Radeon Series RX 6000 Raytracing numbers posted by AMD, match RTX 3070

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Oct 29, 2020.

  1. Jawnys

    Jawnys Master Guru

    Messages:
    225
    Likes Received:
    55
    GPU:
    asus tuf oc 3090
    im gonna wait for official reviews, but i think ill get the 6900xt, ray tracing performance shouldnt be bad considering, that all title will be optimized for this gpu architecture because of the new consoles, i just bought a 1440p 240 hertz, i care more about high fps than raytracing, i mostly play fps game as well so i would turn off the ray tracings anyway, if 6800xt and 6900xt turn out to be like amd benchmark its a big win for me, the monitor and the 6900xt will cost me what the 3090 would have
     
  2. Kaarme

    Kaarme Ancient Guru

    Messages:
    3,513
    Likes Received:
    2,355
    GPU:
    Nvidia 4070 FE
    Yeah. Someone else said it already in another thread I've forgotten, but PS4 and XBox One already have been sporting AMD CPU+GPU, yet on PC games nearly always run better on Nvidia cards. It seems to be some kind of mantra to claim AMD video cards will benefit from the console ecosystem, but somehow that's never the case in practice. Nvidia has a huge machinery running to help game studios to optimise the games for Nvidia video cards, and it shows. If a game runs (relatively) better on an AMD video card, you usually find AMD's direct involvement in the background, not merely the status of being a console port.
     
  3. Jawnys

    Jawnys Master Guru

    Messages:
    225
    Likes Received:
    55
    GPU:
    asus tuf oc 3090
    but this console gen are basicly PCs, if im not mistaken game are even developped on pc and ported to these consoles, it might be different this time around. time will tell us, if i go with an amd gpu it ll be a first for me since the all in wonder 9700pro
     
  4. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    I have unpleasant feeling that while AMD claims huge effective bandwidth thanks to IC, actual story with high resolution textures will be different.
    Maximum theoretical bandwidth of those 16GB memories is 512GB/s w/o OC. That's together Read+Write operations as GDDR6 can't do both at same time through same line.
    That leaves us with usable ~2GB_data/frame if we want to reach 240fps. And that's not much. We will see effect of IC, but I think it will not do much for high fps targets while it will do a lot for bottom line.
     

  5. Jawnys

    Jawnys Master Guru

    Messages:
    225
    Likes Received:
    55
    GPU:
    asus tuf oc 3090
    yep anyway, there is nothing for us to do but wait for reviews, its not like we can buy a 3080 or 3090 anyway :) these are practicly non existent in canada
     
    Fox2232 likes this.
  6. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    I wouldn't worry too much, if you look at the rtx 3070 and its paltry 448gb/s vs the 2080ti's 616gb/s or the 3080 , the extra bandwidth doesn't seem to matter as much when it comes to typical gfx workloads (compute is a different story however),

    I suspect that 512gb/s more than enough, and that the cache mainly decreases the access latency, which keeps the cus better fed, instead of stalling like what older cards suffered from,
     
  7. pharma

    pharma Ancient Guru

    Messages:
    2,485
    Likes Received:
    1,180
    GPU:
    Asus Strix GTX 1080
    Minecraft and Quake 2 via RTX ray-tracing. Anything else?
     
  8. beedoo

    beedoo Member Guru

    Messages:
    149
    Likes Received:
    126
    GPU:
    6900XT Liquid Devil
    Definitely not wanting to poo-poo ray tracing - but looking at videos on YT, comparing RTX on and off - it's a real credit to the designers and artists just how good looking they've made a lot of these games (without it).
     
  9. Gomez Addams

    Gomez Addams Master Guru

    Messages:
    255
    Likes Received:
    164
    GPU:
    RTX 3090
    Yes, I did. Did you? It's difficult to determine that because your post has virtually nothing to do with mine. I will repeat myself : THAT WAS NOT A BENCHMARK! It is just a sample app and AMD's note states exactly that. I am not second guessing them or anyone else because there is no guesswork involved what so ever. This is a fact. Once again, you can download and compile that program for yourself to see what's going on if you want to. I did.

    This sample app really demonstrates very little other than the basic power of the GPU. It has no representation of the performance one will see in games for several reasons. First, as its title states, it is procedural geometry. That means the shapes you see there are computed on the CPU or by the GPU itself - they are not models. Games do very little of that. Second, the number of polygons being rendered is very, very small - in the six digits at most. Games today constantly render millions of polygons per frame. Third, there are zero assets involved in this sample app. None! That means there are no textures of models to manage. What that means is the video memory of the card is not touched for anything other than rendering and the frame buffer.

    If you want to hang your hat on these results from a sample application be my guest. Did you notice the screen shot was rendered on a 3070?
     
  10. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    RT is nothing but buzz word and pr still it way to demanding unless some simple RT test that is only purpose to to show RT
     

  11. pharma

    pharma Ancient Guru

    Messages:
    2,485
    Likes Received:
    1,180
    GPU:
    Asus Strix GTX 1080
    What is being benchmarked is the RT acceleration mechanism used by the GPU. Synthetic benchmarks like this are not representative of anticipated game performance but can be used to draw comparisons between GPU's, ie 3DMark.

    AMD chose to use this app to measure ray tracing progress by showing comparisons between the "new" HW based accelerator on the RX 6000's and the DXR software fallback layer.
    CapFrameX ran his 2080 Ti using default settings and matched the score posted by AMD. He mentioned changing default window size altered the results. The only issue here is it's not known exactly what settings were used by AMD (resolution, etc..), so for all intents and purposes AMD is only showing that their RT accelerator is functional in this application and the resulting benchmark numbers.

    AMD chose this app to show the performance differential made by their HW based RT accelerator and I trust their judgement in choosing how to do that. Like most launches the devil is in the detail, so I can't wait for the upcoming third-party reviews but have a feeling we will see this benchmark again.
     
  12. dirthurts

    dirthurts Guest

    Messages:
    765
    Likes Received:
    13
    GPU:
    Vega 56 Pulse
    These are rather old results on a benchmark with no settings being listed. We also don't know how raw the AMD drivers/hardware was at this point, or what card it was. We really can't take much from this. What we DO know, however, is AMD managed to get the console APUs at least rendering full path tracing (via Minecraft that they showed a long time ago) an games are actually using light RT in their games, even on the APU. Assuming RT is per CU cluster (I think this is confirmed) we can assume it scales well above that. Toss in mature drivers, optimizations bled over from console support (being the same hardware and DX12) and even if their hardware is weaker I think they can have an implementation that is at least highly relevant. Who knows how well Nvidias RTX stuff will run once everything is made with consoles (AMD) in mind first.
     
  13. 0blivious

    0blivious Ancient Guru

    Messages:
    3,301
    Likes Received:
    824
    GPU:
    7800 XT / 5700 XT
    It's looking like these are the 2 cards to get but it's hard to say how they stack up.

    Look forward to seeing the Guru3d 6800XT reviews. Other sites look all flashy but Hilbert has gained my trust over the years.
     
  14. rl66

    rl66 Ancient Guru

    Messages:
    3,924
    Likes Received:
    839
    GPU:
    Sapphire RX 6700 XT
    Sony is a too closed system (it even make look Apple as a totaly open ecosystem lol)... If it succeed to Nintendo it is because nearly all their game are exclusive... sadly it is the the case of Sony, even worse in some case you lost things passing from one console to another.
     
  15. Gomez Addams

    Gomez Addams Master Guru

    Messages:
    255
    Likes Received:
    164
    GPU:
    RTX 3090
    No, we can't and that is my point. Firstly, it is NOT a benchmark and is not intended to be. It is a sample app and its only setting is raytracing in HW or SW. The screenshot is of the sample app running on a 3070 so we have only their word that it can even run on AMD hardware.
     

  16. Corrupt^

    Corrupt^ Ancient Guru

    Messages:
    7,270
    Likes Received:
    600
    GPU:
    Geforce RTX 3090 FE
    Same here. When it becomes mainstream and doesn't murder our framerates anymore, I'll be interested.
     
  17. Stormyandcold

    Stormyandcold Ancient Guru

    Messages:
    5,872
    Likes Received:
    446
    GPU:
    RTX3080ti Founders
    PS5 RT, I think some of you missed this video;
     
  18. MonstroMart

    MonstroMart Maha Guru

    Messages:
    1,397
    Likes Received:
    878
    GPU:
    RX 6800 Red Dragon
    Turanis likes this.
  19. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,011
    Likes Received:
    7,352
    GPU:
    GTX 1080ti
    it doesn't use rdna2

    custom rdna with some custom cerny functions.
     
  20. Turanis

    Turanis Guest

    Messages:
    1,779
    Likes Received:
    489
    GPU:
    Gigabyte RX500
    Sorry,I was wrong about DX12U&DXR for PS5(like XBOX have it),but Playstation 5 have everything RDNA 2 can do.
    PS5 could have a custom Vulkan API.

    [​IMG]

    [​IMG]
     

Share This Page