AMD Radeon VII Announcement and Discussion

Discussion in 'Architecture and Products' started by ToTTenTranz, Jan 9, 2019.

Tags:
  1. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    I can’t speak for GPU vendors, who might have their shit together more than others, and I can’t disclose specifics about the stuff that I do know first or second hand, but the difficulty and cost of a working HBM system is universally underestimated and doesn’t come close to traditional PCB based solutions.

    That should be sufficient, since nothing about my claim was extraordinary to begin with. ;-)
     
    Lightman likes this.
  2. SumoSaki

    Newcomer

    Joined:
    Jul 4, 2014
    Messages:
    71
    Likes Received:
    16
    Given the terrible power efficiency of this card versus a GTX 280 (TDP 215W) what's the thought they can improve per per watt for Navi? AMD have done such a bad job here it doesn't inspire confidence for PS5.
     
    vipa899 likes this.
  3. Rootax

    Veteran Newcomer

    Joined:
    Jan 2, 2006
    Messages:
    1,147
    Likes Received:
    570
    Location:
    France
    Well Navi is a new gpu, you hope they learned from Vega.
     
  4. itsmydamnation

    Veteran Regular

    Joined:
    Apr 29, 2007
    Messages:
    1,294
    Likes Received:
    394
    Location:
    Australia
    The idea that something is so bad at yielding/integration yet has mass market adoption ( google, amd, nv , intel, Fujitsu ,etc) is extraordinary. We see what happens when something yields poorly, its called intel 10nm.

    The next part is what doesn't count as coming close in terms of cost. if we are to break it down to cost per bit/gbit 1.5x, 2x , 10x ? How about cost per Gb of bandwidth?
     
  5. ToTTenTranz

    Legend Veteran Subscriber

    Joined:
    Jul 7, 2008
    Messages:
    9,780
    Likes Received:
    4,431
    I implied absolutely nothing of the sort.

    VII is actually the same GPU as a MI50. In memory amount and CU units enabled.
     
    #125 ToTTenTranz, Jan 10, 2019
    Last edited: Jan 10, 2019
    w0lfram likes this.
  6. vipa899

    Regular Newcomer

    Joined:
    Mar 31, 2017
    Messages:
    922
    Likes Received:
    354
    Location:
    Sweden
    Was thinking the same, all we have is rumors about Navi, but nothing seen yet. One could think the GPU/CPU design should be allready done, if not allready planned to produce? Theres no valid reason to not show Navi at CES if it was ready or anything high-end that performs above nvidias offerings.
    Everything points Navi to being a mid-range/low power GPU/Chip, makes perfect sense for a console to be honest.
     
    Picao84 likes this.
  7. AlphaWolf

    AlphaWolf Specious Misanthrope
    Legend

    Joined:
    May 28, 2003
    Messages:
    8,395
    Likes Received:
    247
    Location:
    Treading Water
    "Hey here's our new product please buy lots of them, oh by the way we'll be shipping a much better product next year."

    Is that how you would handle product announcements?
     
    snarfbot, Cat Merc, Lightman and 2 others like this.
  8. vipa899

    Regular Newcomer

    Joined:
    Mar 31, 2017
    Messages:
    922
    Likes Received:
    354
    Location:
    Sweden
    If AMD has a matching GPU in the works releasing this summer (or isnt it this summer?), then why not showing it off or atleast some specs, or a sign it can actually keep up with Nvidia? I mean this RadeonVII gpu could mean people loose faith in AMD, by again doing another product that doesnt compete with Nvidia? Quite short-sighted otherwise to handle it like this, big companies like AMD think in the long run, very long run. Those stocked vegas would perhaps mean a loss for the first few months if even that, but launching yet another dissapointing product doesnt seem the way to go.

    People where expecting a 'come back' from AMD at CES2019, it was rather a dissapointment. I think we all want competition, it will benefit us all, those Nvidia prices are way too high imo. No matter how good their performance is, it needs to be more affordable to reach a bigger public.
     
  9. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    Google, Nvidia and Fujitsu are using HBM only for HPC products where silicon cost is just a minor fraction of the system price.

    One where silicon yield can be low and component cost high without material impact on gross margins. Products that can’t be categorized as mass market adoption at all.

    Nvidia uses GDDR for all its mass market products despite the fact that HBM is a higher performing solution. If cost were close to equivalent, why would they do that?

    The only valid point you bring up is Intel for its AMD GPU design. One that can only be found in products that are priced much higher than usual, and more than likely a one off. One that also only has 1 HBM site, which lessen packaging yield losses considerably.

    The problem with Intel’s 10nm wasn’t that it was just yielding badly, it was that yield was close to 0. Nobody said that HBM was yielding close to 0. But it’s a solution with low volume and expensive base components, and with an expensive packing system that has much lower yields than the alternative.

    Perfect for high priced products, lousy for mass market ones.

    Unfortunately for AMD, they are forced to use it to work around their power inefficient GPU architecture.
     
    Picao84, pharma, yuri and 1 other person like this.
  10. Tkumpathenurpahl

    Veteran Newcomer

    Joined:
    Apr 3, 2016
    Messages:
    1,013
    Likes Received:
    743
    I think I was quite clear, and your confusion is artificial.

    Vega VII offers comparable performance to the RTX2080, except for RT, of course.

    As long as they can offer lower prices or enticing bundles, they're competitive. Simple.
     
    Picao84 and ToTTenTranz like this.
  11. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    Depends on POV.

    Since we’re in the architecture part of the forum, I’d say it’s a very underwhelming GPU, just like Vega 10.

    Can it be made attractive for a consumer? Of course, by slapping a low price on it. But that doesn’t make it any technically underwhelming.
     
    Lightman likes this.
  12. AlphaWolf

    AlphaWolf Specious Misanthrope
    Legend

    Joined:
    May 28, 2003
    Messages:
    8,395
    Likes Received:
    247
    Location:
    Treading Water
    The goal of them announcing new products is to actually sell them, not to quell your personal disappointment in their lineup.
    It doesn't make sense to sabotage a product announcement with a further out product announcement which might make consumers wait on a purchase, making the immediate product DOA.
    As for pricing, it is a matter of what the market will bear. It is apparent that $700 is not too much if the performance fits.
     
  13. A1xLLcqAgt0qc2RyMz0y

    Regular

    Joined:
    Feb 6, 2010
    Messages:
    980
    Likes Received:
    268
    Or they are selling it for a loss and making it up on volume and software sales like TI tried with their ill fated console.

    HINT: That did not work out for TI.
     
  14. vipa899

    Regular Newcomer

    Joined:
    Mar 31, 2017
    Messages:
    922
    Likes Received:
    354
    Location:
    Sweden
    It is also important to think in the long run, yet another flop might do damage in the long run. Everywhere on youtube, forums etc you see people being dissapointed with AMD's showing, they were expecting Navi perhaps. MS first Xbox lost them 4billion, but thats an investment for the future.
     
  15. Tkumpathenurpahl

    Veteran Newcomer

    Joined:
    Apr 3, 2016
    Messages:
    1,013
    Likes Received:
    743
    Okay then Obi Wan ;)

    It's technically underwhelming in the same sense that Vega, generally, is pretty underwhelming. But it is what it is: a 7nm variant of Vega.

    Do you criticise cheetahs for being poor swimmers? If not, maybe consider not criticising 7nm Vega for being Vega.
     
    ToTTenTranz likes this.
  16. rcf

    rcf
    Regular Newcomer

    Joined:
    Nov 6, 2013
    Messages:
    370
    Likes Received:
    302
    Nvidia's 16nm GP100, released in Q1/2016, would probably beat AMD's 7nm Vega20 released in Q4/2018.
     
    Picao84 and yuri like this.
  17. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    It was perfectly possible for AMD to make some incremental fixes while keeping the same core architecture. Think Polaris compared to earlier GCN chips in terms of IPC or lower level power efficiency changes.

    I wonder if they did even that.
     
    pharma likes this.
  18. silent_guy

    Veteran Subscriber

    Joined:
    Mar 7, 2006
    Messages:
    3,754
    Likes Received:
    1,379
    Not saying anything about Navi instead of overhyping who fantastic their next GPU be half a year down the road is a remarkable improvement!
     
    nnunn and vipa899 like this.
  19. Geeforcer

    Geeforcer Harmlessly Evil
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    2,297
    Likes Received:
    464
    I think VII will have a very tought time competing with 1080TI. At this point Pascal Vs Vega is a settled affair, and VII is not bringing any new features from a consumer perspective. As such, you have comparable performance at much higher price and worse power consumption Vs a card that had been out for two years.
     
    #139 Geeforcer, Jan 11, 2019
    Last edited: Jan 11, 2019
    vipa899 likes this.
  20. mrcorbo

    mrcorbo Foo Fighter
    Veteran

    Joined:
    Dec 8, 2004
    Messages:
    3,563
    Likes Received:
    1,980
    If you could still buy 1080tis, the 2080 wouldn't be competitive with it either. Not at the price point it ended up at.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...