencies
    - measure ability to concentrate on complex tasks
    - measure pain response
    - ...

    At the end of the study, we failed to find reasonable statistical correlation between eating carrot and good eye sight. But we were successful in confirming that carrot increases cholesterol levels, aggressivity and reduces ability to concentrate on complex tasks.
    - - - -
    We measures following events:
    - amount of rain in between January and April
    - peak wind speed in January to February
    - number of jellyfish stings in April and May
    - ...

    Our study was successful in confirming that sightings of Blue Butterflies as early as March is indication with 98% success rate that we will have volcanic eruption in between September and December.
    Having too many things to measure is often unhealthy if they are used in any way it suits purpose.
    That's why proper scientific study is done in order to confirm one thesis and nothing else as chance for false positive drastically increases with any additional measured cause or impact.
     
    Last edited: Feb 21, 2017

Share This Page