Torrent Invites! Buy, Trade, Sell Or Find Free Invites, For EVERY Private Tracker! HDBits.org, BTN, PTP, MTV, Empornium, Orpheus, Bibliotik, RED, IPT, TL, PHD etc!



Results 1 to 1 of 1
Like Tree2Likes
  • 2 Post By sedna

Thread: The Bridge for Titan V SLI is gonna cost you $600, up to $7,196 for 2-way SLI

  1. #1
    sedna
    Guest sedna's Avatar

    The Bridge for Titan V SLI is gonna cost you $600, up to $7,196 for 2-way SLI

    Check update on this article here

    What would it cost to own an NVIDIA TITAN V? USD $2,999. Two cards in SLI? $5,998? Think again. NVIDIA expects a goat sacrifice of $6,597 (or $7,196, we'll tell you why), because the company launched yet another multi-GPU bridge cable, which costs $599! With its "Volta" graphics architecture, NVIDIA is bringing its NVLink high-bandwidth interconnect to the client-segment. Since the TITAN V is essentially a client-graphics implementation of Tesla V100 add-on card, with higher clocks, display I/O and a desktop form-factor thrown in, it retains the NVLink interface. You'll notice the broad NVLink finger that's about two-thirds the length of a PCIe x16 interface; the TITAN V PCB has two of those.

    On its online store, NVIDIA is selling the NVLink bridge for a staggering $599, right next to the $30 SLI-HB bridge that made its debut with the "Pascal" architecture. The NVLink bridge was originally sold for users of Quadro GP100. That's not all, the TITAN V PCB has two such NVLink interfaces, and on the Quadro platform, NVIDIA recommends two of these bridges "for the best performance." It's not entirely known if you'll require one or two of these bridges for SLI, or if it's cable-less. Beginning with the TITAN Xp, NVIDIA positioned its TITAN line of graphics cards in the gray area between GeForce, Quadro, and Tesla; with the quite a bit of HPC muscle and professional-graphics features, exemplified with the company choosing to leave Tensor cores enabled (accelerates neural net building). NVLink could hence, have several HPC and deep-learning acceleration applications, where you're scaling up your neural net across two GPUs. It's unlikely that you'll need 200 GB/s of bandwidth for multi-GPU, even with 3x UHD surround displays.

    Pic
    Pic2
    Last edited by ; 12-08-2017 at 03:06 PM.
    jimmy7 and kirill like this.


Tags for this Thread

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
  •