Announcement

Collapse
No announcement yet.

AMD Radeon "Aldebaran" GPU Support Published For Next-Gen CDNA

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #51
    Originally posted by AlB80 View Post
    Full rate FP64 has the same rate (per byte) as full rate FP32? O rly? Dumb correspondent has been detected.
    Let's please try to keep this civil and maybe we can at least understand each other, if not also learn something.

    Now, I'm just speculating, but if pal666 was talking about GPUs with a 2:1 32:64 ratio, then he's right that because a fp32 is half as big as a fp64, it works out to the same data rate. However, that feels a bit like restating the obvious, so I suspect I'm misunderstanding either the statement or the intent (or both).

    If we just back up to my earlier point that fp64 multipliers aren't free (a point which @AIB80 seems to have acknowledged), then is it so hard to consider that maybe the reason they never previously went for 1:1 32:64 is that they were always including a market (i.e. interactive graphics) that prized fp32 performance? And with CDNA, that's no longer the case. So, it seems natural that they would revisit their silicon apportioning and perhaps even their ABI to target the fp64-heavy workloads that are so important to HPC customers.

    As I suggested before, maybe the new chip is just a natively 64-bit architecture (i.e. each vector lane is 64-bits, instead of 64-bit ops having to use pairs of 32-bit registers). In that case, perhaps they decided it wasn't worth bothering to support packed 2-in-1 fp32, they way they previously offered fp16 support (i.e. since vega10).

    Anyway, we can just wait and see. I'm not about to buy a CDNA card, either way. By the time I'm ready to replace my Radeon VII, perhaps even RDNA cards will be outside my price range. Inflation in both CPU and GPU pricing has really gotten out of hand!
    Last edited by coder; 28 February 2021, 04:16 AM.

    Comment


    • #52
      Originally posted by coder View Post
      Anyway, we can just wait and see. I'm not about to buy a CDNA card, either way. By the time I'm ready to replace my Radeon VII, perhaps even RDNA cards will be outside my price range. Inflation in both CPU and GPU pricing has really gotten out of hand!
      i watch the news very closely and right now the CPU market is cooling down
      a AMD 5800X right now is 20 euros below the recommended retail price from amd.
      https://geizhals.de/amd-ryzen-7-5800...loc=at&hloc=de

      the 5600X is also near this range right now 30euros over the recommended retail price.
      https://geizhals.de/amd-ryzen-5-5600...loc=at&hloc=de

      this means the time of overpriced desktop CPUs is over.

      even if you buy 16core
      3950X is at 679€
      https://geizhals.de/amd-ryzen-9-3950...loc=at&hloc=de

      only the 5950X is still expensive:
      5950x is at 969€
      https://geizhals.de/amd-ryzen-9-5950...loc=at&hloc=de

      but if you show arround AMD still has cheap 16core cpus who are nice:

      the AMD 2950X is at 590€
      https://www.ebay.de/itm/AMD-2950X-CP...AAAOSwuB1gN04b

      this means if you like multicore performance and you don't care for single core performance you buy the 2950X...

      "outside my price range"

      wrong... the chiplet design will make it very cheap to do lowend and highend with the same silicium die.

      if you do some chiplet design with 4 gpu cors for like 4000€ then the 3 chiplet version is at 3000€ and the 2 chiplet version is at 2000€ and the one with 1 chiplet is at 1000€

      but this is only if you think linear amd could easily price the 4 gpu version 8000€ and the 1 chiplet version 700€...

      but see my writing about 5800X/3950X/2950X CPUs these are all chiplet design and the price is coming down.

      the 2950X is chiplet design to and you get 16cores for 590€ right now.
      Phantom circuit Sequence Reducer Dyslexia

      Comment


      • #53
        Originally posted by AlB80 View Post
        Full rate FP64 has the same rate (per byte) as full rate FP32? O rly?
        moron still can't make distinction between "data availability" and "computation"

        Comment


        • #54
          Originally posted by pal666 View Post
          moron still can't make distinction between "data availability" and "computation"
          we have poor education status everywhere--- they really think if you can store 64bit in two 32bit they really believe it is the same than to compute these numbers.

          yes you can do emulation but the performance is poor. for real full speed computation it needs a lot of transistors.
          Phantom circuit Sequence Reducer Dyslexia

          Comment


          • #55
            Originally posted by tildearrow View Post
            What do you specifically mean? According to a spec sheet the 32-bit floating point rate is still double of the 64-bit rate....
            Or are you talking about integer?
            The spec sheet you quoted is for Arcturus - but the recent code push is for Aldebaran.

            Comment

            Working...
            X