Speculation: GPU Performance Comparisons of 2020 *Spawn*

Discussion in 'Architecture and Products' started by eastmen, Jul 20, 2020.

Thread Status:
Not open for further replies.
  1. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    Actually decreased # of xtors.
    A thousand cuts made whole.
    They'll have to throw a properly long and annoying ISSCC presentation to explain all of it.
    Not even twice.
    That's reserved for N31.
     
  2. troyan

    Newcomer

    Joined:
    Sep 1, 2015
    Messages:
    235
    Likes Received:
    426
    So AMD is not doubling bandwidth? 3080 competition is it then. And no, i dont believe that they will beat nVidia with data compression and GDDR efficiency.
     
  3. JasonLD

    Regular

    Joined:
    Apr 3, 2004
    Messages:
    418
    Likes Received:
    60
    I said better reason lol. Series X technical paper is convincing enough that whatever perf/w figure they are touting is purely on node improvement, not on the architecture level, and even that is meaningless without knowing what is being compared.
     
  4. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    No.
    No.
    Ha.
    what.
    What node improvement?
    N7+ over N7p is like 5% better power.
    They're comparing fmax to fmax parts too, so you must understand that fmax went way up.
    You didn't even see it.
    Silence.
    You just know.

    Get you goddamed popcorn ready.
     
    #124 Bondrewd, Aug 23, 2020
    Last edited: Aug 23, 2020
  5. JasonLD

    Regular

    Joined:
    Apr 3, 2004
    Messages:
    418
    Likes Received:
    60
    That is called a node improvement

    Oh, I read it, both old RDNA and Series X hot chips presentation. and telling people to silence shows a lot about your character.
     
  6. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    5% to 50%?
    That's some 0 missing.
    That 5% power in vacuum trasnalated to like 2% fmax and that's not the case.
    At all.
    Read them harder.
    Yeah, gotta make sure no one says anything that could age badly.
    I dropped some unkind words about AMD's chances in mobile and now have to be silent and watch.
    Quietly.
     
  7. madhatter

    Newcomer

    Joined:
    Jul 23, 2020
    Messages:
    21
    Likes Received:
    17
    Going from N7P to N7+ wouldn't bring the huge improvement that we see in the Series X GPU over RDNA GPUs. It has a 52CU GPU at a slightly higher sustained frequency than the 5700 XT's game clock, yet consumes significantly less power than a 5700.
     
    Krteq likes this.
  8. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    Worse, XSX isn't even N7+.
    EUV is deemed too voodoo for consoles.
    Yes, this is the real shit.
    That's how they plan to win gaming mobile with N22/23 and SmartShift.
     
  9. madhatter

    Newcomer

    Joined:
    Jul 23, 2020
    Messages:
    21
    Likes Received:
    17
    True, I completely forgot about that.
    Those look like they will be more than competitive against Ampere given recent events, I just wonder in how many laptops they'll make it. Cezanne+N23 will be a killer combo for sure.
     
  10. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    A lot.
    N22 too.
    CZN is systems on shelf I think mid-late January'21; not much left to wait.
     
    madhatter likes this.
  11. JasonLD

    Regular

    Joined:
    Apr 3, 2004
    Messages:
    418
    Likes Received:
    60
    Plenty enough, it is just RDNA1+RT

    And you don't get to decide that.

    5700xt can be driven with lot less power without a big performance loss when undervolted. Plus, Enhanced N7 they are using on Series X is bringing 10% improvement on power on same frequency. I think this is the reason why Big Navi with RDNA2 is getting overhyped.
     
  12. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    Bold.
    Oh lord.
    It's a console. It's inherently overvolted for yield salvaging.
    Navi1x is already N7p bro.
    The first shipping N7p part at that, ironically enough.
    No there's no hype besides me shitlording here and on twatter.
    AMD isn't even talking about Milan, Vermeer, Van Gogh, Cezanne and RDNA2 parts.
    At all.
    And those are literally 1-5 months away.
     
  13. Kaotik

    Kaotik Drunk Member
    Legend

    Joined:
    Apr 16, 2003
    Messages:
    9,350
    Likes Received:
    3,340
    Location:
    Finland
    So are you saying Sienna Cichlid is N31 now? Or that Sienna Cichlid isn't RDNA at all?
    Since we have Sienna Cichlid in Linux patches/drivers right now with HBM
     
  14. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    No.
    There's also patches for it with GDDR6.
    Paranoid, survive, yadda, yadda, yadda.
     
  15. SimBy

    Regular Newcomer

    Joined:
    Jun 21, 2008
    Messages:
    641
    Likes Received:
    313
    I mean 2 x 5700XT performance is perfectly reasonable to expect without overhyping anything. That's 33% faster than 2080Ti. I mean they even call the damn thing Navi2x. How much more obvious can it get?
     
    Lightman likes this.
  16. Bondrewd

    Veteran Newcomer

    Joined:
    Sep 16, 2017
    Messages:
    1,042
    Likes Received:
    441
    They call it because everyone kinda adopted their die numbering scheme in usual conversations.
    It's a 4die lineup with some extra bits.
     
  17. no-X

    Veteran

    Joined:
    May 28, 2005
    Messages:
    2,349
    Likes Received:
    326
    What did it reveal about RDNA2's RBEs and rasterizer?
     
    Lightman likes this.
  18. pTmdfx

    Regular Newcomer

    Joined:
    May 27, 2014
    Messages:
    341
    Likes Received:
    283
    Also physical design? :razz:

    Physical design optimisations of discrete GPU chips can be more aggressive, unlike monolithic APUs having to provision for the co-existence of high-frequency CPU blocks.

    Also it is not like the architecture will look drastically different at the level of colourful white paper diagrams after just, eh, 1 year since the big revolutionary crop. Heck, these diagrams have barely changed structurally since Tahiti.
     
    #138 pTmdfx, Aug 24, 2020
    Last edited: Aug 24, 2020
  19. Malo

    Malo Yak Mechanicum
    Legend Veteran Subscriber

    Joined:
    Feb 9, 2002
    Messages:
    7,907
    Likes Received:
    4,109
    Location:
    Pennsylvania
    No one else here does. It brings nothing to the discussion.
     
    Kej, Geeforcer, sonen and 15 others like this.
  20. CarstenS

    Legend Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    5,165
    Likes Received:
    2,685
    Location:
    Germany
    Max throughput.
     
Loading...
Thread Status:
Not open for further replies.

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...