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

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

  1. PrEzi

    PrEzi Master Guru

    Messages:
    723
    Likes Received:
    585
    GPU:
    XFX MERC310 7900XTX
    There are a few leaks from Moores law is dead.... He had some good sources talking bout RDNA2 so I assume his info could be something.... but take it with a grain of salt.
    Here it is >
     
    Chess likes this.
  2. ninja750

    ninja750 Master Guru

    Messages:
    366
    Likes Received:
    89
    GPU:
    RX 6800 REF
    at this point I don't think we'll hear news from lisa in january for n22
     
  3. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,564
    Likes Received:
    2,961
    GPU:
    XFX 7900XTX M'310
    https://www.reddit.com/r/Amd/comments/kg8ktw/amd_apus_for_2021_and_2022_rogame_leak/

    https://twitter.com/_rogame/status/1340294195433660422

    https://twitter.com/_rogame/status/1340294511218585601

    Neat.

    Also there is a Zen3+ there. :D
    Vega living on too.

    Course they can't just throw out all the Vega APU work and existing tech like nothing and just shift it all over to Navi even if Navi2 is probably a excellent move from GCN now.

    Interesting little roadmap here though these are the APU's the dedicated desktop models for both Zen and RDNA will look a bit different and what's available and when.


    Curious too what Zen3+ actually gives Zen3 already feels like a solid improvement over Zen2 but if they can take it further before Zen4 and AM5 with DDR5 and what ever else is on the market by 2022 or whenever well that'd be interesting too.

    And of course RDNA3 as well for the GPU side and what's next for NVIDIA if that's Hopper or something else.
    Intel might be ready for some additional competition too on the GPU side and maybe we'll even see CDNA and AMD pushing to the professional software market and these systems but yeah that's going to be tough with especially the CUDA support here.
     
    Last edited: Dec 19, 2020
  4. PrEzi

    PrEzi Master Guru

    Messages:
    723
    Likes Received:
    585
    GPU:
    XFX MERC310 7900XTX
    AMD patents GPU chiplet designs, a future of RDNA architecture?

    A new patent submitted to the US Patent Office on December 31 describing the AMD approach to potential GPU chiplet design. The manufacturer has outlined the problematic structure of such a design and explained how it would be possible to avoid them in the future.

    https://videocardz.com/newz/amd-patents-gpu-chiplet-designs-a-future-of-rdna-architecture


    ++ This >
    AMD's Newly-patented Programmable Execution Unit (PEU) allows Customizable Instructions and Adaptable Computing


    Processor includes one or more reprogrammable execution units which can be programmed to execute different types of customized instructions

    When a processor loads a program, it also loads a bitfile associated with the program which programs the PEU to execute the customized instruction Decode and dispatch unit of the CPU dispatches the specialized instructions to the PEUs
    PEU shares registers with the FP and Int EUs.
    PEU can accelerate Int or FP workloads as well if speedup is desired
    PEU can be virtualized while still using system security features
    Each PEU can be programmed differently from other PEUs in the system
    PEUs can operate on data formats that are not typical FP32/FP64 (e.g. Bfloat16, FP16, Sparse FP16, whatever else they want to come up with) to accelerate machine learning, without needing to wait for new silicon to be made to process those data types.
    PEUs can be reprogrammed on-the-fly (during runtime)
    PEUs can be tuned to maximize performance based on the workload
    PEUs can massively increase IPC by doing more complex work in a single cycle.
     

  5. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,564
    Likes Received:
    2,961
    GPU:
    XFX 7900XTX M'310
    https://www.reddit.com/r/Amd/comments/kp6nh8/amd_navi_2x_xtxh_and_nashira_point_gpus_spotted/

    https://videocardz.com/newz/amd-navi-2x-xtxh-and-nashira-point-gpus-spotted

    Nav21 I think are as follows.

    XL - 6800
    XT - 6800XT
    XTX 6900XT

    And now here's XTXH

    HBM?
    (M as mentioned in the Reddit topic is what they'd use for mobile.)

    Interesting. :)

    EDIT: Suppose this also strengthens the former leaks about there being two Navi 21 chips for the high-end rumored as HBM capable but that's yet to be confirmed what this will be.

    Wonder if that means the rest of the article and the mention of Nashira Point is going to be for RDNA3 whenever that comes out. Likely but it could be something else too.
    (At some point somewhere there were these compute oriented CDNA cards replacing GCN I believe.)
     
    Last edited: Jan 3, 2021
  6. PrEzi

    PrEzi Master Guru

    Messages:
    723
    Likes Received:
    585
    GPU:
    XFX MERC310 7900XTX
    AMD Teases the mid-range cards --- coming in 1H 2021 !

    AMD showcased the upcoming design of its single-fan and dual-fan RDNA2 graphics cards. The SKUs based on Navi 2X GPUs should launch in the first half of 2021, but AMD did not reveal the exact date. The GPUs will also launch for the upcoming gaming laptops.

    The dual-fan design is clearly the Radeon RX 6700 series that we have spoken about for the past 2 months. AMD was expected to unveil the series during the stream, but sadly that was not the case. This SKU will be based on the Navi 22 GPU. Rumors suggest that Radeon RX 6700 series might launch in March.

    The single-fan model is likely based on Navi 23 GPU, the entry-level processor. It is too early to speculate on the naming, but it is very likely that this design will be used by Radeon RX 6500 series.

    https://videocardz.com/newz/amd-teases-mid-range-and-mainstream-radeon-rx-6000-graphics-card-series
     
    Last edited: Jan 13, 2021
    Undying likes this.
  7. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,016
    Likes Received:
    7,353
    GPU:
    GTX 1080ti
    Consumer/gaming gpu's will never go the chiplet route.

    chiplets are for GPGPU designs.
     
  8. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    Never depends on DXR progression. Pure raytracing workloads will not be hindered by even multiple discrete GPUs. Chiplets will be perfect for it.
     
  9. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,016
    Likes Received:
    7,353
    GPU:
    GTX 1080ti
    Definitely not, you want a fixed function units point of contention to be the clockrate, not a bus adding latency into the pipeline.
     
  10. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,564
    Likes Received:
    2,961
    GPU:
    XFX 7900XTX M'310
    If AMD attempts to go this route if it is more cost effective they would likely have to use this infinity cache method to offset latency but considering the size already for the 128 MB on the 6000 GPU series and the cost of having even more it seems like a fairly limited workaround not without it's own issues to retain speed.

    [​IMG]

    (TechPowerup.)

    That's only for cache though unless there's other hardware dedicated to more full ray tracing acceleration instead of partial solutions like de-noising that would be a problem and then that's still a problem for how the workload overall should be processed otherwise you get delays as one waits for the other that would need to be solved as well.

    Navi30 will be a interesting reveal but yeah CDNA and the compute oriented hardware and workstation system might be where AMD will eventually leverage this.
     

  11. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    Pure raytracing can run every single pixel on separate GPU without performance penalty. Only overhead is in telling which pixel should be calculated by given GPU and then time to get result back. In other words, 4 GPUs will run 4 tiles to practically no overhead.

    I think you did overlook word "pure".
     
    JonasBeckman likes this.
  12. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,564
    Likes Received:
    2,961
    GPU:
    XFX 7900XTX M'310
    Interesting, though for consumer GPU's and gaming I would expect the current console generation and hardware to be a big factor and thus a hybrid approach is going to be the standard along with a variety of tweaks and optimizations to manage performance though I am very curious to see ray tracing become a bigger thing other than focusing on updating one or a few specific effects. :)
     
  13. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,016
    Likes Received:
    7,353
    GPU:
    GTX 1080ti
    Reality says no.
     
    itpro likes this.
  14. Chastity

    Chastity Ancient Guru

    Messages:
    3,738
    Likes Received:
    1,662
    GPU:
    Nitro 5700XT/6800M
    AMD has already stated RDNA3 will incorporate a chiplet design.

    What exactly are your credentials again to tell us we are all wrong?
     
  15. itpro

    itpro Maha Guru

    Messages:
    1,364
    Likes Received:
    735
    GPU:
    AMD Testing
    He is somewhat right though. We cannot use given hardware as we like or desire. You all here (including myself) dream that this is a magical breakthrough to double, triple or quadruple the gaming performance. That ain't gonna happen, ever. (because of chiplets anyway, there are many more things to seek)
     

  16. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    As I wrote, pure raytracing can be distributed even across multiple GPUs while they render same frame. And it will not result in loss of efficiency.
    Same goes to CPU raytracing. Only overhead is syncing and telling HW what is on next frame.
    And funny thing, that even nVidia agrees with their old marketing. But as I had to reiterate, key word here is "pure".
    One day we may have that in games. And that's why @Astyanax 's statement which puts emphasis on "never" having chiplets in consumer/gaming space is incorrect.
    (Unless our world comes to end before things get figured out or manufacturing enables chiplets powerful enough to run pure raytracing in real time.)
     
  17. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,016
    Likes Received:
    7,353
    GPU:
    GTX 1080ti
    Last edited: Jan 14, 2021
  18. Astyanax

    Astyanax Ancient Guru

    Messages:
    17,016
    Likes Received:
    7,353
    GPU:
    GTX 1080ti
    "multiple gpu's" is not doing BVH on GPU A and Drawing on GPU B.

    multiple gpu's are doing the work per frame with each card doing the entire work for its assigned frame.
     
  19. Truder

    Truder Ancient Guru

    Messages:
    2,392
    Likes Received:
    1,426
    GPU:
    RX 6700XT Nitro+
    The only thing confirmed about GPU chiplets is a recent patent...

    As for RDNA3, only thing confirmed is that it's going to be on an "advanced node", for now.

    But really, to say consumer GPUs will never have chiplets? That's a very finite statement? AMD engineers are exploring the possibility, I'm sure we can all agree on that? Never say never after all.
     
  20. OnnA

    OnnA Ancient Guru

    Messages:
    17,852
    Likes Received:
    6,741
    GPU:
    TiTan RTX Ampere UV
    Unless they do Sole RT Core Chiplet (to handle RT+De-noise)
    So we will have 2 Chips that can do different jobs - it can be done IMhO.
     

Share This Page