1. #331
    Member Follow User Gallery
    Join Date
    Mar 2013
    Location
    BC, Canada
    Posts
    43

    Default

    Quote Originally Posted by PabloV3D View Post
    hi all,

    I was for awile looking for information to decide between 7900x and a threadripper.
    Seems that Zbrush likes INTEL cores better than AMD. at leas for now or there is some kind problem with AMD XFR. (hyperthread)
    I think the issue comes from the fact that AMD is prioritizing core count over per-core performance. Intel's individual cores are some 20-30% faster than AMD's similarly priced chip cores depending on workload, and not everything in most programs is or can be made fully multi-threaded and thus can be held up by too slow of single-core performance, so finding that right balance between single-core performance and core count is important and I think Intel has it right for most content creator workloads, since for highly-parallel workloads, using the GPU is much more preferable, we need the CPU to process those long serial operations.

  2. #332

    Default

    Hi,
    Here's the result I get on 1950x with 32gigs of RAM in 4 slots (3200mhz, timings 14-14-14-34)
    (no manual overclock, just xmp for ram to work on advertised speed)

    Single Thread Timer= 2.461
    Multi Thread Timer= 0.364998


    Multithreading performance= 674%

    cheers
    Rafal
    Last edited by RafalZ; 01-31-18 at 01:40 AM. Reason: additional details

  3. #333
    Member Follow User Gallery
    Join Date
    Aug 2005
    Location
    ARGENTINA
    Posts
    93

    Default

    oh thats better fot the TR.

    ( I edited my post , the multy thread was 0.31) @ 4.3ghz


    @ stock I get
    2.1 single
    0.34 multy

    so is about th same as the threadripper


    my ram 32gb at 3200 mhz CL 16. T: 16 18 18 36





  4. #334
    New Member Follow User Gallery
    Join Date
    Apr 2018
    Posts
    4

    Default Intel Core i7 8700K 3.7 GHz 12MB

    Intel Core i7 8700K 3.7 GHz 12MB

    Single Thread Timer= 1.94
    Multi Thread Timer= 0.54
    Multithreading performance= 353%

    I would say good value for money

  5. #335
    New Member Follow User Gallery
    Join Date
    Jul 2017
    Posts
    25

    Default

    How is the new performance compared to 4r8?
    Do you measure similar values?

  6. #336
    Member Follow User Gallery
    Join Date
    Aug 2005
    Location
    ARGENTINA
    Posts
    93

    Default

    No, Actutaly im getting totally different values in ZB 2018 about 20% worst numbers

    4r8 multy= 3.1

    2018 multy= 3.6

    , no idea why.

  7. #337
    Senior Member Follow User Gallery
    Join Date
    Jan 2012
    Location
    Montreal
    Posts
    101

    Default

    I get pretty much the same result as you in 2018, about 10% performance loss.

  8. #338
    New Member Follow User Gallery
    Join Date
    Jul 2017
    Posts
    25

    Default

    Yeah, I also have a significant performance loss. Any words from the devs about this? Can this be fixed in a future release?

  9. #339
    Member Follow User Gallery
    Join Date
    Mar 2005
    Posts
    39

    Default

    Windows 10
    Single Xeon @ 2.10GHz
    128 GB RAM

    Single Thread Timer = 3.413010
    Multi Thread Timer = 0.542999


    Multithreading performance = 628%


    Multithreading IS recommended for this system.

  10. #340
    Member Follow User Gallery
    Join Date
    Mar 2005
    Posts
    39

    Default

    @eFROSTDouble xeon - Multithreading

    build a new workstation and it looks like Zbrush doesn't work with double CPU's (Xeon e5 v4) ? or is it just me ?
    loads only one of them and I get 50% load in the task manager on heavy loads
    Max threads I can set in preferences is 40 as well
    I have the quite same result as you, but single Xeon.


    @eFROST
    We just encountered something like this with another user who had an incredible number of cores. He set ZBrush to use a max of 16 cores per CPU and said it then worked like a champ.
    I cant confirm this, because the Max Threads in ZBrush come back to 32 even if you store your settings at Max Threads 16

  11. #341
    New Member Follow User Gallery
    Join Date
    May 2018
    Posts
    1

    Default

    i7 7700k 5.0ghz + 16GB DDR4 3200mhz

    Single Thread Timer= 1.785
    Multi Thread Timer= 0.575
    Multithreading performance= 310%

  12. #342
    New Member Follow User Gallery
    Join Date
    May 2018
    Posts
    1

    Default

    "I think the issue comes from the fact that AMD is prioritizing core count over per-core performance. Intel's individual cores are some 20-30% faster than AMD's similarly priced chip cores depending on workload, and not everything in most programs is or can be made fully multi-threaded and thus can be held up by too slow of single-core performance, so finding that right balance between single-core performance and core count is important and I think Intel has it right for most content creator workloads, since for highly-parallel workloads, using the GPU is much more preferable, we need the CPU to process those long serial operations." -Invertex

    Please refer to page 214 in the following PDF: http://www.agner.org/optimize/microarchitecture.pdf

    From the linked compiler optimization guide:

    The throughput of each core in the Ryzen is higher than on any previous AMD or Intel x86processor, except for 256-bit vector instructions. Loops that fit into the µop cache can havea throughput of five instructions or six µops per clock cycle. Code that does not fit into theµop cache can have a throughput of four instructions or six µops or approximately 16 bytesof code per clock cycle, whichever is smaller. The 16 bytes fetch rate is a likely bottleneckfor CPU intensive code with large loops......

    ....The very high throughput of the Ryzen core places an extra burden on the programmer andthe compiler if you want optimal performance.

    ----------------

    Based on the fact that a Ryzen core has more instruction level parallelism and can execute more instructions per clock cycle than anything from Intel at this time, and does indeed demonstrate higher throughput per cycle than Intel cores in workloads that have been compiled with Zen architecture optimizations, I must dispute your claims that AMD is prioritizing core count over core performance. On the contrary, a Ryzen 7 CPU has significantly more execution resources than an i7-8700K.

    The "gap" in performance in Zbrush, is almost certainly related to a lack of programming and compiling optimizations for the Ryzen architecture in Zrbush.

  13. #343
    Senior Member Follow User Gallery
    Join Date
    Mar 2011
    Location
    Schaumburg IL
    Age
    31
    Posts
    382

    Default

    i7 920 @ 4.3 GHz
    (overclocked quad core)



    9% performance loss

    2018 seems to be less fluid for me and others as seen on this thread
    http://www.zbrushcentral.com/showthr...in-ZBrush-2018

  14. #344
    New Member Follow User Gallery
    Join Date
    May 2005
    Posts
    17

    Default

    (Whoops! I already posted in this thread a page back...dang old brain!)

  15. #345
    Senior Member Follow User Gallery
    Join Date
    Mar 2011
    Location
    Schaumburg IL
    Age
    31
    Posts
    382

    Default

    ( I edited my post , the multy thread was 0.31) @ 4.3ghz


    @ stock I get
    2.1 single
    0.34 multy

    so is about th same as the threadripper
    your i9 wins vs the TR, but i'd like to see what score an Overclocked TR would get.

    @Aurick
    CPU influences how smoothly you interact with the mesh as you are sculpting. So the better your CPU, the better your sculpting experience will be as you increase the polygon count. Here, more cores is typically better than faster core speed. An easy way to compare processors is to take the speed and multiply it by the number of cores. Doing this for each CPU will give you a value that you can compare. That with the highest score, wins. This means that a 6 core CPU at 2.5 GHz would have a score of 15 while a 4 core CPU at 3 GHz would only have a score of 12. The former edges out the latter even though each core is slower.
    how is it that an i9 has a better score over the TR? Would a better process to compare cpu's be with multithread timing or with your method? does it matter on low multithread timings when its a hundredth of a unit such as .31 vs .36

Page 23 of 24 FirstFirst ... 142021222324 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •