Real benefits of HBM?

Discussion in 'Videocards - AMD Radeon' started by Screwdriver, Aug 28, 2015.

Thread Status:
Not open for further replies.
  1. eclap

    eclap Banned

    Messages:
    31,495
    Likes Received:
    4
    GPU:
    Palit GR 1080 2000/11000
    According to many on this forum, 4gb is a bottleneck. These vrammongers have mostly been banned though for spewing vram threads.

    4gb is enough nowadays, so is fast gddr5. That's why nvidia is holding out for HBM2. There's really no need to go HBM atm, when gddr5 is fast enough. By the time HBM2 hits the market, it might have some use. Might.
     
    Last edited: Aug 28, 2015
  2. Screwdriver

    Screwdriver Banned

    Messages:
    120
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury Tri-X
    I read AMD will get first dibs on Hynix HBM2...Nvidia will have to stand in line.
     
  3. Barry J

    Barry J Ancient Guru

    Messages:
    2,790
    Likes Received:
    133
    GPU:
    RTX2080 TRIO Super
    I detect a fanboy I really dislike closed opinions due to brand loyalty. We as gamers need both company's as they push each other to develop better hardware AMDs HBM is awesome who ever uses HBM2 first will be awesome as it will make the other manufacturer work harder be it NVidia or AMD.
    And we will get better hardware
     
  4. eclap

    eclap Banned

    Messages:
    31,495
    Likes Received:
    4
    GPU:
    Palit GR 1080 2000/11000
    Fine by me, I don't really care who gets it first, HBM will never be the selling point to me when HBM2 launches.
     

  5. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,026
    Likes Received:
    1,322
    GPU:
    2070 Super
    Assuming thats true...

    That would mean that capacity will be very limited, production expensive and AMD would get screwed once again; same like they did with 20nm cancellation,
    same like they did Fury and its boutique supplies.

    Remember that definition of insanity?
     
  6. Screwdriver

    Screwdriver Banned

    Messages:
    120
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury Tri-X

    I bet it will be if Nvidia starts using it.....
     
  7. -Tj-

    -Tj- Ancient Guru

    Messages:
    17,208
    Likes Received:
    1,941
    GPU:
    Zotac GTX980Ti OC
    HBM2 is only relevant because it will be possible to use more then 4GB vram, e.g. 8GB and 16GB variants.

    I really doubt nv is holding back for HMB2 because HBM1 its only 4GB vram like some of you nv fans claim.
    They're holding back because its only Pascal that will be the first gpu using HBM, so actually now they're playing catch up again.


    AMD's Fiji & HBM1 already uses texture compression so that 4GB limit doesn't cripple it much. Anything extra is just a bonus and mostly meant for very high resolutions like 4K, surround, etc. And high MSAA values, gpu<>vram communication will crunch the data faster @ 600GB/s then 384bit 350- 380GB/s, imo its mostly a latency thing.
     
    Last edited: Aug 29, 2015
  8. BLEH!

    BLEH! Ancient Guru

    Messages:
    6,107
    Likes Received:
    170
    GPU:
    Sapphire Fury
    Power efficiency, bandwidth, smaller footprint.
     
  9. eclap

    eclap Banned

    Messages:
    31,495
    Likes Received:
    4
    GPU:
    Palit GR 1080 2000/11000
    You're clearly new around here and don't know what you're talking about. Before you start making assumptions, I'll have you know that I had a 7950 cfx before this 970. So keep your troll posts to yourself.

    No brand loyalty here, live with that.
     
  10. Screwdriver

    Screwdriver Banned

    Messages:
    120
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury Tri-X
    Can't wait to get 29k+ posts to get some respect from you. That is a post a day for 79 years.....10 posts a day for 7.9 years.

    Hmmmm:3eyes:
     
    Last edited: Aug 29, 2015

  11. ---TK---

    ---TK--- Ancient Guru

    Messages:
    22,111
    Likes Received:
    2
    GPU:
    2x 980Ti Gaming 1430/7296
    The way your going you won`t hit 500 posts, before a custom title is under your name.
     
  12. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,867
    Likes Received:
    735
    GPU:
    Inno3D RTX 3090
    AMD was one of the primary (if not the primary) developers of HBM as a tech. There are rumors that they have supply agreements with Hynix (hey, look at all that cheap Hynix GDDR5 that found its way into the 300 series). On top of that, they have supply agreements and patent exchanges with both Global Foundries and IBM. It is NVIDIA who has to basically wait in line in TSMC to get chips out, and yes, no matter the money they make, they are still peanuts compared to other TSMC clients like Apple or Qualcomm. I love how everybody in the thread glossed over the results that people get in 3DMark by overclocking their Furies, and just kept repeating the same 300GB/sec is enough for everybody mantra. HBM2 will have double the speed per stack, and we're gonna see cards with 1TB/sec bandwidths. I wonder why both AMD and NVIDIA would bring that if it didn't matter :infinity:

    We tend to kinda forget also that memory controllers peak too, it's not only the VRAM that does. So you might be connecting ridiculously overclocked VRAM to a controller that literally doesn't know what to do with it. Furies (and I'm sure Pascal too) seem to know what to do with it. As for the whole "only HBM2 matters" thing I've seen, it is completely ridiculous. Wait a tiiiny bit more for some more DX12 results guys, it might matter a bit more then.
     
  13. eclap

    eclap Banned

    Messages:
    31,495
    Likes Received:
    4
    GPU:
    Palit GR 1080 2000/11000
    Post count has nothing to do with respect. Not sure why you think it does.
     
    Last edited: Aug 29, 2015
  14. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,026
    Likes Received:
    1,322
    GPU:
    2070 Super

    according to Wcfftch........
    No offense I trust Wfcctech like I trust "1ms response time" tags on LCD boxes.
    There are more credible rumors about this one guy walking on water :)

    [​IMG]

    J/kKK :D Will AMD get preferential treatment with HBM/2?
    They should get something for time and money sank in. How about better prices?
    I doubt that preferential treatment amounts to 80% market leader "waiting in line", if there are enough chips to go around.

    And if there aren't enough chips, that's bad news for everyone.
    More so for AMD who needs HBM/2 to even maintain some resemblance of equilibrium with NV.

    HMB2 is a great equalizer.
    If by then AMD did not improve their arch enough to close the gap - it's GG.
    Then again can it get worse than 82% : 18% ...
     
  15. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,867
    Likes Received:
    735
    GPU:
    Inno3D RTX 3090
    It might be according to that website, but it actually makes sense. They own a sh*tload of patents for HBM (most of the important ones actually), and that gives them leverage. Also NVIDIA only has TSMC as a manufacturing partner, with all of what that entails (like waiting behind Apple and Qualcomm if there is not enough capacity). Unless the fab dynamics change, I sincerely hope that both companies will have fab capacity to do what they must.
     

  16. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,026
    Likes Received:
    1,322
    GPU:
    2070 Super
  17. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,867
    Likes Received:
    735
    GPU:
    Inno3D RTX 3090
  18. Noisiv

    Noisiv Ancient Guru

    Messages:
    8,026
    Likes Received:
    1,322
    GPU:
    2070 Super
    Yes, right now TSMC is safe regarding NV/AMD gpus.

    They've lost Qualcomm though
     
  19. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,867
    Likes Received:
    735
    GPU:
    Inno3D RTX 3090
    I didn't know that. Also I'm not so sure that the next AMD GPU will be made there and not in Global Foundries.
     
  20. Screwdriver

    Screwdriver Banned

    Messages:
    120
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury Tri-X
    Maybe if I go into the Nvidia forum and argue...

    I started this thread specifically about HBM and the Fiji chip.....but you and others troll here and want to push pro Nvidia stuff.
     
Thread Status:
Not open for further replies.

Share This Page