eGPU Results - Clevo P870DM-G + GTX1080 + TB3 HDK / PE4C v4.1 (M.2 NGFF)

Discussion in 'e-GPU (External Graphics) Discussion' started by bloodhawk, Oct 5, 2016.

  1. bloodhawk

    bloodhawk Derailer of threads.

    Reputations:
    2,487
    Messages:
    5,563
    Likes Received:
    7,807
    Trophy Points:
    681
    System :

    17" Clevo P870DM-G
    i7-6700K CPU
    64GB RAM GSkill 2800Mhz?
    512GB Samsung 950 Pro
    GTX980 dGPU
    Win10



    eGPU Setup(s) :

    PE4C V4.1

    Inxtron TB3 HDK

    Asus Strix GTX1080 OC


    [​IMG]
    [​IMG]
    [​IMG]


    TB3 Results :


    [​IMG] [​IMG]



    http://www.3dmark.com/3dm/15193649 -Standard, GPU = 18536

    http://www.3dmark.com/3dm/15191071 - Ultra, GPU = 10007

    http://www.3dmark.com/3dm/15193921 - Extreme, GPU = 5206

    http://www.3dmark.com/3dm/15193762 - TimeSpy, GPU = 6855



    [​IMG]
    [​IMG]


    [​IMG]
    [​IMG]
    [​IMG]





    PE4C V4.1 - M.2 NGFF -




    With the cables routed a bit more effectively :
    http://imgur.com/a/Z0mx6
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    http://imgur.com/a/BtVwJ

    [​IMG]

    CUDA-Z :

    CUDA-Z Report
    =============
    Version: 0.10.251 64 bit http://cuda-z.sf.net/
    OS Version: Windows x86 6.2.9200
    Driver Version: 372.70
    Driver Dll Version: 8.0 (6.14.13.7270)
    Runtime Dll Version: 6.50

    Core Information
    ----------------
    Name: GeForce GTX 1080
    Compute Capability: 6.1
    Clock Rate: 1809.5 MHz
    PCI Location: 0:5:0
    Multiprocessors: 20
    Threads Per Multiproc.: 2048
    Warp Size: 32
    Regs Per Block: 65536
    Threads Per Block: 1024
    Threads Dimensions: 1024 x 1024 x 64
    Grid Dimensions: 2147483647 x 65535 x 65535
    Watchdog Enabled: Yes
    Integrated GPU: No
    Concurrent Kernels: Yes
    Compute Mode: Default
    Stream Priorities: Yes

    Memory Information
    ------------------
    Total Global: 8192 MiB
    Bus Width: 256 bits
    Clock Rate: 5005 MHz
    Error Correction: No
    L2 Cache Size: 48 KiB
    Shared Per Block: 48 KiB
    Pitch: 2048 MiB
    Total Constant: 64 KiB
    Texture Alignment: 512 B
    Texture 1D Size: 131072
    Texture 2D Size: 131072 x 65536
    Texture 3D Size: 16384 x 16384 x 16384
    GPU Overlap: Yes
    Map Host Memory: Yes
    Unified Addressing: Yes
    Async Engine: Yes, Bidirectional

    Performance Information
    -----------------------
    Memory Copy
    Host Pinned to Device: 2842.01 MiB/s
    Host Pageable to Device: 2414.87 MiB/s
    Device to Host Pinned: 2981.8 MiB/s
    Device to Host Pageable: 2559.01 MiB/s
    Device to Device: 111.991 GiB/s
    GPU Core Performance
    Single-precision Float: 9246.7 Gflop/s
    Double-precision Float: 295.548 Gflop/s
    64-bit Integer: 465.641 Giop/s
    32-bit Integer: 2742.09 Giop/s
    24-bit Integer: 2084.85 Giop/s

    Generated: Tue Oct 04 23:08:26 2016


    http://www.3dmark.com/fs/10425255 - Standard, GPU = 23143

    http://www.3dmark.com/3dm/15359522 - Extreme, GPU = 10938

    http://www.3dmark.com/3dm/15250559 - Ultra, GPU = 5451
    http://www.3dmark.com/fs/10398914 - Ultra, GPU = 5635

    http://www.3dmark.com/3dm/15360121 - TimeSpy, GPU = 7672




    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]


    One of the advantages of the PE4C implementation is that i can still use the 980 as the dedicated card for the 4k Internal panel or as a CUDA card or just disable it.
    Also unlike the TB3 setup, where i needed to turn the PSU off manually to have the TB3 device shutdown, the PE4C shuts down the GPU as soon as the system shuts down.







    Post your questions peoples. Let me know if you need me to run any specific tests. But please be reasonable, im at work all day so have very limited amount of time once im back.


    @Mr. Fox @Prema @Phoenix (<-Dont complain, thought it would be cool to show it to you :p) @Papusan @Johnksss@iBUYPOWER @Meaker@Sager @Mobius 1 @TBoneSan @D2 Ultima @Q937 @ssj92
     
    Last edited: Oct 29, 2016
  2. bloodhawk

    bloodhawk Derailer of threads.

    Reputations:
    2,487
    Messages:
    5,563
    Likes Received:
    7,807
    Trophy Points:
    681
    Reserved
     
  3. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    2,592
    Messages:
    9,699
    Likes Received:
    6,160
    Trophy Points:
    681
    ha, we were actually thinking about that exact setup @M.2 in order to make our Batmans eGPU capable :) very interesting to see that this actually gives better results than the TB3 enclosure. could be that the M.2 lanes are maybe more independent from other components than the TB3 port?

    on the other hand, not a very ergonomic setup, to say the least @Keyboard half removed :D

    since youre still able to use the internal 980, heres an interesting idea: asynchronous SLI via DX12 with external 1080 :D

    in any case, +rep!
     
    Dr. AMK, Ashtrix, ajc9988 and 3 others like this.
  4. bloodhawk

    bloodhawk Derailer of threads.

    Reputations:
    2,487
    Messages:
    5,563
    Likes Received:
    7,807
    Trophy Points:
    681
    nando seems to agree with. And the m.2 ports are connected directly, so no TB3 to PCIe conversion overheads. (https://www.techinferno.com/index.p...ffm2-pe4c-v41-win10-bloodhawk/#comment-150544)

    That is definitely one of the drawbacks. For now ;). But on our systems, either its a $1300 MXM card or living with the keyboard open, with a < $800 setup. Definitely not for most people though. Might be really helpful for the P775 / P770 folks.
    These speeds made me wonder, if the Proprietary AGA connector is just a fancy looking M.2 port extension. Id actually be willing to buy one and tear it down if someone can get me the service manual with the pin outs. Maybe i can just use their connectors and make the AGA usable on other systems with the M.2 slots. Or even use those connectors to make an external port on our systems :)

    I didnt know about asynchronous SLI! That is definitely interesting. Will look into it, whats the performance bump like ?
     
    Dr. AMK, Ashtrix, Phoenix and 3 others like this.
  5. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    2,592
    Messages:
    9,699
    Likes Received:
    6,160
    Trophy Points:
    681
  6. triturbo

    triturbo Long live 16:10 and MXM-B

    Reputations:
    1,553
    Messages:
    3,644
    Likes Received:
    1,141
    Trophy Points:
    231
    I doubt that we'll see a schematic of this thing, or if we do, it wont be soon. The best way is to get one and reverse-engineer it. After all, the PCIe pins on the GPU side are not a secret :) Multimeter is your best friend.

    As for the TB3 vs M.2, I'm pretty certain that it is indeed the overhead. TB3 is a multi-purpose port, where the M.2's PCIe is just that - PCIe communicating with PCIe devices.
     
  7. bloodhawk

    bloodhawk Derailer of threads.

    Reputations:
    2,487
    Messages:
    5,563
    Likes Received:
    7,807
    Trophy Points:
    681
    Hmm. I'll probably pick one up next week. Let's see how it goes.
     
    Dr. AMK, Phoenix and triturbo like this.
  8. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,744
    Trophy Points:
    581
    Just a reminder, the ZM is PCIe 2.0x4 (if memory serves), meaning we may take a hit whereas the DM will be able to recreate this with relative ease with the bottom open and no keyboard issue. Throw an AIO cooler on the card, maybe better fans on the cooler, and do some good works...
     
  9. jaybee83

    jaybee83 Biotech-Doc

    Reputations:
    2,592
    Messages:
    9,699
    Likes Received:
    6,160
    Trophy Points:
    681
    true, but then again that would be our best bet for an egpu setup, seeing as were lacking TB3. either that or go directly through the mxm slot.

    Sent from my Huawei Mate 8 NXT-AL10
     
  10. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,009
    Messages:
    4,291
    Likes Received:
    5,744
    Trophy Points:
    581
    I definitely agree. My point is more that which card we pick might have to be based on the bandwidth limitations of 4 lane PCIe2.0, and if you have an NVMe drive, forget about it (said with NY accent). Meanwhile, the MXM would allow support for any card, just about...
     
    Dr. AMK, jaybee83 and bloodhawk like this.
Loading...

Share This Page