NBR Folding@home Team

Discussion in 'Windows OS and Software' started by DR650SE, Dec 9, 2011.

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

    HopelesslyFaithful Notebook Virtuoso

    Reputations:
    1,552
    Messages:
    3,274
    Likes Received:
    164
    Trophy Points:
    0
    obviously ^^ The difference in transistor count and GLOPS/TFLOPS are huge. You have a 100-200GFLOPS tops CPU verse a 2,000GFLOPS/2TFLOPS GPU
     
  2. AESdecryption

    AESdecryption Notebook Evangelist

    Reputations:
    290
    Messages:
    472
    Likes Received:
    9
    Trophy Points:
    31
    Transistors do not equate to computational power. What matters is how optimized the software is for the hardware and architecture (that's when computations come in). Depending on the task, certain processes work better with the CPU than GPU.
     
  3. HopelesslyFaithful

    HopelesslyFaithful Notebook Virtuoso

    Reputations:
    1,552
    Messages:
    3,274
    Likes Received:
    164
    Trophy Points:
    0
    transistors do play a role -_- my point was the fact that GPUs are also far bigger than a CPU. They are not just better at these computations it is also the fact that they are larger. If you scaled a CPU up to the size of a GPU (intel express card...forget the name) they too would have larger performance but still not the same due to different technologies.

    you entirely missed my point

    Here i will elaborate some more.

    The difference exists for these reasons:

    GPUs are developed to be for more raw power. They have a lot less junk thrown in like a CPU has. (CISC vs semi-RISC) I say semi because all current RISC chips are really CISC chips just with a lot less instructions ^^
    -This is why they excel in this type of use.


    GPUs are also bigger than a CPU. If you scale a CPU to be comparable to a GPU they would closer in computational power but GPU would still lead by a mile in these areas.

    I am sure i am missing others as well but those are the two big ones


    Also if i remember correctly F@H is heavily floating point based and it runs better on AMD cpus....for some reason i remember this being true but i could be mistaken. At least i thought AMD cpus are better at FP :/ my memory escapes me. (mostly basing this off of what i remember form reviews a couple years old from my memory.)
     
  4. bigtonyman

    bigtonyman Desktop Powa!!!

    Reputations:
    2,371
    Messages:
    5,023
    Likes Received:
    253
    Trophy Points:
    251
    Pretty sure the flagship AMD enthusiast chip gets slapped around in folding by the intel processors like they do in almost every other test. AMD just can't compete on the high end anymore versus intel. Their GPU's are an entirely different matter though. :p
     
  5. HopelesslyFaithful

    HopelesslyFaithful Notebook Virtuoso

    Reputations:
    1,552
    Messages:
    3,274
    Likes Received:
    164
    Trophy Points:
    0
    like i said that may have been a long time ago...my ability to remember when things happen is horrible. You ask me when did this happen 90% of the time my date in my head is off. It may have been reviews i was reading 3 years ago for all i know. I just thought AMD was pretty good at floating point...maybe they were -_-
     
  6. AESdecryption

    AESdecryption Notebook Evangelist

    Reputations:
    290
    Messages:
    472
    Likes Received:
    9
    Trophy Points:
    31
    The thing about floating point (single precision) is that it is not so precise as double precision (or better), scientific research needs to be done as accurate as possible. GPUs were originally made to gain a boost from CPU gaming. When rendering games (1990s to early 2000s), you don't need insane accuracy (single precision) to get graphics that can convince the gamer that it is realistic (shaders, lighting, movement, etc). So, GPUs were designed to handle single precision calculations quickly. Lately, recent GPUs have been improving their double precision calculations. For nVidia, they pretty much hid their double precision calculations from public view and boasted their floating precision. This was pretty much true for their consumer (GTX series) level GPUs until their release of the GTX Titan (they started boasting about single and double precision).
     
  7. HopelesslyFaithful

    HopelesslyFaithful Notebook Virtuoso

    Reputations:
    1,552
    Messages:
    3,274
    Likes Received:
    164
    Trophy Points:
    0
    and your point???
     
  8. AESdecryption

    AESdecryption Notebook Evangelist

    Reputations:
    290
    Messages:
    472
    Likes Received:
    9
    Trophy Points:
    31
    My point is that your argument (GPUs have more FLOPs than CPUs) is true for some cases. For folding, there is no problem (which I had also said earlier). Making realtime-simulations on a large scale with highly accurate calculations becomes a different case.
     
  9. AESdecryption

    AESdecryption Notebook Evangelist

    Reputations:
    290
    Messages:
    472
    Likes Received:
    9
    Trophy Points:
    31
    Here's a screenshot of 4x6990s folding, not as much PPD as you'd expect compared to nVidia (there was little improvement over the months for AMD folding).

    [​IMG]
     
  10. HopelesslyFaithful

    HopelesslyFaithful Notebook Virtuoso

    Reputations:
    1,552
    Messages:
    3,274
    Likes Received:
    164
    Trophy Points:
    0
    when did you get 4 of those?
     
Loading...
Thread Status:
Not open for further replies.

Share This Page