Quick test: Wolfenstein II: The New Colossus Adaptive Shading Benchmarks

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Nov 20, 2018.

  1. Darren Hodgson

    Darren Hodgson Ancient Guru

    Messages:
    15,996
    Likes Received:
    511
    GPU:
    EVGA GTX 1080 Ti SC2
    Seems a bit pointless to me this as the performance increase is so minimal anyway that you will almost certainly not notice any difference while playing. The RTX 2080 (and presumably RTX 2080 Ti) already run this game at ridiculously high framerates that a few more FPS at 1440p and 4K is really not that worthwhile.

    This technology might be more useful on lower end Turing cards such as an RTX 2050/2060 which I assume will be out at some point.

    Not really been impressed with the RTX range so far. Sure, the performance increase is nice in the RTX 2080 Ti vs. my own GTX 1080 Ti but that is par for the course with any new range of GPUs and that card isn't really offering a huge leap over what we've had before, just the usual 25-40% increase on average vs. the previous generation's equivalent GPU. The RTX features so far have been very underwhelming IMO and definitely not worth the price hike in my view.

    Like any new technology though, the first generation is always the proof of concept that doesn't quite run as expected but I'm sure that future hardware will see significant improvements and also lower prices. I am more than happy to wait as it's not like there is any game out there that demands anything more powerful than my current GTX 1080 Ti and those RTX enhancements are so sparse that there is absolutely no reason to upgrade.
     
    Robbo9999 likes this.
  2. Ratatosk

    Ratatosk New Member

    Messages:
    1
    Likes Received:
    0
    GPU:
    RX580 Nitro L.E.
    You should really remove those misleading graphs.

    Conceptually interesting tech, but pretty useless, at least on this game.
     
  3. StewieTech

    StewieTech Chuck Norris

    Messages:
    2,537
    Likes Received:
    903
    GPU:
    MSI gtx 960 Gaming
    Thanks for the article boss. To me these are way to many settings, im not a geek enough. Why not just fire a game and enjoy it?
     
    fantaskarsef likes this.
  4. Moonbogg

    Moonbogg Member Guru

    Messages:
    134
    Likes Received:
    37
    GPU:
    GTX1080Ti@2.1GHz
    Getting that extra 9fps when you already have over 140fps is easily worth spending the $1300+ on an RTX 2080ti, and don't you dare try to say it isn't because you know darn well that it is. Stop being a bunch of picky fools and go out there and get you a piece of that next generation technology. Go get it. You go do it.
     

  5. Noisiv

    Noisiv Ancient Guru

    Messages:
    7,232
    Likes Received:
    814
    GPU:
    2070 Super
    Anyone who can be misled by removal of empty space in x/y graph shouldn't be reading graphs.

    There is nothing, absolutely nothing of interest happening at 0fps.
    Everything that's needed to analyze the performance uplift and get a complete mental picture of the Adaptive Shading benefits and that of the total fps performance is right there on the graph.
    The only way one can be misled by Hilbert's graph is if he's too smart for his own good by associating point height with the total performance, yet stupid enough to miss the clearly labeled ordinate values (142,144,146...). Therefore he shouldn't be reading graphs to begin with.

    Truncation of line graph in order to get a better and immediate insight of data trend is a standard method in data visualization.
    Notice I said line graph.
    OTOH truncating ordinate on a bar graph is a bad practice and can be considered misleading.
    Because unlike a point in 2d space of line graph, an area (or a volume in 3d charts) with its clearly visible borders/demarcation lines suggests totality of the value it represents.

    --------------------------------------------------------------------------------------------------------------------------

    Example 1:

    [​IMG]
    • This is no-no. Because it is highly suggestive of tripling of the observed value in 2013 compared to 2011. Notice that the bar areas do not accurately represent their respective ordinate values.
    --------------------------------------------------------------------------------------------------------------------------

    Example 2:

    [​IMG]
    • This is perfectly fine.
    • Straight lines would be a better choice in order to emphasize the lack of data between the observed points.
    • Use smooth line (like HH here) to emphasize the trend.
    --------------------------------------------------------------------------------------------------------------------------

    Example 3:

    [​IMG]
    • This is fine as well
    • Use this if your targeted audience consists of half-blind, illiterate, AMD lunatics easily offended by Turing awesomeness, or some such mix :)
    • No seriously - use this if you're dealing with uneducated crowd(!)
    • Notice that unlike in the Example 2, the information about the inflection point lying between Quality and Balanced (visualizing perf. de-acceleration and subsequent acceleration) is lost on a quick graph inspection.
    • OTOH what's lost on locality is gained on the immediate sense of the total scale.
     
    Last edited: Nov 21, 2018
    yasamoka, alanm, Aura89 and 3 others like this.
  6. matox

    matox Member

    Messages:
    29
    Likes Received:
    4
    GPU:
    8870M - 2GB GDDR5
    Don't listen to these idiots HH. The purpose of a graph is to illustrate the difference. Maximum you could do was a %delta between the two.
    If people can't spend an extra second looking at the axis they should not be looking at graphs in the first place.

    Go HH
     
    yasamoka likes this.
  7. BlackZero

    BlackZero Ancient Guru

    Messages:
    8,879
    Likes Received:
    478
    GPU:
    RX Vega
    Some people may have difficulties with numbers, so the full graph illustration may help them somewhat better understand the graph using assumed magnitudes.

    Certainly not suggesting anyone here does, just raising a point.

    p.s.

    On the usefulness of this new tech, if you had 10 similar threads highlighting relatively small improvements, you'd find the same people criticising the improvement in every one of them. o_O
     
    Last edited: Nov 21, 2018
  8. Rx4speed

    Rx4speed Active Member

    Messages:
    99
    Likes Received:
    36
    GPU:
    R9 295x2
    LMAO. I thought "Eight Thoughts" from Youtube made this comment.
     
  9. Robbo9999

    Robbo9999 Maha Guru

    Messages:
    1,499
    Likes Received:
    262
    GPU:
    GTX1070 @2050Mhz
    I disagree mostly, I think the graph that starts at 0 fps shows the most information - ie that there is just not much performance difference with this added feature.
     
  10. Rich_Guy

    Rich_Guy Ancient Guru

    Messages:
    12,578
    Likes Received:
    600
    GPU:
    MSI 2070S X-Trio

    Thats bloody great! :D
     

  11. schmidtbag

    schmidtbag Ancient Guru

    Messages:
    5,419
    Likes Received:
    1,965
    GPU:
    HIS R9 290
    I'm sure it also depends on the hardware. The GPU might be powerful enough to not really be struggling with shading in the first place.
     
  12. Denial

    Denial Ancient Guru

    Messages:
    13,111
    Likes Received:
    2,576
    GPU:
    EVGA 1080Ti
    It's not a bad perf uplift considering Nvidia is pushing it into the DX/Vulkan spec (currently support only via NVAPI). On lower end/mid-range cards it might bring a bigger performance delta.
     
    pharma likes this.
  13. schmidtbag

    schmidtbag Ancient Guru

    Messages:
    5,419
    Likes Received:
    1,965
    GPU:
    HIS R9 290
    DX12 and Vulkan have very little to do with improving GPU performance; the main performance enhancement is from reducing CPU and PCIe overhead. So, if the GPU is already bottlenecked, DX12 and Vulkan aren't going to offer much of a performance gain vs DX11 or OGL4.x.

    In other words, the a lower-end GPU could potentially see a greater performance difference, but only because of adaptive shading, not because of DX12/Vulkan.

    All that being said, DX12/Vulkan will see greater performance difference (over DX11/OGL4.x) if you swap out a worse CPU.
     
  14. Denial

    Denial Ancient Guru

    Messages:
    13,111
    Likes Received:
    2,576
    GPU:
    EVGA 1080Ti
    I think you misunderstood what I was saying - content adaptive shading is currently implemented via NVAPI.. but Nvidia is working with Microsoft to get it implemented directly into the DX library which will make it easier for developers to support and open it up to AMD.
     
  15. fry178

    fry178 Ancient Guru

    Messages:
    1,599
    Likes Received:
    234
    GPU:
    2080S WaterForceWB
    @bobmitch
    did u install RTSS as well (iirc its selected by default)?
    gets detected by some games as if your trying to cheat.

    @Bender82
    do you also whine about the price/insurance/maintenance cost when a new porsche/lambo etc gets released?
    last time i checked no one has a gun to their head forcing you to buy anything.
    guess your still catching up on the whole how-capitalism-works thing.
     

  16. Robbo9999

    Robbo9999 Maha Guru

    Messages:
    1,499
    Likes Received:
    262
    GPU:
    GTX1070 @2050Mhz
    The RTX lineup from NVidia is hardly worthy of porsche/lambo comparisons! They just happen to be the next gen cards from NVidia that will eventually replace all the Pascal cards, they're still overpriced. Yes, they should cost more than AMD, because they're better in performance & features, but they're milking it too much with the 2080 & 2080ti (2070 is slightly better value). Titan V is your Porsche & Lambo though, but the RTX should be fairly priced.
     
    Last edited: Nov 21, 2018
  17. angelgraves13

    angelgraves13 Ancient Guru

    Messages:
    2,138
    Likes Received:
    592
    GPU:
    RTX 2080 Ti FE
    This is a great new feature that will be very useful in new games that push GPUs to the limit.

    I’m actually more interested in seeing their parallel compute which isn’t enabled yet in drivers. Should make async compute irrelevant. DICE mentioned it is waiting for it to speed up RTX and the performance uplift will be massive.
     
  18. fry178

    fry178 Ancient Guru

    Messages:
    1,599
    Likes Received:
    234
    GPU:
    2080S WaterForceWB
    @Robbo9999
    oh i didnt mean comparing it,
    but that ppl who dont have the money to buy one, complain that they cost a lot to own.

    at least i doubt that 80% of the ppl here that are talking about milking, are complaining to their local car dealer that they cant buy their top of the line +500hp truck/car because its too expensive for them.
    e. g.
    if its not in "your" budget, doesn't mean you should sour it for the ppl that do buy it.
     
    Last edited: Nov 25, 2018
    yasamoka and Aura89 like this.
  19. Robbo9999

    Robbo9999 Maha Guru

    Messages:
    1,499
    Likes Received:
    262
    GPU:
    GTX1070 @2050Mhz
    Well, ok, that's right, you either decide to buy it or you don't, it's a choice like you say, but I still think RTX should be a bit cheaper. I think NVidia should have significantly lowered price of Pascal (which would help shift them) when RTX came out, and RTX should be similar price to launch price of Pascal - that seems fair to me.
     
    fry178 likes this.
  20. Valken

    Valken Ancient Guru

    Messages:
    1,571
    Likes Received:
    126
    GPU:
    Forsa 1060 3GB Temp GPU
    Nice quick review. Can you do this test when the 2060 comes out? I want to see how this affects the 2060/2070 because those GPU need the performance boost MORE than the 2080/Ti GPUs...
     
    Robbo9999 and fry178 like this.

Share This Page