1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

NVIDIA Could Tease Its Next-Gen Ampere GPU on 7nm at GTC 19

Discussion in 'Frontpage news' started by Rich_Guy, Mar 15, 2019.

  1. bemaniac

    bemaniac Master Guru

    Messages:
    260
    Likes Received:
    9
    GPU:
    MSI Trio X 2080ti
    Likewise. The VRR UHD screens coming out next month from LG are going to need an AMD GPU though.
     
    Dragam1337 likes this.
  2. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    who told you it requires double the TFLOPS for 4K 60fps? in a lot of cases the 30 FPS lock is because of the CPU! you have developers literally talk about how the CPU's on the consoles today are just awful and limit them mostly, it's not the GPU. with checker board rendering and console optimization you don't need double the TFLOPS.

    remember that next-gen consoles will have a new architecture, it's not a GCN refresh. and ofc 7nm will double density for same area size & power.
     
  3. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    1. I am running 4k myself, with gpu's at 10 TFLOPS each with OC.

    2. The ps4 pro and xbox one x are nearly identical in every way, aside of gpu - and while the xbox one x can do 4k30 native in many titles, the ps4 pro always has to upscale, cause its 4,5 TFLOPS gpu doesn't have the reqruired power to do 4k30 native.

    3. The amount of power reqruired to run something scales linearly with the fps. Twice as high fps reqruires twice as much power.
     
  4. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    PC =/= Consoles. consoles get much better optimization. the GPU horse power on the one X is basically an RX 480! they hit 4K native with that....
     

  5. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    Read the entire post before responding...
     
  6. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    You simply ignore the fact that we still don't know anything about Navi on 7nm. with current GPU tech, no doubt.
    I'm willing to bet that i'm right, the codenames and targets were already leaked,
    Lockhart and Anaconda are legit and developers already confirmed it.


    https://www.windowscentral.com/xbox-anaconda

    https://www.thurrott.com/xbox/19539...conda-anthem-danta-lockhart-maverick-and-roma


    Next gen will target 4K 60 fps, and I'll willing to bet on it. Consoles always show big leaps. the current one X hits 4K 30 fps, what is the next possible step consider we will get from a 7nm console with GDDR6?
     
  7. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    Obviously we don't have a crystal ball, but considering how little progress amd has made in the gpu department in recent years, i highly doubt that they will make any significant improvements. Vega on 7nm is quite underwhelming. There is no way that they will literally double the TFLOPS with navi on 7nm vs vega on 14nm, which is what it would reqruire for them to hit 4k60 native with the consoles.

    I will be generous and guess at a 30% improvement, which would still put an affordable chip they could put into a console well below 10 TFLOPS.

    I am sure you are right that they will target 60 fps, as the cpu's should be more than capable. But native 4k at 60 fps? Not a chance. It would be upscaled from something like 1440p, maybe 1800p. And then all the console plebs can say that they are getting 4k60fps... but they aren't really.
     
  8. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    Yes but that whole console will be cheaper than a RTX 2080(probably even 2070)...... the most popular gpu's are not the 2080 / 2080Ti. sorry.

    RTX exists for a reason, the reason is quite clear.
     
    Last edited: Mar 17, 2019
  9. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    Yes, and a fiat is cheaper than a ferrari... but i am quite sure that they ferrari guys are also quite a bit more happy with their cars ;)

    Yes, RTX exists as a way for nvidia to justify that a massive part of the die is used on RT cores, which would otherwise be unusable for gaming and a complete waste of die space, because nvidia hasn't bothered separately developing gaming gpu's, and instead just give us watered down quadro cards that are really not made for gaming.
    This might have worked in the past, but the needs for gaming and the needs for datacenters etc are becoming so different to oneanother, that they need start developing the architectures separately.
     
  10. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    They can't develop 2 different architectures, it's too expensive. this is why all their AI acceleration cores are found on their consumer graphics.
    while the AI / datacenter markets show good growth, they simply don't make enough money to cover a full architecture.

    this is why AMD is using chiplets. for a higher end product you just add more chiplets. no need for 2 architectures
     

  11. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    They could easily do it if they wanted to, as they make an insane amount of money... they just simply choose not to, as they are ngreedia.
     
  12. Denial

    Denial Ancient Guru

    Messages:
    12,341
    Likes Received:
    1,529
    GPU:
    EVGA 1080Ti
    I hate this argument. It makes it sound like Nvidia execs are just shoving all the money into their pockets - which isn't the case. They've expanded their R&D budget significantly in the last few years, hiring thousands of people and just purchased Mellanox for ~$7 billion. Obviously some of those things aren't directly impacting us gamers as customers but it's not like Jens is just sitting there going "I'm not going to build two architectures and instead give myself a payout of a few hundred million!!!"

    Also Nvidia has developed separate gaming GPUs. GP100 was significantly different than GP102. As far as the RT cores go, I don't think it would make a difference. GV100 has significantly more CUDA cores than Turing yet both architectures perform nearly identically because they are capped by ~250w TDP. So it's not like trading RT/Tensor for CUDA would yield some massive performance improvement.
     
    HandR and fantaskarsef like this.
  13. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    the issue is not TDP, the issue is cost. RT/Tensor cores require a lot of area-size and transistors. that means higher pricing.

    but I agree that Nvidia won't build 2 architectures, instead they try to push the gaming industry into the same place their datacenter/AI customers are.
    that way they benefit from developments in AI accelerations both ways. it makes sense.
     
  14. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    no they can't. if they could they would have. spliting their best engineers between 2 completely different architectures is a disaster
     
  15. H83

    H83 Ancient Guru

    Messages:
    2,717
    Likes Received:
    395
    GPU:
    MSI Duke GTX1080Ti
    I agree with Dragam on this one. They have the resources, both personal and financial, to build different 2 diferent architectures both they choose not to for economical reasons. It´s a choice with it´s advantages and disadvantages like any choice. On the consumer space that choice is backfiring because of the silly prices of their current cards.
     
    Dragam1337 likes this.

  16. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    I think Nvidia's suffering is only beginning, they lost half their value and my guess is that they will never recover. they have decided to fully commit on RTX years ago as it takes that long to an architecture to mature. there's no going back.

    the gaming industry is not going to get easier with streaming services and Intel getting into the Graphics market. APU's will become more potent and the mid-range will be gone for Nvidia.
    their only positive is that AMD fails again in their GPU timeline. imagine RTX low sales and Navi challenging the mid-range.
     
  17. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    if your stock value looks like this in the last 6 months:
    https://********/Y2R96Zr/Untitled.png

    and then you proceed to buy Mellanox for 7B$ which has nothing to do with gaming, literally NOTHING. it clearly indicates what is the main focus for Nvidia right now.
     
  18. H83

    H83 Ancient Guru

    Messages:
    2,717
    Likes Received:
    395
    GPU:
    MSI Duke GTX1080Ti
    I agree that Nvidia is on a bad run right now but i think they are still doing extremely well considering the big picture. They only have a product, GPUs, but they have managed to create several markets for that single product and they are doing well in all of those markets. As for future problems/challenges that happens to all companies but companies like Nvidia are better equipped to deal with them.
     
  19. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    if you lose half your value in 6 months, you're not doing "extremely well". even Intel which is losing market share to AMD is still going strong. why? because they have a strong grasp over certain markets and make multiple products.

    a single product is putting all the eggs in one basket. if that basket drops..... then ya.....
     
  20. Dragam1337

    Dragam1337 Maha Guru

    Messages:
    1,281
    Likes Received:
    603
    GPU:
    1080 Gaming X SLI
    I think it is exactly like that.

    https://www.bloomberg.com/billionaires/profiles/jensen-huang/

    GP102 isn't significantly different from GP100... the chip itself is identical with a few clusters disabled - it just isn't built on the HBM platform like GP100.

    Volta is an older architecture, so obviously it won't perform as well when comparing the amount of cuda cores vs performance. GP102 clocks alot higher, and also has the major advantage of gddr6, which performs better for gaming.
     

Share This Page