New Upcoming ATI/AMD GPU's Thread: Leaks, Hopes & Aftermarket GPU's

Discussion in 'Videocards - AMD Radeon' started by OnnA, Jul 9, 2016.

  1. Anarion

    Anarion Ancient Guru

    Messages:
    13,599
    Likes Received:
    386
    GPU:
    GeForce RTX 3060 Ti
    GeForce 2 MX and GeForce 4 MX aren't the same. The main issue was that it didn't offer DirectX 8 support but it had many improvements over GeForce 2 architecture. Also MX440 was faster than the to tier GeForce 2 cards.
     
    Last edited: Apr 23, 2017
  2. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,547
    Likes Received:
    608
    GPU:
    6800 XT
    It was faster yes. "Instead, the NV17 would basically be a GeForce2 MX with an improved memory controller, multisample AA unit, and updated video features; another way of looking at it would be the GeForce3 without two pixel pipelines or DirectX 8 compliance." ye it's a tad more then rebrand but not much.

    Having way higher clocks helped for sure. But anyway not that straight rebrand but same arch none the less. With some tweaks.

    http://www.anandtech.com/show/875/6
     
  3. Undying

    Undying Ancient Guru

    Messages:
    25,343
    Likes Received:
    12,754
    GPU:
    XFX RX6800XT 16GB
  4. Truder

    Truder Ancient Guru

    Messages:
    2,392
    Likes Received:
    1,426
    GPU:
    RX 6700XT Nitro+
    No no you misunderstood, I was trying to explain albeit badly that the 470 is the successor to the 380X with AMD taking the Tonga design and improving it for efficiency making GCN1.3 or GCN4 whichever way you want to call it.
     

  5. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
    New is new, so:

    5xxx
    6xxx
    Thaiti - 7xxx aka GCN 1.0
    Hawaii - 2xx aka GCN 1.1
    Tonga - aka GCN 1.2
    Fiji - Aka GCN 1.3 with add. Features + NextGen HBM
    and GCN 4.0 (1.4) aka Polaris build for FinFet 14nm + add. Features
    and VEGA as GCN 5.0 (1.5) it packs so many Features, so we need whole article to cover it up :nerd: -> Performance wise: HBM Cache, Color Compresion v.5 ? , Primitive Discard Accelerator v.2 etc.

    Then you have same for nV (but we need some nV knowing guy to tell us ;) )

    But AFAIK:
    480 and 580 was almost the same chip
    Then 680 comes, then Upgraded to 780
    Then Maxwell 9xx (it has major change: Tile Based Renderer) and Die shrink -> Pascal 1xxx
    NextGen will be Volta with GDDR5X then HBM_2.... correct me if im wrong

    But in nV you always get also Titan based Big CHips: Yes 1080Ti can easily swallow ~300tW :)
    So you got 980 then Titan and 980Ti as cut down ver. of Titan (similar to Intels HEDT at 1000$ bracket )

    Lets us hope Vega will get 220tW at 1400-1500MHz... (but frankly it's not that Era Yet ;) so propably we can have ~250-270tW)
     
    Last edited: Apr 23, 2017
  6. Anarion

    Anarion Ancient Guru

    Messages:
    13,599
    Likes Received:
    386
    GPU:
    GeForce RTX 3060 Ti
    Basically equal situation to minor GCN architecture update.
     
  7. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    I hope that our hearts have settled on the fact that Vega probably won't be faster than the Ti. Not initially at least. I have no problem with that, if the price is fair.
     
  8. warlord

    warlord Guest

    Messages:
    2,760
    Likes Received:
    927
    GPU:
    Null
    It will be neither faster than refreshed "1080"+.
     
  9. Tuga

    Tuga Guest

    Messages:
    48
    Likes Received:
    0
    GPU:
    nvidia 670 gtx
    Didn't some vega card beat 1080 on Doom @4k? See that's the thing, we don't know whether that was "top end" card, or the "second" card.
     
  10. Chastity

    Chastity Ancient Guru

    Messages:
    3,738
    Likes Received:
    1,662
    GPU:
    Nitro 5700XT/6800M
    Why don't we put speculation aside and wait for some proper hardware reviews? :)
     

  11. Loophole35

    Loophole35 Guest

    Messages:
    9,797
    Likes Received:
    1,161
    GPU:
    EVGA 1080ti SC
    Might want to read the title of the thread.
     
  12. Loophole35

    Loophole35 Guest

    Messages:
    9,797
    Likes Received:
    1,161
    GPU:
    EVGA 1080ti SC
    580 was basically the same GPU as 480 but without the problems.
    680 became the 770. Besides that there were actually no rebrands in the 7xx cards even the 770 got an updated reference PCB and GPU boost 2.0. Maxwell was supposed to be on 20nm IIRC but that was a cluster so it had to be respun as 28nm. That is why Maxwell cards carry a 2 not a 1 as the first number.

    The reason the Titan cards exist is because there has been such slow responses from AMD. Volta is actually 2 years late and Pascal was never supposed to happen but here we are.


    To get the Tflops that AMD have claimed Vega at 4096CU needs 1,475MHz as they claim 12.5 Tflop SP32. Respun Polaris shows us AMD can make silicon go that fast but it's gonna be hungry and hot unless they figured something out.


    Edit: sorry for the double post
     
  13. user1

    user1 Ancient Guru

    Messages:
    2,748
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    I think it would be a good thing to clarify what die shrink means,

    An example of a classic silicon die shrink is the G92 v G92B
    the design is the same but the g92 is on 65nm vs 55nm on the g92b

    There are also are revisions and/or die shrinks such as the g80 - g92
    g80 has an older video encode/decode hw, pcie 1.1 and is on 90nm , but the core design is pretty much identical.

    I would put the differences between the fiji and tonga uarchs in this league of type of revision


    then there are what I would call major revisions, i would consider the gt200 core to be a major revision of the tesla uarch of which the g80-g92b are part of.
    I would put tonga - polaris here as would I put hawaii - fiji.


    And finally what I would call a uarch Rework, which is where I would assume to put vega(though it remains to be seen exactly how extreme the differences are)

    An equivalent nvidia example of a uarch rework would be tesla to fermi in my opinion

    This of course is a Simplification, this is not the complete picture but it should give a rough idea of what how to categorize these types of differences

    Gotta thank nvidia for making so many g80 revsions makes finding examples easy.
     
  14. flow

    flow Maha Guru

    Messages:
    1,023
    Likes Received:
    17
    GPU:
    Asus TUF RTX3080 oc
    In the end of the day consumers want faster cards. If the upcoming RX580 equals the RX480 which in turn is about the same as the 390, which is not much faster than the 290, then there is little improvement on the gaming front.
    Most users wont care about power usage and die shrinks.

    So sure, lets wait and see reviews when those cards are available. But if they aren't much faster, then these rebrands aren't improving things either for us, the users.
    All in all I don't have a good feeling about this.
     
  15. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,547
    Likes Received:
    608
    GPU:
    6800 XT
    That would be insanely weak. To be honest. Fury X is not that far from 1080. If the 1.5ghz clocks and all of the improvements only accounted to 1080+ hahaha. Just scrap the gpu side.
     

  16. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
    Yes it seems as an logical assumption :formal:

    1, 2 ,3 :nerd:

    For 1080Ti/TitanXp levels we need 1450-1550MHz on Big Vega
    I still hope for some hudge step forward (look at ZEN ;) )
    But it's more ATI than AMD so lets Hope they can deliver some VEGA :heh:
    YES ATI is more than capable (we got History to look at)
     
    Last edited: Apr 24, 2017
  17. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
    Yup

    [​IMG]

    -> https://developer.nvidia.com/sites/...orks/images/lifeofatriangle/fermipipeline.png
     
  18. Denial

    Denial Ancient Guru

    Messages:
    14,206
    Likes Received:
    4,118
    GPU:
    EVGA RTX 3080
    Idk, I think Volta simply got split into two architectures for whatever reason. It's main feature was stacked DRAM and that's done with Pascal. I think Nvidia bit off more than it wanted to chew with what it originally had planned for Volta, so they took the die shrink/hbm/nvlink and split it out. The rest of the changes are more likely architectural stuff that they didn't want to implement with the die shrink.

    You also have to remember that Nvidia isn't only competing with AMD with it's GPUs - on the compute side you have Intel and even companies like Google who are developing their own Deep Learning devices (although Google's is very workload specific). AMD is facing similar competition.

    I think Vega is going to be pretty competitive - trading blows with the Ti in DX12 titles. AMD shouldn't have a problem competing with Pascal directly - the problem is going to be the lost sales between now and when it actually ships. I'm really surprised they haven't been showing off more technical stuff about Vega to disrupt some Nvidia sales.
     
  19. Picolete

    Picolete Master Guru

    Messages:
    494
    Likes Received:
    261
    GPU:
    Sapphire Pulse 6800
    This is one of the reasons why i haven't upgraded my graphics card
     
  20. warlord

    warlord Guest

    Messages:
    2,760
    Likes Received:
    927
    GPU:
    Null
    The problem with AMD these days surely is we have not great breakthroughs in performance department. 290/x, 390/x, 480, 580, almost near eachother.

    The days since hd 5XXX are long gone. It was soemthing new. It resulted for failing tessellation performance then, because it was a purely new architecture. After HD 5000 series, red is like trolling us, it took them years to improve tessellation and again not by much they still need until 2017 to control it via ccc :bang: But they are still best bang for buck. That's why many of us love ATI. :)

    I hope I'll get a true replacement for my 390X next year finally. :bang: :banana:
     

Share This Page