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

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

  1. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    They do much more than most, but they don't have it. It's such a simple statistical tool, I don't get why nobody has it.
     
  2. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,230
    Likes Received:
    1,494
    GPU:
    2070 Super
    Right... there's no numerical frame variance analysis, but you can make judgment based on frametime graph.

    variance +- avg(variance) gives you quick glance of frametime distribution, but carries less information and is less descriptive than educated look at framtime graph.


    otoh according to TR, 99% percentile analysis is less kind to Zen, than avg. fps.
     
  3. Dekaohtoura

    Dekaohtoura Master Guru

    Messages:
    382
    Likes Received:
    56
    GPU:
    XFX 6700XT SWFT
    ?

    Different TDP, much lower base freq to begin with.

    The real question is how well it will oc compared to it's bigger brothers.

    If it can easily reach >3.7GHz, then it's a much better choice.
     
  4. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
    Vega is designed to be able to easily access memory outside of their frame buffer, effectively allowing the GPU to use DRAM and even system memory to create a larger frame buffer.

    In a VRAM overspend scenario, Vega is designed to keep all of the most important info in the GPU's onboard HBM memory, keeping other data in external storage, effectively allowing the GPU to use larger datasets with minimal performance degradation.

    We speak about this and several other Vega features in out Vega GPU architectural analysis below.

    Vega essentially has a very smart/innovative solution for using external storage as a frame buffer with a minimal performance loss. This feature is most useful for compute applications, though it will be very useful for games that require lots of VRAM.

    Imagine a user of an older GPU, perhaps a 3GB or 2GB model, who frequently runs out of VRAM in some modern games but have plenty of GPU performance. This will allow those users to get past certain traditional VRAM limitations, which will be handy for gamers moving forward.

    More here -> https://www. overclock3d net/news/gpu...ral_analysis/1

    [​IMG]
     
    Last edited: Mar 12, 2017

  5. Denial

    Denial Ancient Guru

    Messages:
    14,206
    Likes Received:
    4,118
    GPU:
    EVGA RTX 3080
    Eh, it doubles the minimums if you run out of VRAM which is nice - but it doesn't completely eliminate the need for having enough VRAM in the first place. It's definitely good for compute and obviously the workstation uses they've shown off (video editing and stuff) it's awesome for. But if AMD gives people the option between a 4GB HBM2 card and a 8GB HBM2 card - I would definitely still recommend the 8GB card.

    I think the current rumor is 8GB for big Vega right? That seems like enough for 4K for the next 4 years so. I doubt Scorpio is going to have more than 8GB so that will be the "standard".
     
  6. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    Scorpio will be 12GB. The double minimum framerates might also make it almost imperceptible when the GPU fetches, so it could have less stutter by "default", even for titles where the onboard VRAM is more than enough. Don't forget they also promised something like 50% VRAM usage reduction, which in itself is a bit crazy. Let's wait and see.
     
  7. Hussnash

    Hussnash Guest

    Messages:
    20
    Likes Received:
    0
    GPU:
    GTX 1080

    how is this any different from unified memory and DMA from the GPU ?
    sounds like a lot of marketing hoo-wa for a now standard feature

    https://devblogs.nvidia.com/parallelforall/beyond-gpu-memory-limits-unified-memory-pascal/
     
  8. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    If it doubles the minimums it's not "just marketing". AMD also has a history of utilizing external devices and the bus better. Don't forget that when it's working, Crossfire scales better and it doesn't require a bridge. It also works over x4 links.
     
  9. Hussnash

    Hussnash Guest

    Messages:
    20
    Likes Received:
    0
    GPU:
    GTX 1080
    It doubles the minimums? It's not just marketing? Link to benchmarks please
     
  10. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    They showed it in GDC, and the controller is a large part of their compute push and a large die investment.
     

  11. Mugsy

    Mugsy Guest

    Messages:
    287
    Likes Received:
    15
    GPU:
    RX 5700 XT Pulse
  12. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
  13. DerSchniffles

    DerSchniffles Ancient Guru

    Messages:
    1,665
    Likes Received:
    148
    GPU:
    MSI 3080Ti
    May I ask your opinion on why recording minimum frames is retarded? Most of what you notice when frame rate is fluctuating is the dips to the minimum, so why would you not want to see that? I feel that minimum frame rate is highly underrated for that exact point. It does not matter if you are getting 200 fps if it dips to 40 when the scene being rendered gets intense...am I missing something critical here? I would love to hear what you have to say on this topic.
     
  14. mtrai

    mtrai Maha Guru

    Messages:
    1,183
    Likes Received:
    374
    GPU:
    PowerColor RD Vega
    Yes BIGGER IS ALWAYS BETTER Please pay no attention to the small man behind the curtain. Seriously the min frames are important too.

    Now that I have said that...the reasoning and rational is that if you can get a high max FPS then in theory you min FPS should be just fine.
     
  15. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    Becausee minimum framerates literally mean nothing. You can have a GPU that gives you evenly paced gameplay with extremely small variance (under 1%), which does one loading hitch for 1/3 of a second, and your "minimum" suddenly will be 5fps.

    Then you have a card that has huge frametime variance (more than 20% let's say), gameplay is jerky, yet it won't do the one loading hitch that might be the reason that the original card is able to provide you with such great experience. And it's minimum is 22fps, let's say.

    Do you see why it's worthless?
     

  16. Redemption80

    Redemption80 Guest

    Messages:
    18,491
    Likes Received:
    267
    GPU:
    GALAX 970/ASUS 970
    Is that not comparing a loading hitch to actual consistent low framerates though?

    Minimum is the only thing i look at when comparing.
     
  17. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    "Minimum" and "maximum" are literally the single highest fps and the single lowest FPS. At least average gives you an idea of the whole run. Still worthless, but less so.

    What you mean by "actually consistent low framerates" could only be tested by 95 or 99 percentile numbers along with variance.
     
  18. Redemption80

    Redemption80 Guest

    Messages:
    18,491
    Likes Received:
    267
    GPU:
    GALAX 970/ASUS 970
    I get what they are, and i agree it would be unfair to discard something as poor for a one off drop but you can't just discard the minimums as meaningless.

    I've capped many a game at 30fps even though the average was closer to 60, and years ago I remember capping RE5 at 50 even though the average was much much higher.
     
  19. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    The minimums are no indication of either the actual sustained minimum framerate, neither of the stutter experience in a game. Variance and percentages are that. The minimums and the maximums are merely bursts either to the bottom or to the top, and unless we know the frequency of the bursts compared to where the game actually renders, they mean nothing.
     
  20. Redemption80

    Redemption80 Guest

    Messages:
    18,491
    Likes Received:
    267
    GPU:
    GALAX 970/ASUS 970
    I know, but it's not uncommon to get just min/avg/max.

    It's not a figure you can just discard, well maybe you can but I can't and without more detail the average is is something I would discard before the minimum.
     

Share This Page