Results 1 to 6 of 6

Thread: Interesting Findings... (Threadripper)

  1. #1
    Xtreme Addict
    Join Date
    Sep 2005
    Location
    Las Vegas, NV
    Posts
    1,771

    Interesting Findings... (Threadripper)

    So, finally picked up a Threadripper.

    Things have just over all been interesting with it.. Build went pretty nice and easy.

    But trying to overclock the thing, appears that just raising the speed of the memory is far better than actually overclocking the cores.

    Baseline 3.4GHz with 3333MHz memory, versus 3.8GHz with 3333MHz memory.

    https://browser.geekbench.com/v4/cpu...seline=4501567

    What am I missing here?

    Mainboard: Asus ROG Zenith Extreme
    CPU: 1950X
    Memory: G-Skill 3333MHz 16-16-16-36 2T
    Cooling: EK 280mm CE Rad, EK TR4 Water Block, D5 Pump.



    Thanks,

    Michael
    Main Rig: Intel Core i7 7700k @ 4.2GHz, 64GB of memory, 512GB m.2 SSD, nVidia GTX1080Ti
    NAS: QNAP TVS-1282, 8 x 4TB WD Golds(Main Storage Pool), 4 x 960GB M4 Crucial (VM Storage) , 2 x 512GB M.2 Caching
    Private Cloud: 4 Nodes (2 x Xeon 5645, 48GB DDR3 ECC/REG, 1 x 1TB HDD, 1 x 960GB SSD/Each)
    Distributed Encoding Cloud: 4 Nodes (2 x Xeon x5690, 24GB DDR3 ECC/REG, 1 x 128GB SSD/Each)
    Feedback
    EBAY:HEAT

  2. #2
    Brilliant Idiot
    Join Date
    Jan 2005
    Location
    Hell on Earth
    Posts
    11,015
    baseline has turbo and 3.8 does not single core will get worse. same with CB.
    heatware chew*
    I've got no strings to hold me down.
    To make me fret, or make me frown.
    I had strings but now I'm free.
    There are no strings on me

  3. #3
    Brilliant Idiot
    Join Date
    Jan 2005
    Location
    Hell on Earth
    Posts
    11,015
    Here I ran this off for you. If I had to guess your system and OC is not stable thus throttling or stability is effecting results.

    Like I said stock single core will be better than OC unless you have a great chip.

    This machine is going to be heavy multicore use so makes more sense to lock in a fixed oc.

    Asrock Taichi
    1950x
    h220x2 (2x120mm mcp30 pump) with a Bykski TR block
    gskill 3400 4x16g C16 @ 3200 C14

    fixed 3.9 OC


    Stock ( the lower the temps the better it auto Oc's turbo/xfr )
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	geekbench 4.jpg 
Views:	477 
Size:	263.3 KB 
ID:	132469   Click image for larger version. 

Name:	geekbench 4 stock.jpg 
Views:	459 
Size:	252.3 KB 
ID:	132470  
    Last edited by chew*; 10-17-2017 at 11:53 AM.
    heatware chew*
    I've got no strings to hold me down.
    To make me fret, or make me frown.
    I had strings but now I'm free.
    There are no strings on me

  4. #4
    Xtreme Addict
    Join Date
    Sep 2005
    Location
    Las Vegas, NV
    Posts
    1,771
    Even my stock scores are pretty low.

    What I've noticed is stock, my voltages are in the 1.5v area, while OC'ed my voltages are lower.

    So is the idea to just leave it stock, and let it do its thing? Seems like your multi-core scores are higher being OC'ed across the board, but stock seems to be lower on that.

    Thanks,
    Main Rig: Intel Core i7 7700k @ 4.2GHz, 64GB of memory, 512GB m.2 SSD, nVidia GTX1080Ti
    NAS: QNAP TVS-1282, 8 x 4TB WD Golds(Main Storage Pool), 4 x 960GB M4 Crucial (VM Storage) , 2 x 512GB M.2 Caching
    Private Cloud: 4 Nodes (2 x Xeon 5645, 48GB DDR3 ECC/REG, 1 x 1TB HDD, 1 x 960GB SSD/Each)
    Distributed Encoding Cloud: 4 Nodes (2 x Xeon x5690, 24GB DDR3 ECC/REG, 1 x 128GB SSD/Each)
    Feedback
    EBAY:HEAT

  5. #5
    Xtreme Addict
    Join Date
    Sep 2005
    Location
    Las Vegas, NV
    Posts
    1,771
    https://browser.geekbench.com/v4/cpu/4514072

    My latest, just ran today... Still seems low compared to what you were able to get. Wondering if this mainboard is wonky, or if the chip is wonky.

    No idea.
    Main Rig: Intel Core i7 7700k @ 4.2GHz, 64GB of memory, 512GB m.2 SSD, nVidia GTX1080Ti
    NAS: QNAP TVS-1282, 8 x 4TB WD Golds(Main Storage Pool), 4 x 960GB M4 Crucial (VM Storage) , 2 x 512GB M.2 Caching
    Private Cloud: 4 Nodes (2 x Xeon 5645, 48GB DDR3 ECC/REG, 1 x 1TB HDD, 1 x 960GB SSD/Each)
    Distributed Encoding Cloud: 4 Nodes (2 x Xeon x5690, 24GB DDR3 ECC/REG, 1 x 128GB SSD/Each)
    Feedback
    EBAY:HEAT

  6. #6
    Brilliant Idiot
    Join Date
    Jan 2005
    Location
    Hell on Earth
    Posts
    11,015
    yah something is definitely wonky with your setup not sure what though. it almost looks like your ram is not running dual channel but you have it all installed in proper slots. Like it has dropped a channel.

    Try rolling back to 3200 maybe? also untuned subtimings will hurt performance. Since your on the ASUS I would suggest trying TheStilts profile loaded in your bios.

    Your welcome to try mine but the Asrock I work with is tuned a tad different likewise I tune the ram a bit different.

    The idea is if temps are decent load is not severe Auto CPU the single core boost will hit 4.2 thus boosting single core scores but the all core boost peak is 3.7 which is why single threaded is better at stock and multithreaded is not better than a 3.8 OC.

    Also keep in mind tdie of 68C load is desired however I checked temps in this particular benchmark and unless you have a really bad mount on your block temps should be no issue for geekbench 4.

    You need about 3950 to 4000 all core to break even on single threaded and have a significant boost in multithreaded.

    All depends on your use.

    I am working on chasing this goal via cooling cpu better so trying various setups till I find the right recipe without breaking the bank.
    Last edited by chew*; 10-18-2017 at 01:33 AM.
    heatware chew*
    I've got no strings to hold me down.
    To make me fret, or make me frown.
    I had strings but now I'm free.
    There are no strings on me

Bookmarks

Bookmarks

Posting Permissions

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