Article: BaseMark GPUScore Relic Of Life benchmarks with 22 GPUs

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Mar 8, 2022.

  1. Stormyandcold

    Stormyandcold Ancient Guru

    Messages:
    5,780
    Likes Received:
    391
    GPU:
    RTX3080ti Founders
    Looks like a decent RT benchmark. However, I feel like the results just confirm what we already knew; that the top AMD cards have RTX2080ti-like RT performance.
     
  2. kanenas

    kanenas Master Guru

    Messages:
    388
    Likes Received:
    293
    GPU:
    rtx 2070 ,5700xt
    almost as powerful as a 6800xt Screenshot 2022-03-09 120549.png Screenshot 2022-03-09 120638.png
     
  3. AlmondMan

    AlmondMan Master Guru

    Messages:
    918
    Likes Received:
    277
    GPU:
    5700 XT Red Devil
    "So here we see the GeForce RTX 3080 is faster on the Vulkan API, at the same workload but only marginally."

    But the graph shows that Vulkan is slower?
     
    Makaveli and Stormyandcold like this.
  4. Fierce Guppy

    Fierce Guppy Member Guru

    Messages:
    105
    Likes Received:
    27
    GPU:
    2 x GTX 980 / 4GB
    Gigabyte RTX 3080 Gaming OC
    Res: 2560x1440

    API: Vulcan 1.2
    Linux (Manjaro) 5643
    Windows 11 5751

    API: DirectX 12
    Windows 11 5796
     
    Last edited: Mar 9, 2022

  5. schmidtbag

    schmidtbag Ancient Guru

    Messages:
    6,564
    Likes Received:
    2,911
    GPU:
    HIS R9 290
    Why is anyone complaining about this when:
    A. This is a synthetic benchmark, and those don't tend to represent realistic workloads
    B. It's pushing the latest technology. It's not exactly news that AMD lags when it comes to RT performance. Anyone who bought such a GPU is assumed to not care that much about RT, in which case, this benchmark isn't relevant to you.
     
    Airbud likes this.
  6. Astyanax

    Astyanax Ancient Guru

    Messages:
    13,438
    Likes Received:
    5,325
    GPU:
    GTX 1080ti
    its relevant when they want to post claims of favouritism.
     
  7. vMax1965

    vMax1965 Member

    Messages:
    47
    Likes Received:
    10
    GPU:
    MSI RTX 2080 Gaming
    Nvidia bet big on RT and risked quiet a bit but it is starting to pay off especially in conjunction with DLSS. RT is showing up more and more and will be mainstream now that the consoles have added RT and are promoting it even though it is pretty much RT lite! I am sure AMD will get on top of it with the next gen but for now Nvidia has the edge with AMD competing more than well enough on the rasterization side.
     
    Venix and GoldenTiger like this.
  8. Venix

    Venix Ancient Guru

    Messages:
    2,464
    Likes Received:
    1,207
    GPU:
    Palit 1060 6gb
    Actually I will disagree . While the architectural design started long before 2020 when the 2xxx series begun , Nvidia pushed prices high and introduced dlss and rtx in a generation where the rasterisation performance crown was not ever in danger being contested. So they had the feature advantage and still do, but at the time there was not even an alternative for top raster performance , you thought rtx is not for you maybe gimmick or what ever it did not matter at all if you still wanted one of the top 3 raster cards Nvidia was the only option there too. So yes they took a bet but dear lord when they lunched the 2xxx series was THE ideal time for such thing :p.
     
  9. suty455

    suty455 Master Guru

    Messages:
    525
    Likes Received:
    222
    GPU:
    Nvidia 3090
    The fact is Nvidias implementation of RT and DLSS is propriety to them, its not something a number of producers can just dip into and provide a comparison, its hardware that needs the software to be written in a particular way to work and bring out its best for which they charge a nice licence amount, AMD has the same issue its approach is unique but the main difference is they give away the tech FOC much like Gsynch and freesynch.
    For this reason the benchmark is flawed its focusing too much on one competitors implementation not on a universal test that all competitors use, RT in games is pointless right now even with my 3090, it just doesn't make a massive difference to my experience and running my 6900xt and the 3090 side by side same monitor etc they look identical, honestly am hard pressed to see what RT brings to the game other than a much larger power draw, in this situation both systems run the game within a few frames of each other especially as the resolution ramps, benchmarks like this provide a nice chart where one producer sits at the top but it doesn't offer a fair comparison in fact it just confirms what we already know Nvidias RT cores are better then AMDs at the moment, raster performance however is another matter, try a raster comparison and the table will look a lot different.
     
  10. Krizby

    Krizby Maha Guru

    Messages:
    1,395
    Likes Received:
    466
    GPU:
    3090 Watercooled
    What RT games have you played? Godfall? RE Village?
     

  11. Kaleid

    Kaleid Ancient Guru

    Messages:
    2,693
    Likes Received:
    282
    GPU:
    Gigabyte 3070 OC
    Don't know why G3D has so high scores for 3070, I only get about 3400
     
  12. AlmondMan

    AlmondMan Master Guru

    Messages:
    918
    Likes Received:
    277
    GPU:
    5700 XT Red Devil
    They always seem to be a bit higher. I figure because of super optimized testbench setups where most of us are running it on our daily drivers. I get 630ish on my cpu-z singlecore on my 5800x for instance... the reviews have it at 660.
     
  13. Toss3

    Toss3 Member Guru

    Messages:
    199
    Likes Received:
    17
    GPU:
    WC Inno3D GTX 980 TI
    My scores for my RTX 3090 were a bit higher than Guru3D's: 7258 for DX12 and 6936 for Vulkan => Windows 11 Enterprise.
     
  14. Remekra

    Remekra Active Member

    Messages:
    87
    Likes Received:
    7
    GPU:
    AMD Radeon RX6900XT
    I'm just waiting for Radeon 7xxx series to launch and be good at Raytracing to see all the fanboys changing their mind that RT is good now. You could already see it before 6xxx launched, when there was no info on RT performance, only after it turned out that AMD is not that great at it was brushed off as unwanted and unneeded feature again same as when RTX2xxx launched.
    I've had 6900XT, it was a really performant card in raster but RT was not good enough.

    AMD is well aware of it and all the games that use RT and are sponsored by AMD like FC6, DIRT 5 and RE: Village use RT in a light way with for example half-resolution reflections and no advanced features like GI.
    It's not matter of benchmark preferring one manufacter to the other. RT cores in RTX cards do full BVH traversal calculation and triangle/bounding box intersection while AMD solution does just the intersection part.
    I think their solution to RT in RDNA3 will be a full blown solution that is as good as nvidia or better at it but for now, they have it and can do it, but it's not on par on nvidia solution and every benchmark that uses RT will show it.
     
    Krizby likes this.
  15. Luc

    Luc Active Member

    Messages:
    94
    Likes Received:
    57
    GPU:
    RX 480 | Gt 710
    It's only me or no one think this benchmark looks really bad? Talking about animation, design, geometry and textures, of course.

    Combining the reflection with shadows, global ilumination and volumetric lightning... multipliying them in a mirror doesn't sound smart at all, only taxating, maybe is what they intend, but it sounds silly to me, because they failed with really simplistic models and everything they put togheter.

    Maybe I didn't get the point, but I think they failed in their promise of delivering "high-end gamelike 3D content".

     

  16. suty455

    suty455 Master Guru

    Messages:
    525
    Likes Received:
    222
    GPU:
    Nvidia 3090
    Cyberpunk, Doom a few others on a 4k and a ultrawide monitor just dont see a massive change AMD to Nvidia unless of course RT and DLSS is on even then its a small amount of candy and i mean very small if at all noticeable whilst playing
     
  17. Odarnok

    Odarnok Active Member

    Messages:
    62
    Likes Received:
    25
    GPU:
    32 GB DDR3 2133 MHz
    Your statement is false.
    The code of this benchmark (or another) is the same for all vendors.
    Therefore the code must be hardware agnostic.

    The 3DMark DX12 RT benchmark (Port Royal) show a similar difference, and the games that uses diverse RT caracteristics, too.

    You know the real fact because your claims about traditional rasterization, for a minnimal difference that favors AMD.
     
  18. Odarnok

    Odarnok Active Member

    Messages:
    62
    Likes Received:
    25
    GPU:
    32 GB DDR3 2133 MHz
    Not true, you can see this site:

    https://gamegpu.com/

    https://gamegpu.com/action-/-fps-/-tps/cyberpunk-2077-v-1-5-test-gpu-cpu

    RT test without DLSS or FSR (1080 and 1440)

    upload_2022-3-10_17-34-11.png

    Same at 2160

    upload_2022-3-10_17-35-50.png

    2160p and RT with DLSS or FSR - only playable with 3080, 3080ti and 3090

    upload_2022-3-10_17-37-5.png
     
  19. Makaveli

    Makaveli Member

    Messages:
    15
    Likes Received:
    7
    GPU:
    RX 6800 XT
    This!!

    There is numerous text in this short review saying Vulkan is faster yet all the graphs show DX12 being faster????

    I do 674 Single core with PBO on and single core clocks hitting 5.05Ghz
     
    Last edited: Mar 11, 2022
  20. Mineria

    Mineria Ancient Guru

    Messages:
    5,409
    Likes Received:
    644
    GPU:
    Asus RTX 3080 Ti
    As shown in the results above, just RT alone already makes a huge difference.

    DLSS is just an added bonus to get more fps at the cost of some details, almost the same as FSR with the exception that FSR doesn't add something as close to AA without enabling any specific AA.
    Both have their ups and downs too, although the issues are mainly caused by how different game engines are developed and updated, which unfortunately does make FSR look much worse compared to DLSS when it comes to certain game implementations.
    If LOD etc. was done properly in those games it wouldn't be as bad at all.
     

Share This Page