Announcement

Collapse
No announcement yet.

Intel Launches The Core i7 5960X, Mighty Powerful Haswell-E CPUs

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

  • #21
    Originally posted by nerdopolis View Post
    I would probably have to include the 5GB of sample data to compress with the test if I was to do so?
    Sure, but you could take advantage of other tests. For example the kernel build test, there you have real-world data and a tree full of object files.

    Comment


    • #22
      Not exactly irrelevent for multitheaded

      Originally posted by Delgarde View Post
      It doesn't need to be 4x more powerful to justify it. If someone simply needs the fastest cores they can get their hands on, it's irrelevant that AMD has a better price/performance ratio if AMD can't provide the desired performance.
      It's not irrelevant if for the same price I can build a multinode AMD cluster that outperforms the Intel chip, so long as the job in question is fully multithreaded and scales very well, such as libx264 video rendering (my main CPU intensive task).

      Comment


      • #23
        Originally posted by Luke View Post
        It's not irrelevant if for the same price I can build a multinode AMD cluster that outperforms the Intel chip, so long as the job in question is fully multithreaded and scales very well, such as libx264 video rendering (my main CPU intensive task).
        Sure, but that "so long as" you mention is a pretty big qualifier. If you have a task that's infinitely parallelizable and doesn't require much communication between instances, then a cluster is ideal. On the other hand, if you have a task that's mostly single-threaded or which requires tight interaction between threads, clusters *are* irrelevant, because they don't solve the problem.

        Comment


        • #24
          If the job isn't multithreaded or doesn't scale well

          Originally posted by Delgarde View Post
          Sure, but that "so long as" you mention is a pretty big qualifier. If you have a task that's infinitely parallelizable and doesn't require much communication between instances, then a cluster is ideal. On the other hand, if you have a task that's mostly single-threaded or which requires tight interaction between threads, clusters *are* irrelevant, because they don't solve the problem.
          If the job isn't multithreaded or doesn't scale well, you don't need an 8-core processer either. Neither cluster nor $1K Intel proc will outperform a much cheaper chip with few cores but high clocks in that role. Bulldozer will KILL a dual-core Intel for video editing, but for gaming the reverse is true from what I hear.

          I've sometimes thought I should disable all but one Bulldozer module in BIOS for normal websurfing, video playback, and gaming (0ad doesn't make effective use of multicore) to save power, then re-enable all cores when editing video. This would be to save power, but having to remember to go back into setup each time would be a pain. These procs are supposed to "park" unused cores, but I've never observed that in Linux and never tested Windows at all as I don't have or use it. I would like to find a way to force-park and unpark cores from the OS.

          Comment


          • #25
            Love to see what DDR 4 does for this CPU!

            I haven't even looked to see what availability is like with DDR 4 but it would be very interesting to see what DDR 4 can do for this platform. Obviously you would need a test suite that is very good at loading up all of the cores in this chip. The other problem is that I don't know if the initial supply of DDR 4 RAM is faste enough to make a difference.

            However that is what Phoronix is all about, testing and verification.

            Comment

            Working...
            X